Tag Archives: machine learning

Crowdsourcing brain research at Princeton University to discover 6 new neuron types

Spritely music!

There were already 1/4M registered players as of May 17, 2018 but I’m sure there’s room for more should you be inspired. A May 17, 2018 Princeton University news release (also on EurekAlert) reveals more about the game and about the neurons,

With the help of a quarter-million video game players, Princeton researchers have created and shared detailed maps of more than 1,000 neurons — and they’re just getting started.

“Working with Eyewirers around the world, we’ve made a digital museum that shows off the intricate beauty of the retina’s neural circuits,” said Sebastian Seung, the Evnin Professor in Neuroscience and a professor of computer science and the Princeton Neuroscience Institute (PNI). The related paper is publishing May 17 [2018] in the journal Cell.

Seung is unveiling the Eyewire Museum, an interactive archive of neurons available to the general public and neuroscientists around the world, including the hundreds of researchers involved in the federal Brain Research through Advancing Innovative Neurotechnologies (BRAIN) Initiative.

“This interactive viewer is a huge asset for these larger collaborations, especially among people who are not physically in the same lab,” said Amy Robinson Sterling, a crowdsourcing specialist with PNI and the executive director of Eyewire, the online gaming platform for the citizen scientists who have created this data set.

“This museum is something like a brain atlas,” said Alexander Bae, a graduate student in electrical engineering and one of four co-first authors on the paper. “Previous brain atlases didn’t have a function where you could visualize by individual cell, or a subset of cells, and interact with them. Another novelty: Not only do we have the morphology of each cell, but we also have the functional data, too.”

The neural maps were developed by Eyewirers, members of an online community of video game players who have devoted hundreds of thousands of hours to painstakingly piecing together these neural cells, using data from a mouse retina gathered in 2009.

Eyewire pairs machine learning with gamers who trace the twisting and branching paths of each neuron. Humans are better at visually identifying the patterns of neurons, so every player’s moves are recorded and checked against each other by advanced players and Eyewire staffers, as well as by software that is improving its own pattern recognition skills.

Since Eyewire’s launch in 2012, more than 265,000 people have signed onto the game, and they’ve collectively colored in more than 10 million 3-D “cubes,” resulting in the mapping of more than 3,000 neural cells, of which about a thousand are displayed in the museum.

Each cube is a tiny subset of a single cell, about 4.5 microns across, so a 10-by-10 block of cubes would be the width of a human hair. Every cell is reviewed by between 5 and 25 gamers before it is accepted into the system as complete.

“Back in the early years it took weeks to finish a single cell,” said Sterling. “Now players complete multiple neurons per day.” The Eyewire user experience stays focused on the larger mission — “For science!” is a common refrain — but it also replicates a typical gaming environment, with achievement badges, a chat feature to connect with other players and technical support, and the ability to unlock privileges with increasing skill. “Our top players are online all the time — easily 30 hours a week,” Sterling said.

Dedicated Eyewirers have also contributed in other ways, including donating the swag that gamers win during competitions and writing program extensions “to make game play more efficient and more fun,” said Sterling, including profile histories, maps of player activity, a top 100 leaderboard and ever-increasing levels of customizability.

“The community has really been the driving force behind why Eyewire has been successful,” Sterling said. “You come in, and you’re not alone. Right now, there are 43 people online. Some of them will be admins from Boston or Princeton, but most are just playing — now it’s 46.”

For science!

With 100 billion neurons linked together via trillions of connections, the brain is immeasurably complex, and neuroscientists are still assembling its “parts list,” said Nicholas Turner, a graduate student in computer science and another of the co-first authors. “If you know what parts make up the machine you’re trying to break apart, you’re set to figure out how it all works,” he said.

The researchers have started by tackling Eyewire-mapped ganglion cells from the retina of a mouse. “The retina doesn’t just sense light,” Seung said. “Neural circuits in the retina perform the first steps of visual perception.”

The retina grows from the same embryonic tissue as the brain, and while much simpler than the brain, it is still surprisingly complex, Turner said. “Hammering out these details is a really valuable effort,” he said, “showing the depth and complexity that exists in circuits that we naively believe are simple.”

The researchers’ fundamental question is identifying exactly how the retina works, said Bae. “In our case, we focus on the structural morphology of the retinal ganglion cells.”

“Why the ganglion cells of the eye?” asked Shang Mu, an associate research scholar in PNI and fellow first author. “Because they’re the connection between the retina and the brain. They’re the only cell class that go back into the brain.” Different types of ganglion cells are known to compute different types of visual features, which is one reason the museum has linked shape to functional data.

Using Eyewire-produced maps of 396 ganglion cells, the researchers in Seung’s lab successfully classified these cells more thoroughly than has ever been done before.

“The number of different cell types was a surprise,” said Mu. “Just a few years ago, people thought there were only 15 to 20 ganglion cell types, but we found more than 35 — we estimate between 35 and 50 types.”

Of those, six appear to be novel, in that the researchers could not find any matching descriptions in a literature search.

A brief scroll through the digital museum reveals just how remarkably flat the neurons are — nearly all of the branching takes place along a two-dimensional plane. Seung’s team discovered that different cells grow along different planes, with some reaching high above the nucleus before branching out, while others spread out close to the nucleus. Their resulting diagrams resemble a rainforest, with ground cover, an understory, a canopy and an emergent layer overtopping the rest.

All of these are subdivisions of the inner plexiform layer, one of the five previously recognized layers of the retina. The researchers also identified a “density conservation principle” that they used to distinguish types of neurons.

One of the biggest surprises of the research project has been the extraordinary richness of the original sample, said Seung. “There’s a little sliver of a mouse retina, and almost 10 years later, we’re still learning things from it.”

Of course, it’s a mouse’s brain that you’ll be examining and while there are differences between a mouse brain and a human brain, mouse brains still provide valuable data as they did in the case of some groundbreaking research published in October 2017. James Hamblin wrote about it in an Oct. 7, 2017 article for The Atlantic (Note: Links have been removed),

 

Scientists Somehow Just Discovered a New System of Vessels in Our Brains

It is unclear what they do—but they likely play a central role in aging and disease.

A transparent model of the brain with a network of vessels filled in
Daniel Reich / National Institute of Neurological Disorders and Stroke

You are now among the first people to see the brain’s lymphatic system. The vessels in the photo above transport fluid that is likely crucial to metabolic and inflammatory processes. Until now, no one knew for sure that they existed.

Doctors practicing today have been taught that there are no lymphatic vessels inside the skull. Those deep-purple vessels were seen for the first time in images published this week by researchers at the U.S. National Institute of Neurological Disorders and Stroke.

In the rest of the body, the lymphatic system collects and drains the fluid that bathes our cells, in the process exporting their waste. It also serves as a conduit for immune cells, which go out into the body looking for adversaries and learning how to distinguish self from other, and then travel back to lymph nodes and organs through lymphatic vessels.

So how was it even conceivable that this process wasn’t happening in our brains?

Reich (Daniel Reich, senior investigator) started his search in 2015, after a major study in Nature reported a similar conduit for lymph in mice. The University of Virginia team wrote at the time, “The discovery of the central-nervous-system lymphatic system may call for a reassessment of basic assumptions in neuroimmunology.” The study was regarded as a potential breakthrough in understanding how neurodegenerative disease is associated with the immune system.

Around the same time, researchers discovered fluid in the brains of mice and humans that would become known as the “glymphatic system.” [emphasis mine] It was described by a team at the University of Rochester in 2015 as not just the brain’s “waste-clearance system,” but as potentially helping fuel the brain by transporting glucose, lipids, amino acids, and neurotransmitters. Although since “the central nervous system completely lacks conventional lymphatic vessels,” the researchers wrote at the time, it remained unclear how this fluid communicated with the rest of the body.

There are occasional references to the idea of a lymphatic system in the brain in historic literature. Two centuries ago, the anatomist Paolo Mascagni made full-body models of the lymphatic system that included the brain, though this was dismissed as an error. [emphases mine]  A historical account in The Lancet in 2003 read: “Mascagni was probably so impressed with the lymphatic system that he saw lymph vessels even where they did not exist—in the brain.”

I couldn’t resist the reference to someone whose work had been dismissed summarily being proved right, eventually, and with the help of mouse brains. Do read Hamblin’s article in its entirety if you have time as these excerpts don’t do it justice.

Getting back to Princeton’s research, here’s their research paper,

Digital museum of retinal ganglion cells with dense anatomy and physiology,” by Alexander Bae, Shang Mu, Jinseop Kim, Nicholas Turner, Ignacio Tartavull, Nico Kemnitz, Chris Jordan, Alex Norton, William Silversmith, Rachel Prentki, Marissa Sorek, Celia David, Devon Jones, Doug Bland, Amy Sterling, Jungman Park, Kevin Briggman, Sebastian Seung and the Eyewirers, was published May 17 in the journal Cell with DOI 10.1016/j.cell.2018.04.040.

The research was supported by the Gatsby Charitable Foundation, National Institute of Health-National Institute of Neurological Disorders and Stroke (U01NS090562 and 5R01NS076467), Defense Advanced Research Projects Agency (HR0011-14-2- 0004), Army Research Office (W911NF-12-1-0594), Intelligence Advanced Research Projects Activity (D16PC00005), KT Corporation, Amazon Web Services Research Grants, Korea Brain Research Institute (2231-415) and Korea National Research Foundation Brain Research Program (2017M3C7A1048086).

This paper is behind a paywall. For the players amongst us, here’s the Eyewire website. Go forth,  play, and, maybe, discover new neurons!

A potpourri of robot/AI stories: killers , kindergarten teachers, a Balenciaga-inspired AI fashion designer, a conversational android, and more

Following on my August 29, 2018 post (Sexbots, sexbot ethics, families, and marriage), I’m following up with a more general piece.

Robots, AI (artificial intelligence), and androids (humanoid robots), the terms can be confusing since there’s a tendency to use them interchangeably. Confession: I do it too, but, not this time. That said, I have multiple news bits.

Killer ‘bots and ethics

The U.S. military is already testing a Modular Advanced Armed Robotic System. Credit: Lance Cpl. Julien Rodarte, U.S. Marine Corps

That is a robot.

For the purposes of this posting, a robot is a piece of hardware which may or may not include an AI system and does not mimic a human or other biological organism such that you might, under circumstances, mistake the robot for a biological organism.

As for what precipitated this feature (in part), it seems there’s been a United Nations meeting in Geneva, Switzerland held from August 27 – 31, 2018 about war and the use of autonomous robots, i.e., robots equipped with AI systems and designed for independent action. BTW, it’s the not first meeting the UN has held on this topic.

Bonnie Docherty, lecturer on law and associate director of armed conflict and civilian protection, international human rights clinic, Harvard Law School, has written an August 21, 2018 essay on The Conversation (also on phys.org) describing the history and the current rules around the conduct of war, as well as, outlining the issues with the military use of autonomous robots (Note: Links have been removed),

When drafting a treaty on the laws of war at the end of the 19th century, diplomats could not foresee the future of weapons development. But they did adopt a legal and moral standard for judging new technology not covered by existing treaty language.

This standard, known as the Martens Clause, has survived generations of international humanitarian law and gained renewed relevance in a world where autonomous weapons are on the brink of making their own determinations about whom to shoot and when. The Martens Clause calls on countries not to use weapons that depart “from the principles of humanity and from the dictates of public conscience.”

I was the lead author of a new report by Human Rights Watch and the Harvard Law School International Human Rights Clinic that explains why fully autonomous weapons would run counter to the principles of humanity and the dictates of public conscience. We found that to comply with the Martens Clause, countries should adopt a treaty banning the development, production and use of these weapons.

Representatives of more than 70 nations will gather from August 27 to 31 [2018] at the United Nations in Geneva to debate how to address the problems with what they call lethal autonomous weapon systems. These countries, which are parties to the Convention on Conventional Weapons, have discussed the issue for five years. My co-authors and I believe it is time they took action and agreed to start negotiating a ban next year.

Docherty elaborates on her points (Note: A link has been removed),

The Martens Clause provides a baseline of protection for civilians and soldiers in the absence of specific treaty law. The clause also sets out a standard for evaluating new situations and technologies that were not previously envisioned.

Fully autonomous weapons, sometimes called “killer robots,” would select and engage targets without meaningful human control. They would be a dangerous step beyond current armed drones because there would be no human in the loop to determine when to fire and at what target. Although fully autonomous weapons do not yet exist, China, Israel, Russia, South Korea, the United Kingdom and the United States are all working to develop them. They argue that the technology would process information faster and keep soldiers off the battlefield.

The possibility that fully autonomous weapons could soon become a reality makes it imperative for those and other countries to apply the Martens Clause and assess whether the technology would offend basic humanity and the public conscience. Our analysis finds that fully autonomous weapons would fail the test on both counts.

I encourage you to read the essay in its entirety and for anyone who thinks the discussion about ethics and killer ‘bots is new or limited to military use, there’s my July 25, 2016 posting about police use of a robot in Dallas, Texas. (I imagine the discussion predates 2016 but that’s the earliest instance I have here.)

Teacher bots

Robots come in many forms and this one is on the humanoid end of the spectum,

Children watch a Keeko robot at the Yiswind Institute of Multicultural Education in Beijing, where the intelligent machines are telling stories and challenging kids with logic problems  [donwloaded from https://phys.org/news/2018-08-robot-teachers-invade-chinese-kindergartens.html]

Don’t those ‘eyes’ look almost heart-shaped? No wonder the kids love these robots, if an August  29, 2018 news item on phys.org can be believed,

The Chinese kindergarten children giggled as they worked to solve puzzles assigned by their new teaching assistant: a roundish, short educator with a screen for a face.

Just under 60 centimetres (two feet) high, the autonomous robot named Keeko has been a hit in several kindergartens, telling stories and challenging children with logic problems.

Round and white with a tubby body, the armless robot zips around on tiny wheels, its inbuilt cameras doubling up both as navigational sensors and a front-facing camera allowing users to record video journals.

In China, robots are being developed to deliver groceries, provide companionship to the elderly, dispense legal advice and now, as Keeko’s creators hope, join the ranks of educators.

At the Yiswind Institute of Multicultural Education on the outskirts of Beijing, the children have been tasked to help a prince find his way through a desert—by putting together square mats that represent a path taken by the robot—part storytelling and part problem-solving.

Each time they get an answer right, the device reacts with delight, its face flashing heart-shaped eyes.

“Education today is no longer a one-way street, where the teacher teaches and students just learn,” said Candy Xiong, a teacher trained in early childhood education who now works with Keeko Robot Xiamen Technology as a trainer.

“When children see Keeko with its round head and body, it looks adorable and children love it. So when they see Keeko, they almost instantly take to it,” she added.

Keeko robots have entered more than 600 kindergartens across the country with its makers hoping to expand into Greater China and Southeast Asia.

Beijing has invested money and manpower in developing artificial intelligence as part of its “Made in China 2025” plan, with a Chinese firm last year unveiling the country’s first human-like robot that can hold simple conversations and make facial expressions.

According to the International Federation of Robots, China has the world’s top industrial robot stock, with some 340,000 units in factories across the country engaged in manufacturing and the automotive industry.

Moving on from hardware/software to a software only story.

AI fashion designer better than Balenciaga?

Despite the title for Katharine Schwab’s August 22, 2018 article for Fast Company, I don’t think this AI designer is better than Balenciaga but from the pictures I’ve seen the designs are as good and it does present some intriguing possibilities courtesy of its neural network (Note: Links have been removed),

The AI, created by researcher Robbie Barat, has created an entire collection based on Balenciaga’s previous styles. There’s a fabulous pink and red gradient jumpsuit that wraps all the way around the model’s feet–like a onesie for fashionistas–paired with a dark slouchy coat. There’s a textural color-blocked dress, paired with aqua-green tights. And for menswear, there’s a multi-colored, shimmery button-up with skinny jeans and mismatched shoes. None of these looks would be out of place on the runway.

To create the styles, Barat collected images of Balenciaga’s designs via the designer’s lookbooks, ad campaigns, runway shows, and online catalog over the last two months, and then used them to train the pix2pix neural net. While some of the images closely resemble humans wearing fashionable clothes, many others are a bit off–some models are missing distinct limbs, and don’t get me started on how creepy [emphasis mine] their faces are. Even if the outfits aren’t quite ready to be fabricated, Barat thinks that designers could potentially use a tool like this to find inspiration. Because it’s not constrained by human taste, style, and history, the AI comes up with designs that may never occur to a person. “I love how the network doesn’t really understand or care about symmetry,” Barat writes on Twitter.

You can see the ‘creepy’ faces and some of the designs here,

Image: Robbie Barat

In contrast to the previous two stories, this all about algorithms, no machinery with independent movement (robot hardware) needed.

Conversational android: Erica

Hiroshi Ishiguro and his lifelike (definitely humanoid) robots have featured here many, many times before. The most recent posting is a March 27, 2017 posting about his and his android’s participation at the 2017 SXSW festival.

His latest work is featured in an August 21, 2018 news news item on ScienceDaily,

We’ve all tried talking with devices, and in some cases they talk back. But, it’s a far cry from having a conversation with a real person.

Now a research team from Kyoto University, Osaka University, and the Advanced Telecommunications Research Institute, or ATR, have significantly upgraded the interaction system for conversational android ERICA, giving her even greater dialog skills.

ERICA is an android created by Hiroshi Ishiguro of Osaka University and ATR, specifically designed for natural conversation through incorporation of human-like facial expressions and gestures. The research team demonstrated the updates during a symposium at the National Museum of Emerging Science in Tokyo.

Here’s the latest conversational android, Erica

Caption: The experimental set up when the subject (left) talks with ERICA (right) Credit: Kyoto University / Kawahara lab

An August 20, 2018 Kyoto University press release on EurekAlert, which originated the news item, offers more details,

When we talk to one another, it’s never a simple back and forward progression of information,” states Tatsuya Kawahara of Kyoto University’s Graduate School of Informatics, and an expert in speech and audio processing.

“Listening is active. We express agreement by nodding or saying ‘uh-huh’ to maintain the momentum of conversation. This is called ‘backchanneling’, and is something we wanted to implement with ERICA.”

The team also focused on developing a system for ‘attentive listening’. This is when a listener asks elaborating questions, or repeats the last word of the speaker’s sentence, allowing for more engaging dialogue.

Deploying a series of distance sensors, facial recognition cameras, and microphone arrays, the team began collecting data on parameters necessary for a fluid dialog between ERICA and a human subject.

“We looked at three qualities when studying backchanneling,” continues Kawahara. “These were: timing — when a response happens; lexical form — what is being said; and prosody, or how the response happens.”

Responses were generated through machine learning using a counseling dialogue corpus, resulting in dramatically improved dialog engagement. Testing in five-minute sessions with a human subject, ERICA demonstrated significantly more dynamic speaking skill, including the use of backchanneling, partial repeats, and statement assessments.

“Making a human-like conversational robot is a major challenge,” states Kawahara. “This project reveals how much complexity there is in listening, which we might consider mundane. We are getting closer to a day where a robot can pass a Total Turing Test.”

Erica seems to have been first introduced publicly in Spring 2017, from an April 2017 Erica: Man Made webpage on The Guardian website,

Erica is 23. She has a beautiful, neutral face and speaks with a synthesised voice. She has a degree of autonomy – but can’t move her hands yet. Hiroshi Ishiguro is her ‘father’ and the bad boy of Japanese robotics. Together they will redefine what it means to be human and reveal that the future is closer than we might think.

Hiroshi Ishiguro and his colleague Dylan Glas are interested in what makes a human. Erica is their latest creation – a semi-autonomous android, the product of the most funded scientific project in Japan. But these men regard themselves as artists more than scientists, and the Erica project – the result of a collaboration between Osaka and Kyoto universities and the Advanced Telecommunications Research Institute International – is a philosophical one as much as technological one.

Erica is interviewed about her hope and dreams – to be able to leave her room and to be able to move her arms and legs. She likes to chat with visitors and has one of the most advanced speech synthesis systems yet developed. Can she be regarded as being alive or as a comparable being to ourselves? Will she help us to understand ourselves and our interactions as humans better?

Erica and her creators are interviewed in the science fiction atmosphere of Ishiguro’s laboratory, and this film asks how we might form close relationships with robots in the future. Ishiguro thinks that for Japanese people especially, everything has a soul, whether human or not. If we don’t understand how human hearts, minds and personalities work, can we truly claim that humans have authenticity that machines don’t?

Ishiguro and Glas want to release Erica and her fellow robots into human society. Soon, Erica may be an essential part of our everyday life, as one of the new children of humanity.

Key credits

  • Director/Editor: Ilinca Calugareanu
  • Producer: Mara Adina
  • Executive producers for the Guardian: Charlie Phillips and Laurence Topham
  • This video is produced in collaboration with the Sundance Institute Short Documentary Fund supported by the John D and Catherine T MacArthur Foundation

You can also view the 14 min. film here.

Artworks generated by an AI system are to be sold at Christie’s auction house

KC Ifeanyi’s August 22, 2018 article for Fast Company may send a chill down some artists’ spines,

For the first time in its 252-year history, Christie’s will auction artwork generated by artificial intelligence.

Created by the French art collective Obvious, “Portrait of Edmond de Belamy” is part of a series of paintings of the fictional Belamy family that was created using a two-part algorithm. …

The portrait is estimated to sell anywhere between $7,000-$10,000, and Obvious says the proceeds will go toward furthering its algorithm.

… Famed collector Nicolas Laugero-Lasserre bought one of Obvious’s Belamy works in February, which could’ve been written off as a novel purchase where the story behind it is worth more than the piece itself. However, with validation from a storied auction house like Christie’s, AI art could shake the contemporary art scene.

“Edmond de Belamy” goes up for auction from October 23-25 [2018].

Jobs safe from automation? Are there any?

Michael Grothaus expresses more optimism about future job markets than I’m feeling in an August 30, 2018 article for Fast Company,

A 2017 McKinsey Global Institute study of 800 occupations across 46 countries found that by 2030, 800 million people will lose their jobs to automation. That’s one-fifth of the global workforce. A further one-third of the global workforce will need to retrain if they want to keep their current jobs as well. And looking at the effects of automation on American jobs alone, researchers from Oxford University found that “47 percent of U.S. workers have a high probability of seeing their jobs automated over the next 20 years.”

The good news is that while the above stats are rightly cause for concern, they also reveal that 53% of American jobs and four-fifths of global jobs are unlikely to be affected by advances in artificial intelligence and robotics. But just what are those fields? I spoke to three experts in artificial intelligence, robotics, and human productivity to get their automation-proof career advice.

Creatives

“Although I believe every single job can, and will, benefit from a level of AI or robotic influence, there are some roles that, in my view, will never be replaced by technology,” says Tom Pickersgill, …

Maintenance foreman

When running a production line, problems and bottlenecks are inevitable–and usually that’s a bad thing. But in this case, those unavoidable issues will save human jobs because their solutions will require human ingenuity, says Mark Williams, head of product at People First, …

Hairdressers

Mat Hunter, director of the Central Research Laboratory, a tech-focused co-working space and accelerator for tech startups, have seen startups trying to create all kinds of new technologies, which has given him insight into just what machines can and can’t pull off. It’s lead him to believe that jobs like the humble hairdresser are safer from automation than those of, says, accountancy.

Therapists and social workers

Another automation-proof career is likely to be one involved in helping people heal the mind, says Pickersgill. “People visit therapists because there is a need for emotional support and guidance. This can only be provided through real human interaction–by someone who can empathize and understand, and who can offer advice based on shared experiences, rather than just data-driven logic.”

Teachers

Teachers are so often the unsung heroes of our society. They are overworked and underpaid–yet charged with one of the most important tasks anyone can have: nurturing the growth of young people. The good news for teachers is that their jobs won’t be going anywhere.

Healthcare workers

Doctors and nurses will also likely never see their jobs taken by automation, says Williams. While automation will no doubt better enhance the treatments provided by doctors and nurses the fact of the matter is that robots aren’t going to outdo healthcare workers’ ability to connect with patients and make them feel understood the way a human can.

Caretakers

While humans might be fine with robots flipping their burgers and artificial intelligence managing their finances, being comfortable with a robot nannying your children or looking after your elderly mother is a much bigger ask. And that’s to say nothing of the fact that even today’s most advanced robots don’t have the physical dexterity to perform the movements and actions carers do every day.

Grothaus does offer a proviso in his conclusion: certain types of jobs are relatively safe until developers learn to replicate qualities such as empathy in robots/AI.

It’s very confusing

There’s so much news about robots, artificial intelligence, androids, and cyborgs that it’s hard to keep up with it let alone attempt to get a feeling for where all this might be headed. When you add the fact that the term robots/artificial inteligence are often used interchangeably and that the distinction between robots/androids/cyborgs is not always clear any attempts to peer into the future become even more challenging.

At this point I content myself with tracking the situation and finding definitions so I can better understand what I’m tracking. Carmen Wong’s August 23, 2018 posting on the Signals blog published by Canada’s Centre for Commercialization of Regenerative Medicine (CCRM) offers some useful definitions in the context of an article about the use of artificial intelligence in the life sciences, particularly in Canada (Note: Links have been removed),

Artificial intelligence (AI). Machine learning. To most people, these are just buzzwords and synonymous. Whether or not we fully understand what both are, they are slowly integrating into our everyday lives. Virtual assistants such as Siri? AI is at work. The personalized ads you see when you are browsing on the web or movie recommendations provided on Netflix? Thank AI for that too.

AI is defined as machines having intelligence that imitates human behaviour such as learning, planning and problem solving. A process used to achieve AI is called machine learning, where a computer uses lots of data to “train” or “teach” itself, without human intervention, to accomplish a pre-determined task. Essentially, the computer keeps on modifying its algorithm based on the information provided to get to the desired goal.

Another term you may have heard of is deep learning. Deep learning is a particular type of machine learning where algorithms are set up like the structure and function of human brains. It is similar to a network of brain cells interconnecting with each other.

Toronto has seen its fair share of media-worthy AI activity. The Government of Canada, Government of Ontario, industry and multiple universities came together in March 2018 to launch the Vector Institute, with the goal of using AI to promote economic growth and improve the lives of Canadians. In May, Samsung opened its AI Centre in the MaRS Discovery District, joining a network of Samsung centres located in California, United Kingdom and Russia.

There has been a boom in AI companies over the past few years, which span a variety of industries. This year’s ranking of the top 100 most promising private AI companies covers 25 fields with cybersecurity, enterprise and robotics being the hot focus areas.

Wong goes on to explore AI deployment in the life sciences and concludes that human scientists and doctors will still be needed although she does note this in closing (Note: A link has been removed),

More importantly, empathy and support from a fellow human being could never be fully replaced by a machine (could it?), but maybe this will change in the future. We will just have to wait and see.

Artificial empathy is the term used in Lisa Morgan’s April 25, 2018 article for Information Week which unfortunately does not include any links to actual projects or researchers working on artificial empathy. Instead, the article is focused on how business interests and marketers would like to see it employed. FWIW, I have found a few references: (1) Artificial empathy Wikipedia essay (look for the references at the end of the essay for more) and (2) this open access article: Towards Artificial Empathy; How Can Artificial Empathy Follow the Developmental Pathway of Natural Empathy? by Minoru Asada.

Please let me know in the comments if you should have an insights on the matter in the comments section of this blog.

D-Wave and the first large-scale quantum simulation of a* topological state of matter

This is all about a local (Burnaby is one of the metro Vancouver municipalities) quantum computing companies, D-Wave Systems. The company has been featured here from time to time. It’s usually about about their quantum technology (they are considered a technology star in local and [I think] other circles) but my March 9, 2018 posting about the SXSW (South by Southwest) festival noted that Bo Ewald, President, D-Wave Systems US, was a member of the ‘Quantum Computing: Science Fiction to Science Fact’ panel.

Now, they’re back making technology announcements like this August 22, 2018 news item on phys.org (Note: Links have been removed),

D-Wave Systems today [August 22, 2018] published a milestone study demonstrating a topological phase transition using its 2048-qubit annealing quantum computer. This complex quantum simulation of materials is a major step toward reducing the need for time-consuming and expensive physical research and development.

The paper, entitled “Observation of topological phenomena in a programmable lattice of 1,800 qubits”, was published in the peer-reviewed journal Nature. This work marks an important advancement in the field and demonstrates again that the fully programmable D-Wave quantum computer can be used as an accurate simulator of quantum systems at a large scale. The methods used in this work could have broad implications in the development of novel materials, realizing Richard Feynman’s original vision of a quantum simulator. This new research comes on the heels of D-Wave’s recent Science paper demonstrating a different type of phase transition in a quantum spin-glass simulation. The two papers together signify the flexibility and versatility of the D-Wave quantum computer in quantum simulation of materials, in addition to other tasks such as optimization and machine learning.

An August 22, 2108 D-Wave Systems news release (also on EurekAlert), which originated the news item, delves further (Note: A link has been removed),

In the early 1970s, theoretical physicists Vadim Berezinskii, J. Michael Kosterlitz and David Thouless predicted a new state of matter characterized by nontrivial topological properties. The work was awarded the Nobel Prize in Physics in 2016. D-Wave researchers demonstrated this phenomenon by programming the D-Wave 2000Q™ system to form a two-dimensional frustrated lattice of artificial spins. The observed topological properties in the simulated system cannot exist without quantum effects and closely agree with theoretical predictions.

“This paper represents a breakthrough in the simulation of physical systems which are otherwise essentially impossible,” said 2016 Nobel laureate Dr. J. Michael Kosterlitz. “The test reproduces most of the expected results, which is a remarkable achievement. This gives hope that future quantum simulators will be able to explore more complex and poorly understood systems so that one can trust the simulation results in quantitative detail as a model of a physical system. I look forward to seeing future applications of this simulation method.”

“The work described in the Nature paper represents a landmark in the field of quantum computation: for the first time, a theoretically predicted state of matter was realized in quantum simulation before being demonstrated in a real magnetic material,” said Dr. Mohammad Amin, chief scientist at D-Wave. “This is a significant step toward reaching the goal of quantum simulation, enabling the study of material properties before making them in the lab, a process that today can be very costly and time consuming.”

“Successfully demonstrating physics of Nobel Prize-winning importance on a D-Wave quantum computer is a significant achievement in and of itself. But in combination with D-Wave’s recent quantum simulation work published in Science, this new research demonstrates the flexibility and programmability of our system to tackle recognized, difficult problems in a variety of areas,” said Vern Brownell, D-Wave CEO.

“D-Wave’s quantum simulation of the Kosterlitz-Thouless transition is an exciting and impactful result. It not only contributes to our understanding of important problems in quantum magnetism, but also demonstrates solving a computationally hard problem with a novel and efficient mapping of the spin system, requiring only a limited number of qubits and opening new possibilities for solving a broader range of applications,” said Dr. John Sarrao, principal associate director for science, technology, and engineering at Los Alamos National Laboratory.

“The ability to demonstrate two very different quantum simulations, as we reported in Science and Nature, using the same quantum processor, illustrates the programmability and flexibility of D-Wave’s quantum computer,” said Dr. Andrew King, principal investigator for this work at D-Wave. “This programmability and flexibility were two key ingredients in Richard Feynman’s original vision of a quantum simulator and open up the possibility of predicting the behavior of more complex engineered quantum systems in the future.”

The achievements presented in Nature and Science join D-Wave’s continued work with world-class customers and partners on real-world prototype applications (“proto-apps”) across a variety of fields. The 70+ proto-apps developed by customers span optimization, machine learning, quantum material science, cybersecurity, and more. Many of the proto-apps’ results show that D-Wave systems are approaching, and sometimes surpassing, conventional computing in terms of performance or solution quality on real problems, at pre-commercial scale. As the power of D-Wave systems and software expands, these proto-apps point to the potential for scaled customer application advantage on quantum computers.

The company has prepared a video describing Richard Feynman’s proposal about quantum computing and celebrating their latest achievement,

Here’s the company’s Youtube video description,

In 1982, Richard Feynman proposed the idea of simulating the quantum physics of complex systems with a programmable quantum computer. In August 2018, his vision was realized when researchers from D-Wave Systems and the Vector Institute demonstrated the simulation of a topological phase transition—the subject of the 2016 Nobel Prize in Physics—in a fully programmable D-Wave 2000Q™ annealing quantum computer. This complex quantum simulation of materials is a major step toward reducing the need for time-consuming and expensive physical research and development.

You may want to check out the comments in response to the video.

Here’s a link to and a citation for the Nature paper,

Observation of topological phenomena in a programmable lattice of 1,800 qubits by Andrew D. King, Juan Carrasquilla, Jack Raymond, Isil Ozfidan, Evgeny Andriyash, Andrew Berkley, Mauricio Reis, Trevor Lanting, Richard Harris, Fabio Altomare, Kelly Boothby, Paul I. Bunyk, Colin Enderud, Alexandre Fréchette, Emile Hoskinson, Nicolas Ladizinsky, Travis Oh, Gabriel Poulin-Lamarre, Christopher Rich, Yuki Sato, Anatoly Yu. Smirnov, Loren J. Swenson, Mark H. Volkmann, Jed Whittaker, Jason Yao, Eric Ladizinsky, Mark W. Johnson, Jeremy Hilton, & Mohammad H. Amin. Nature volume 560, pages456–460 (2018) DOI: https://doi.org/10.1038/s41586-018-0410-x Published 22 August 2018

This paper is behind a paywall but, for those who don’t have access, there is a synopsis here.

For anyone curious about the earlier paper published in July 2018, here’s a link and a citation,

Phase transitions in a programmable quantum spin glass simulator by R. Harris, Y. Sato, A. J. Berkley, M. Reis, F. Altomare, M. H. Amin, K. Boothby, P. Bunyk, C. Deng, C. Enderud, S. Huang, E. Hoskinson, M. W. Johnson, E. Ladizinsky, N. Ladizinsky, T. Lanting, R. Li, T. Medina, R. Molavi, R. Neufeld, T. Oh, I. Pavlov, I. Perminov, G. Poulin-Lamarre, C. Rich, A. Smirnov, L. Swenson, N. Tsai, M. Volkmann, J. Whittaker, J. Yao. Science 13 Jul 2018: Vol. 361, Issue 6398, pp. 162-165 DOI: 10.1126/science.aat2025

This paper too is behind a paywall.

You can find out more about D-Wave here.

*ETA ‘a’ to the post title on February 24, 2021.

Being smart about using artificial intelligence in the field of medicine

Since my August 20, 2018 post featured an opinion piece about the possibly imminent replacement of radiologists with artificial intelligence systems and the latest research about employing them for diagnosing eye diseases, it seems like a good time to examine some of the mythology embedded in the discussion about AI and medicine.

Imperfections in medical AI systems

An August 15, 2018 article for Slate.com by W. Nicholson Price II (who teaches at the University of Michigan School of Law; in addition to his law degree he has a PhD in Biological Sciences from Columbia University) begins with the peppy, optimistic view before veering into more critical territory (Note: Links have been removed),

For millions of people suffering from diabetes, new technology enabled by artificial intelligence promises to make management much easier. Medtronic’s Guardian Connect system promises to alert users 10 to 60 minutes before they hit high or low blood sugar level thresholds, thanks to IBM Watson, “the same supercomputer technology that can predict global weather patterns.” Startup Beta Bionics goes even further: In May, it received Food and Drug Administration approval to start clinical trials on what it calls a “bionic pancreas system” powered by artificial intelligence, capable of “automatically and autonomously managing blood sugar levels 24/7.”

An artificial pancreas powered by artificial intelligence represents a huge step forward for the treatment of diabetes—but getting it right will be hard. Artificial intelligence (also known in various iterations as deep learning and machine learning) promises to automatically learn from patterns in medical data to help us do everything from managing diabetes to finding tumors in an MRI to predicting how long patients will live. But the artificial intelligence techniques involved are typically opaque. We often don’t know how the algorithm makes the eventual decision. And they may change and learn from new data—indeed, that’s a big part of the promise. But when the technology is complicated, opaque, changing, and absolutely vital to the health of a patient, how do we make sure it works as promised?

Price describes how a ‘closed loop’ artificial pancreas with AI would automate insulin levels for diabetic patients, flaws in the automated system, and how companies like to maintain a competitive advantage (Note: Links have been removed),

[…] a “closed loop” artificial pancreas, where software handles the whole issue, receiving and interpreting signals from the monitor, deciding when and how much insulin is needed, and directing the insulin pump to provide the right amount. The first closed-loop system was approved in late 2016. The system should take as much of the issue off the mind of the patient as possible (though, of course, that has limits). Running a close-loop artificial pancreas is challenging. The way people respond to changing levels of carbohydrates is complicated, as is their response to insulin; it’s hard to model accurately. Making it even more complicated, each individual’s body reacts a little differently.

Here’s where artificial intelligence comes into play. Rather than trying explicitly to figure out the exact model for how bodies react to insulin and to carbohydrates, machine learning methods, given a lot of data, can find patterns and make predictions. And existing continuous glucose monitors (and insulin pumps) are excellent at generating a lot of data. The idea is to train artificial intelligence algorithms on vast amounts of data from diabetic patients, and to use the resulting trained algorithms to run a closed-loop artificial pancreas. Even more exciting, because the system will keep measuring blood glucose, it can learn from the new data and each patient’s artificial pancreas can customize itself over time as it acquires new data from that patient’s particular reactions.

Here’s the tough question: How will we know how well the system works? Diabetes software doesn’t exactly have the best track record when it comes to accuracy. A 2015 study found that among smartphone apps for calculating insulin doses, two-thirds of the apps risked giving incorrect results, often substantially so. … And companies like to keep their algorithms proprietary for a competitive advantage, which makes it hard to know how they work and what flaws might have gone unnoticed in the development process.

There’s more,

These issues aren’t unique to diabetes care—other A.I. algorithms will also be complicated, opaque, and maybe kept secret by their developers. The potential for problems multiplies when an algorithm is learning from data from an entire hospital, or hospital system, or the collected data from an entire state or nation, not just a single patient. …

The [US Food and Drug Administraiont] FDA is working on this problem. The head of the agency has expressed his enthusiasm for bringing A.I. safely into medical practice, and the agency has a new Digital Health Innovation Action Plan to try to tackle some of these issues. But they’re not easy, and one thing making it harder is a general desire to keep the algorithmic sauce secret. The example of IBM Watson for Oncology has given the field a bit of a recent black eye—it turns out that the company knew the algorithm gave poor recommendations for cancer treatment but kept that secret for more than a year. …

While Price focuses on problems with algorithms and with developers and their business interests, he also hints at some of the body’s complexities.

Can AI systems be like people?

Susan Baxter, a medical writer with over 20 years experience, a PhD in health economics, and author of countless magazine articles and several books, offers a more person-centered approach to the discussion in her July 6, 2018 posting on susanbaxter.com,

The fascination with AI continues to irk, given that every second thing I read seems to be extolling the magic of AI and medicine and how It Will Change Everything. Which it will not, trust me. The essential issue of illness remains perennial and revolves around an individual for whom no amount of technology will solve anything without human contact. …

But in this world, or so we are told by AI proponents, radiologists will soon be obsolete. [my August 20, 2018 post] The adaptational learning capacities of AI mean that reading a scan or x-ray will soon be more ably done by machines than humans. The presupposition here is that we, the original programmers of this artificial intelligence, understand the vagaries of real life (and real disease) so wonderfully that we can deconstruct these much as we do the game of chess (where, let’s face it, Big Blue ate our lunch) and that analyzing a two-dimensional image of a three-dimensional body, already problematic, can be reduced to a series of algorithms.

Attempting to extrapolate what some “shadow” on a scan might mean in a flesh and blood human isn’t really quite the same as bishop to knight seven. Never mind the false positive/negatives that are considered an acceptable risk or the very real human misery they create.

Moravec called it

It’s called Moravec’s paradox, the inability of humans to realize just how complex basic physical tasks are – and the corresponding inability of AI to mimic it. As you walk across the room, carrying a glass of water, talking to your spouse/friend/cat/child; place the glass on the counter and open the dishwasher door with your foot as you open a jar of pickles at the same time, take a moment to consider just how many concurrent tasks you are doing and just how enormous the computational power these ostensibly simple moves would require.

Researchers in Singapore taught industrial robots to assemble an Ikea chair. Essentially, screw in the legs. A person could probably do this in a minute. Maybe two. The preprogrammed robots took nearly half an hour. And I suspect programming those robots took considerably longer than that.

Ironically, even Elon Musk, who has had major production problems with the Tesla cars rolling out of his high tech factory, has conceded (in a tweet) that “Humans are underrated.”

I wouldn’t necessarily go that far given the political shenanigans of Trump & Co. but in the grand scheme of things I tend to agree. …

Is AI going the way of gene therapy?

Susan draws a parallel between the AI and medicine discussion with the discussion about genetics and medicine (Note: Links have been removed),

On a somewhat similar note – given the extent to which genetics discourse has that same linear, mechanistic  tone [as AI and medicine] – it turns out all this fine talk of using genetics to determine health risk and whatnot is based on nothing more than clever marketing, since a lot of companies are making a lot of money off our belief in DNA. Truth is half the time we don’t even know what a gene is never mind what it actually does;  geneticists still can’t agree on how many genes there are in a human genome, as this article in Nature points out.

Along the same lines, I was most amused to read about something called the Super Seniors Study, research following a group of individuals in their 80’s, 90’s and 100’s who seem to be doing really well. Launched in 2002 and headed by Angela Brooks Wilson, a geneticist at the BC [British Columbia] Cancer Agency and SFU [Simon Fraser University] Chair of biomedical physiology and kinesiology, this longitudinal work is examining possible factors involved in healthy ageing.

Turns out genes had nothing to do with it, the title of the Globe and Mail article notwithstanding. (“Could the DNA of these super seniors hold the secret to healthy aging?” The answer, a resounding “no”, well hidden at the very [end], the part most people wouldn’t even get to.) All of these individuals who were racing about exercising and working part time and living the kind of life that makes one tired just reading about it all had the same “multiple (genetic) factors linked to a high probability of disease”. You know, the gene markers they tell us are “linked” to cancer, heart disease, etc., etc. But these super seniors had all those markers but none of the diseases, demonstrating (pretty strongly) that the so-called genetic links to disease are a load of bunkum. Which (she said modestly) I have been saying for more years than I care to remember. You’re welcome.

The fundamental error in this type of linear thinking is in allowing our metaphors (genes are the “blueprint” of life) and propensity towards social ideas of determinism to overtake common sense. Biological and physiological systems are not static; they respond to and change to life in its entirety, whether it’s diet and nutrition to toxic or traumatic insults. Immunity alters, endocrinology changes, – even how we think and feel affects the efficiency and effectiveness of physiology. Which explains why as we age we become increasingly dissimilar.

If you have the time, I encourage to read Susan’s comments in their entirety.

Scientific certainties

Following on with genetics, gene therapy dreams, and the complexity of biology, the June 19, 2018 Nature article by Cassandra Willyard (mentioned in Susan’s posting) highlights an aspect of scientific research not often mentioned in public,

One of the earliest attempts to estimate the number of genes in the human genome involved tipsy geneticists, a bar in Cold Spring Harbor, New York, and pure guesswork.

That was in 2000, when a draft human genome sequence was still in the works; geneticists were running a sweepstake on how many genes humans have, and wagers ranged from tens of thousands to hundreds of thousands. Almost two decades later, scientists armed with real data still can’t agree on the number — a knowledge gap that they say hampers efforts to spot disease-related mutations.

In 2000, with the genomics community abuzz over the question of how many human genes would be found, Ewan Birney launched the GeneSweep contest. Birney, now co-director of the European Bioinformatics Institute (EBI) in Hinxton, UK, took the first bets at a bar during an annual genetics meeting, and the contest eventually attracted more than 1,000 entries and a US$3,000 jackpot. Bets on the number of genes ranged from more than 312,000 to just under 26,000, with an average of around 40,000. These days, the span of estimates has shrunk — with most now between 19,000 and 22,000 — but there is still disagreement (See ‘Gene Tally’).

… the inconsistencies in the number of genes from database to database are problematic for researchers, Pruitt says. “People want one answer,” she [Kim Pruitt, a genome researcher at the US National Center for Biotechnology Information {NCB}] in Bethesda, Maryland] adds, “but biology is complex.”

I wanted to note that scientists do make guesses and not just with genetics. For example, Gina Mallet’s 2005 book ‘Last Chance to Eat: The Fate of Taste in a Fast Food World’ recounts the story of how good and bad levels of cholesterol were established—the experts made some guesses based on their experience. That said, Willyard’s article details the continuing effort to nail down the number of genes almost 20 years after the human genome project was completed and delves into the problems the scientists have uncovered.

Final comments

In addition to opaque processes with developers/entrepreneurs wanting to maintain their secrets for competitive advantages and in addition to our own poor understanding of the human body (how many genes are there anyway?), there are same major gaps (reflected in AI) in our understanding of various diseases. Angela Lashbrook’s August 16, 2018 article for The Atlantic highlights some issues with skin cancer and shade of your skin (Note: Links have been removed),

… While fair-skinned people are at the highest risk for contracting skin cancer, the mortality rate for African Americans is considerably higher: Their five-year survival rate is 73 percent, compared with 90 percent for white Americans, according to the American Academy of Dermatology.

As the rates of melanoma for all Americans continue a 30-year climb, dermatologists have begun exploring new technologies to try to reverse this deadly trend—including artificial intelligence. There’s been a growing hope in the field that using machine-learning algorithms to diagnose skin cancers and other skin issues could make for more efficient doctor visits and increased, reliable diagnoses. The earliest results are promising—but also potentially dangerous for darker-skinned patients.

… Avery Smith, … a software engineer in Baltimore, Maryland, co-authored a paper in JAMA [Journal of the American Medical Association] Dermatology that warns of the potential racial disparities that could come from relying on machine learning for skin-cancer screenings. Smith’s co-author, Adewole Adamson of the University of Texas at Austin, has conducted multiple studies on demographic imbalances in dermatology. “African Americans have the highest mortality rate [for skin cancer], and doctors aren’t trained on that particular skin type,” Smith told me over the phone. “When I came across the machine-learning software, one of the first things I thought was how it will perform on black people.”

Recently, a study that tested machine-learning software in dermatology, conducted by a group of researchers primarily out of Germany, found that “deep-learning convolutional neural networks,” or CNN, detected potentially cancerous skin lesions better than the 58 dermatologists included in the study group. The data used for the study come from the International Skin Imaging Collaboration, or ISIC, an open-source repository of skin images to be used by machine-learning algorithms. Given the rise in melanoma cases in the United States, a machine-learning algorithm that assists dermatologists in diagnosing skin cancer earlier could conceivably save thousands of lives each year.

… Chief among the prohibitive issues, according to Smith and Adamson, is that the data the CNN relies on come from primarily fair-skinned populations in the United States, Australia, and Europe. If the algorithm is basing most of its knowledge on how skin lesions appear on fair skin, then theoretically, lesions on patients of color are less likely to be diagnosed. “If you don’t teach the algorithm with a diverse set of images, then that algorithm won’t work out in the public that is diverse,” says Adamson. “So there’s risk, then, for people with skin of color to fall through the cracks.”

As Adamson and Smith’s paper points out, racial disparities in artificial intelligence and machine learning are not a new issue. Algorithms have mistaken images of black people for gorillas, misunderstood Asians to be blinking when they weren’t, and “judged” only white people to be attractive. An even more dangerous issue, according to the paper, is that decades of clinical research have focused primarily on people with light skin, leaving out marginalized communities whose symptoms may present differently.

The reasons for this exclusion are complex. According to Andrew Alexis, a dermatologist at Mount Sinai, in New York City, and the director of the Skin of Color Center, compounding factors include a lack of medical professionals from marginalized communities, inadequate information about those communities, and socioeconomic barriers to participating in research. “In the absence of a diverse study population that reflects that of the U.S. population, potential safety or efficacy considerations could be missed,” he says.

Adamson agrees, elaborating that with inadequate data, machine learning could misdiagnose people of color with nonexistent skin cancers—or miss them entirely. But he understands why the field of dermatology would surge ahead without demographically complete data. “Part of the problem is that people are in such a rush. This happens with any new tech, whether it’s a new drug or test. Folks see how it can be useful and they go full steam ahead without thinking of potential clinical consequences. …

Improving machine-learning algorithms is far from the only method to ensure that people with darker skin tones are protected against the sun and receive diagnoses earlier, when many cancers are more survivable. According to the Skin Cancer Foundation, 63 percent of African Americans don’t wear sunscreen; both they and many dermatologists are more likely to delay diagnosis and treatment because of the belief that dark skin is adequate protection from the sun’s harmful rays. And due to racial disparities in access to health care in America, African Americans are less likely to get treatment in time.

Happy endings

I’ll add one thing to Price’s article, Susan’s posting, and Lashbrook’s article about the issues with AI , certainty, gene therapy, and medicine—the desire for a happy ending prefaced with an easy solution. If the easy solution isn’t possible accommodations will be made but that happy ending is a must. All disease will disappear and there will be peace on earth. (Nod to Susan Baxter and her many discussions with me about disease processes and happy endings.)

The solutions, for the most part, are seen as technological despite the mountain of evidence suggesting that technology reflects our own imperfect understanding of health and disease therefore providing what is at best an imperfect solution.

Also, we tend to underestimate just how complex humans are not only in terms of disease and health but also with regard to our skills, understanding, and, perhaps not often enough, our ability to respond appropriately in the moment.

There is much to celebrate in what has been accomplished: no more black death, no more smallpox, hip replacements, pacemakers, organ transplants, and much more. Yes, we should try to improve our medicine. But, maybe alongside the celebration we can welcome AI and other technologies with a lot less hype and a lot more skepticism.

AI x 2: the Amnesty International and Artificial Intelligence story

Amnesty International and artificial intelligence seem like an unexpected combination but it all makes sense when you read a June 13, 2018 article by Steven Melendez for Fast Company (Note: Links have been removed),

If companies working on artificial intelligence don’t take steps to safeguard human rights, “nightmare scenarios” could unfold, warns Rasha Abdul Rahim, an arms control and artificial intelligence researcher at Amnesty International in a blog post. Those scenarios could involve armed, autonomous systems choosing military targets with little human oversight, or discrimination caused by biased algorithms, she warns.

Rahim pointed at recent reports of Google’s involvement in the Pentagon’s Project Maven, which involves harnessing AI image recognition technology to rapidly process photos taken by drones. Google recently unveiled new AI ethics policies and has said it won’t continue with the project once its current contract expires next year after high-profile employee dissent over the project. …

“Compliance with the laws of war requires human judgement [sic] –the ability to analyze the intentions behind actions and make complex decisions about the proportionality or necessity of an attack,” Rahim writes. “Machines and algorithms cannot recreate these human skills, and nor can they negotiate, produce empathy, or respond to unpredictable situations. In light of these risks, Amnesty International and its partners in the Campaign to Stop Killer Robots are calling for a total ban on the development, deployment, and use of fully autonomous weapon systems.”

Rasha Abdul Rahim’s June 14, 2018 posting (I’m putting the discrepancy in publication dates down to timezone differences) on the Amnesty International website (Note: Links have been removed),

Last week [June 7, 2018] Google released a set of principles to govern its development of AI technologies. They include a broad commitment not to design or deploy AI in weaponry, and come in the wake of the company’s announcement that it will not renew its existing contract for Project Maven, the US Department of Defense’s AI initiative, when it expires in 2019.

The fact that Google maintains its existing Project Maven contract for now raises an important question. Does Google consider that continuing to provide AI technology to the US government’s drone programme is in line with its new principles? Project Maven is a litmus test that allows us to see what Google’s new principles mean in practice.

As details of the US drone programme are shrouded in secrecy, it is unclear precisely what role Google plays in Project Maven. What we do know is that US drone programme, under successive administrations, has been beset by credible allegations of unlawful killings and civilian casualties. The cooperation of Google, in any capacity, is extremely troubling and could potentially implicate it in unlawful strikes.

As AI technology advances, the question of who will be held accountable for associated human rights abuses is becoming increasingly urgent. Machine learning, and AI more broadly, impact a range of human rights including privacy, freedom of expression and the right to life. It is partly in the hands of companies like Google to safeguard these rights in relation to their operations – for us and for future generations. If they don’t, some nightmare scenarios could unfold.

Warfare has already changed dramatically in recent years – a couple of decades ago the idea of remote controlled bomber planes would have seemed like science fiction. While the drones currently in use are still controlled by humans, China, France, Israel, Russia, South Korea, the UK and the US are all known to be developing military robots which are getting smaller and more autonomous.

For example, the UK is developing a number of autonomous systems, including the BAE [Systems] Taranis, an unmanned combat aircraft system which can fly in autonomous mode and automatically identify a target within a programmed area. Kalashnikov, the Russian arms manufacturer, is developing a fully automated, high-calibre gun that uses artificial neural networks to choose targets. The US Army Research Laboratory in Maryland, in collaboration with BAE Systems and several academic institutions, has been developing micro drones which weigh less than 30 grams, as well as pocket-sized robots that can hop or crawl.

Of course, it’s not just in conflict zones that AI is threatening human rights. Machine learning is already being used by governments in a wide range of contexts that directly impact people’s lives, including policing [emphasis mine], welfare systems, criminal justice and healthcare. Some US courts use algorithms to predict future behaviour of defendants and determine their sentence lengths accordingly. The potential for this approach to reinforce power structures, discrimination or inequalities is huge.

In july 2017, the Vancouver Police Department announced its use of predictive policing software, the first such jurisdiction in Canada to make use of the technology. My Nov. 23, 2017 posting featured the announcement.

The almost too aptly named Campaign to Stop Killer Robots can be found here. Their About Us page provides a brief history,

Formed by the following non-governmental organizations (NGOs) at a meeting in New York on 19 October 2012 and launched in London in April 2013, the Campaign to Stop Killer Robots is an international coalition working to preemptively ban fully autonomous weapons. See the Chronology charting our major actions and achievements to date.

Steering Committee

The Steering Committee is the campaign’s principal leadership and decision-making body. It is comprised of five international NGOs, a regional NGO network, and four national NGOs that work internationally:

Human Rights Watch
Article 36
Association for Aid and Relief Japan
International Committee for Robot Arms Control
Mines Action Canada
Nobel Women’s Initiative
PAX (formerly known as IKV Pax Christi)
Pugwash Conferences on Science & World Affairs
Seguridad Humana en América Latina y el Caribe (SEHLAC)
Women’s International League for Peace and Freedom

For more information, see this Overview. A Terms of Reference is also available on request, detailing the committee’s selection process, mandate, decision-making, meetings and communication, and expected commitments.

For anyone who may be interested in joining Amnesty International, go here.

May 28, 2018 release date for inorganic material database ‘AtomWork-Adv’

Announced in a May 23, 2018 news item on Nanowerk,

[Japan National Institute for Materials Science] NIMS will make its inorganic materials database, AtomWork-Adv (pronounced “atom work advanced”), available to the general public as a fee-based service starting Monday, May 28, 2018. This service will be provided by the Data Platform from the Center for Materials Research by Information Integration (CMI2), Research and Services Division of the Materials Data and Integrated System (MaDIS), NIMS.

A May 23, 2018 NIMS press release, which originated the news item, fills out the details (Note: Paragraph 1 is largely repetitive but there is contact information in there),

  1. NIMS will make its inorganic materials database, “AtomWork-Adv” (pronounced “atom work advanced”), available to the general public as a fee-based service starting Monday, May 28, 2018. This service will be provided by the Data Platform (Yibin Xu, Director) from the Center for Materials Research by Information Integration (CMI2), Research and Services Division of the Materials Data and Integrated System (MaDIS), NIMS. AtomWork-Adv markedly improves upon the amount of data available and the usability of the current web-based “AtomWork” database. We hope that this service will promote data-driven materials development using AI and machine learning.
  2. The increasingly popular use of AI and machine learning in materials development requires a high-quality materials database. NIMS publicized the AtomWork inorganic materials database—constructed using literature published up to 2002—on its MatNavi webpage (http://mits.nims.go.jp/index.html). While AtomWork is available free of charge, the data it contains is not updated and its functions, such as allowing users to copy, download and search data, are limited.
  3. The AtomWork-Adv database compiles crystal structure data (approx. 274,000 datasets), X-ray diffraction data (approx. 496,000 datasets), material properties data (approx. 298,000 datasets) and phase diagram data (approx. 40,000 datasets) collected from literature published up to 2014. The number of datasets in these categories are about three to five times greater than those in the Atom Work database. In addition, AtomWork-Adv is equipped with user-friendly functions, such as a versatile search tool which enables searches by element, composition, crystal structure and material property, a matrix function which identifies the number of datasets available for a specific binary material combination and an automatic charting function which allows the plotting of a graph between two material property variables and which displays material names in relation to these variables. Users can download data for use in data science-driven materials research and development (the number of downloads is restricted).
  4. Data will be continuously added to and updated in the fee-based AtomWork-Adv database. We are planning to add new data collected from literature between 2015 and 2016 to the database during FY2018.
  5. This database project was supported by the “Materials Research by Information Integration” Initiative (MI2I) sponsored by the Japan Science and Technology Agency (JST)’s Support Program for Starting up Innovation Hub.

Happy atom hunting!

The Hedy Lamarr of international research: Canada’s Third assessment of The State of Science and Technology and Industrial Research and Development in Canada (2 of 2)

Taking up from where I left off with my comments on Competing in a Global Innovation Economy: The Current State of R and D in Canada or as I prefer to call it the Third assessment of Canadas S&T (science and technology) and R&D (research and development). (Part 1 for anyone who missed it).

Is it possible to get past Hedy?

Interestingly (to me anyway), one of our R&D strengths, the visual and performing arts, features sectors where a preponderance of people are dedicated to creating culture in Canada and don’t spend a lot of time trying to make money so they can retire before the age of 40 as so many of our start-up founders do. (Retiring before the age of 40 just reminded me of Hollywood actresses {Hedy] who found and still do find that work was/is hard to come by after that age. You may be able but I’m not sure I can get past Hedy.) Perhaps our business people (start-up founders) could take a leaf out of the visual and performing arts handbook? Or, not. There is another question.

Does it matter if we continue to be a ‘branch plant’ economy? Somebody once posed that question to me when I was grumbling that our start-ups never led to larger businesses and acted more like incubators (which could describe our R&D as well),. He noted that Canadians have a pretty good standard of living and we’ve been running things this way for over a century and it seems to work for us. Is it that bad? I didn’t have an  answer for him then and I don’t have one now but I think it’s a useful question to ask and no one on this (2018) expert panel or the previous expert panel (2013) seems to have asked.

I appreciate that the panel was constrained by the questions given by the government but given how they snuck in a few items that technically speaking were not part of their remit, I’m thinking they might have gone just a bit further. The problem with answering the questions as asked is that if you’ve got the wrong questions, your answers will be garbage (GIGO; garbage in, garbage out) or, as is said, where science is concerned, it’s the quality of your questions.

On that note, I would have liked to know more about the survey of top-cited researchers. I think looking at the questions could have been quite illuminating and I would have liked some information on from where (geographically and area of specialization) they got most of their answers. In keeping with past practice (2012 assessment published in 2013), there is no additional information offered about the survey questions or results. Still, there was this (from the report released April 10, 2018; Note: There may be some difference between the formatting seen here and that seen in the document),

3.1.2 International Perceptions of Canadian Research
As with the 2012 S&T report, the CCA commissioned a survey of top-cited researchers’ perceptions of Canada’s research strength in their field or subfield relative to that of other countries (Section 1.3.2). Researchers were asked to identify the top five countries in their field and subfield of expertise: 36% of respondents (compared with 37% in the 2012 survey) from across all fields of research rated Canada in the top five countries in their field (Figure B.1 and Table B.1 in the appendix). Canada ranks fourth out of all countries, behind the United States, United Kingdom, and Germany, and ahead of France. This represents a change of about 1 percentage point from the overall results of the 2012 S&T survey. There was a 4 percentage point decrease in how often France is ranked among the top five countries; the ordering of the top five countries, however, remains the same.

When asked to rate Canada’s research strength among other advanced countries in their field of expertise, 72% (4,005) of respondents rated Canadian research as “strong” (corresponding to a score of 5 or higher on a 7-point scale) compared with 68% in the 2012 S&T survey (Table 3.4). [pp. 40-41 Print; pp. 78-70 PDF]

Before I forget, there was mention of the international research scene,

Growth in research output, as estimated by number of publications, varies considerably for the 20 top countries. Brazil, China, India, Iran, and South Korea have had the most significant increases in publication output over the last 10 years. [emphases mine] In particular, the dramatic increase in China’s output means that it is closing the gap with the United States. In 2014, China’s output was 95% of that of the United States, compared with 26% in 2003. [emphasis mine]

Table 3.2 shows the Growth Index (GI), a measure of the rate at which the research output for a given country changed between 2003 and 2014, normalized by the world growth rate. If a country’s growth in research output is higher than the world average, the GI score is greater than 1.0. For example, between 2003 and 2014, China’s GI score was 1.50 (i.e., 50% greater than the world average) compared with 0.88 and 0.80 for Canada and the United States, respectively. Note that the dramatic increase in publication production of emerging economies such as China and India has had a negative impact on Canada’s rank and GI score (see CCA, 2016).

As long as I’ve been blogging (10 years), the international research community (in particular the US) has been looking over its shoulder at China.

Patents and intellectual property

As an inventor, Hedy got more than one patent. Much has been made of the fact that  despite an agreement, the US Navy did not pay her or her partner (George Antheil) for work that would lead to significant military use (apparently, it was instrumental in the Bay of Pigs incident, for those familiar with that bit of history), GPS, WiFi, Bluetooth, and more.

Some comments about patents. They are meant to encourage more innovation by ensuring that creators/inventors get paid for their efforts .This is true for a set time period and when it’s over, other people get access and can innovate further. It’s not intended to be a lifelong (or inheritable) source of income. The issue in Lamarr’s case is that the navy developed the technology during the patent’s term without telling either her or her partner so, of course, they didn’t need to compensate them despite the original agreement. They really should have paid her and Antheil.

The current patent situation, particularly in the US, is vastly different from the original vision. These days patents are often used as weapons designed to halt innovation. One item that should be noted is that the Canadian federal budget indirectly addressed their misuse (from my March 16, 2018 posting),

Surprisingly, no one else seems to have mentioned a new (?) intellectual property strategy introduced in the document (from Chapter 2: Progress; scroll down about 80% of the way, Note: The formatting has been changed),

Budget 2018 proposes measures in support of a new Intellectual Property Strategy to help Canadian entrepreneurs better understand and protect intellectual property, and get better access to shared intellectual property.

What Is a Patent Collective?
A Patent Collective is a way for firms to share, generate, and license or purchase intellectual property. The collective approach is intended to help Canadian firms ensure a global “freedom to operate”, mitigate the risk of infringing a patent, and aid in the defence of a patent infringement suit.

Budget 2018 proposes to invest $85.3 million over five years, starting in 2018–19, with $10 million per year ongoing, in support of the strategy. The Minister of Innovation, Science and Economic Development will bring forward the full details of the strategy in the coming months, including the following initiatives to increase the intellectual property literacy of Canadian entrepreneurs, and to reduce costs and create incentives for Canadian businesses to leverage their intellectual property:

  • To better enable firms to access and share intellectual property, the Government proposes to provide $30 million in 2019–20 to pilot a Patent Collective. This collective will work with Canada’s entrepreneurs to pool patents, so that small and medium-sized firms have better access to the critical intellectual property they need to grow their businesses.
  • To support the development of intellectual property expertise and legal advice for Canada’s innovation community, the Government proposes to provide $21.5 million over five years, starting in 2018–19, to Innovation, Science and Economic Development Canada. This funding will improve access for Canadian entrepreneurs to intellectual property legal clinics at universities. It will also enable the creation of a team in the federal government to work with Canadian entrepreneurs to help them develop tailored strategies for using their intellectual property and expanding into international markets.
  • To support strategic intellectual property tools that enable economic growth, Budget 2018 also proposes to provide $33.8 million over five years, starting in 2018–19, to Innovation, Science and Economic Development Canada, including $4.5 million for the creation of an intellectual property marketplace. This marketplace will be a one-stop, online listing of public sector-owned intellectual property available for licensing or sale to reduce transaction costs for businesses and researchers, and to improve Canadian entrepreneurs’ access to public sector-owned intellectual property.

The Government will also consider further measures, including through legislation, in support of the new intellectual property strategy.

Helping All Canadians Harness Intellectual Property
Intellectual property is one of our most valuable resources, and every Canadian business owner should understand how to protect and use it.

To better understand what groups of Canadians are benefiting the most from intellectual property, Budget 2018 proposes to provide Statistics Canada with $2 million over three years to conduct an intellectual property awareness and use survey. This survey will help identify how Canadians understand and use intellectual property, including groups that have traditionally been less likely to use intellectual property, such as women and Indigenous entrepreneurs. The results of the survey should help the Government better meet the needs of these groups through education and awareness initiatives.

The Canadian Intellectual Property Office will also increase the number of education and awareness initiatives that are delivered in partnership with business, intermediaries and academia to ensure Canadians better understand, integrate and take advantage of intellectual property when building their business strategies. This will include targeted initiatives to support underrepresented groups.

Finally, Budget 2018 also proposes to invest $1 million over five years to enable representatives of Canada’s Indigenous Peoples to participate in discussions at the World Intellectual Property Organization related to traditional knowledge and traditional cultural expressions, an important form of intellectual property.

It’s not wholly clear what they mean by ‘intellectual property’. The focus seems to be on  patents as they are the only intellectual property (as opposed to copyright and trademarks) singled out in the budget. As for how the ‘patent collective’ is going to meet all its objectives, this budget supplies no clarity on the matter. On the plus side, I’m glad to see that indigenous peoples’ knowledge is being acknowledged as “an important form of intellectual property” and I hope the discussions at the World Intellectual Property Organization are fruitful.

As for the patent situation in Canada (from the report released April 10, 2018),

Over the past decade, the Canadian patent flow in all technical sectors has consistently decreased. Patent flow provides a partial picture of how patents in Canada are exploited. A negative flow represents a deficit of patented inventions owned by Canadian assignees versus the number of patented inventions created by Canadian inventors. The patent flow for all Canadian patents decreased from about −0.04 in 2003 to −0.26 in 2014 (Figure 4.7). This means that there is an overall deficit of 26% of patent ownership in Canada. In other words, fewer patents were owned by Canadian institutions than were invented in Canada.

This is a significant change from 2003 when the deficit was only 4%. The drop is consistent across all technical sectors in the past 10 years, with Mechanical Engineering falling the least, and Electrical Engineering the most (Figure 4.7). At the technical field level, the patent flow dropped significantly in Digital Communication and Telecommunications. For example, the Digital Communication patent flow fell from 0.6 in 2003 to −0.2 in 2014. This fall could be partially linked to Nortel’s US$4.5 billion patent sale [emphasis mine] to the Rockstar consortium (which included Apple, BlackBerry, Ericsson, Microsoft, and Sony) (Brickley, 2011). Food Chemistry and Microstructural [?] and Nanotechnology both also showed a significant drop in patent flow. [p. 83 Print; p. 121 PDF]

Despite a fall in the number of parents for ‘Digital Communication’, we’re still doing well according to statistics elsewhere in this report. Is it possible that patents aren’t that big a deal? Of course, it’s also possible that we are enjoying the benefits of past work and will miss out on future work. (Note: A video of the April 10, 2018 report presentation by Max Blouw features him saying something like that.)

One last note, Nortel died many years ago. Disconcertingly, this report, despite more than one reference to Nortel, never mentions the company’s demise.

Boxed text

While the expert panel wasn’t tasked to answer certain types of questions, as I’ve noted earlier they managed to sneak in a few items.  One of the strategies they used was putting special inserts into text boxes including this (from the report released April 10, 2018),

Box 4.2
The FinTech Revolution

Financial services is a key industry in Canada. In 2015, the industry accounted for 4.4%

of Canadia jobs and about 7% of Canadian GDP (Burt, 2016). Toronto is the second largest financial services hub in North America and one of the most vibrant research hubs in FinTech. Since 2010, more than 100 start-up companies have been founded in Canada, attracting more than $1 billion in investment (Moffatt, 2016). In 2016 alone, venture-backed investment in Canadian financial technology companies grew by 35% to $137.7 million (Ho, 2017). The Toronto Financial Services Alliance estimates that there are approximately 40,000 ICT specialists working in financial services in Toronto alone.

AI, blockchain, [emphasis mine] and other results of ICT research provide the basis for several transformative FinTech innovations including, for example, decentralized transaction ledgers, cryptocurrencies (e.g., bitcoin), and AI-based risk assessment and fraud detection. These innovations offer opportunities to develop new markets for established financial services firms, but also provide entry points for technology firms to develop competing service offerings, increasing competition in the financial services industry. In response, many financial services companies are increasing their investments in FinTech companies (Breznitz et al., 2015). By their own account, the big five banks invest more than $1 billion annually in R&D of advanced software solutions, including AI-based innovations (J. Thompson, personal communication, 2016). The banks are also increasingly investing in university research and collaboration with start-up companies. For instance, together with several large insurance and financial management firms, all big five banks have invested in the Vector Institute for Artificial Intelligence (Kolm, 2017).

I’m glad to see the mention of blockchain while AI (artificial intelligence) is an area where we have innovated (from the report released April 10, 2018),

AI has attracted researchers and funding since the 1960s; however, there were periods of stagnation in the 1970s and 1980s, sometimes referred to as the “AI winter.” During this period, the Canadian Institute for Advanced Research (CIFAR), under the direction of Fraser Mustard, started supporting AI research with a decade-long program called Artificial Intelligence, Robotics and Society, [emphasis mine] which was active from 1983 to 1994. In 2004, a new program called Neural Computation and Adaptive Perception was initiated and renewed twice in 2008 and 2014 under the title, Learning in Machines and Brains. Through these programs, the government provided long-term, predictable support for high- risk research that propelled Canadian researchers to the forefront of global AI development. In the 1990s and early 2000s, Canadian research output and impact on AI were second only to that of the United States (CIFAR, 2016). NSERC has also been an early supporter of AI. According to its searchable grant database, NSERC has given funding to research projects on AI since at least 1991–1992 (the earliest searchable year) (NSERC, 2017a).

The University of Toronto, the University of Alberta, and the Université de Montréal have emerged as international centres for research in neural networks and deep learning, with leading experts such as Geoffrey Hinton and Yoshua Bengio. Recently, these locations have expanded into vibrant hubs for research in AI applications with a diverse mix of specialized research institutes, accelerators, and start-up companies, and growing investment by major international players in AI development, such as Microsoft, Google, and Facebook. Many highly influential AI researchers today are either from Canada or have at some point in their careers worked at a Canadian institution or with Canadian scholars.

As international opportunities in AI research and the ICT industry have grown, many of Canada’s AI pioneers have been drawn to research institutions and companies outside of Canada. According to the OECD, Canada’s share of patents in AI declined from 2.4% in 2000 to 2005 to 2% in 2010 to 2015. Although Canada is the sixth largest producer of top-cited scientific publications related to machine learning, firms headquartered in Canada accounted for only 0.9% of all AI-related inventions from 2012 to 2014 (OECD, 2017c). Canadian AI researchers, however, remain involved in the core nodes of an expanding international network of AI researchers, most of whom continue to maintain ties with their home institutions. Compared with their international peers, Canadian AI researchers are engaged in international collaborations far more often than would be expected by Canada’s level of research output, with Canada ranking fifth in collaboration. [p. 97-98 Print; p. 135-136 PDF]

The only mention of robotics seems to be here in this section and it’s only in passing. This is a bit surprising given its global importance. I wonder if robotics has been somehow hidden inside the term artificial intelligence, although sometimes it’s vice versa with robot being used to describe artificial intelligence. I’m noticing this trend of assuming the terms are synonymous or interchangeable not just in Canadian publications but elsewhere too.  ’nuff said.

Getting back to the matter at hand, t he report does note that patenting (technometric data) is problematic (from the report released April 10, 2018),

The limitations of technometric data stem largely from their restricted applicability across areas of R&D. Patenting, as a strategy for IP management, is similarly limited in not being equally relevant across industries. Trends in patenting can also reflect commercial pressures unrelated to R&D activities, such as defensive or strategic patenting practices. Finally, taxonomies for assessing patents are not aligned with bibliometric taxonomies, though links can be drawn to research publications through the analysis of patent citations. [p. 105 Print; p. 143 PDF]

It’s interesting to me that they make reference to many of the same issues that I mention but they seem to forget and don’t use that information in their conclusions.

There is one other piece of boxed text I want to highlight (from the report released April 10, 2018),

Box 6.3
Open Science: An Emerging Approach to Create New Linkages

Open Science is an umbrella term to describe collaborative and open approaches to
undertaking science, which can be powerful catalysts of innovation. This includes
the development of open collaborative networks among research performers, such
as the private sector, and the wider distribution of research that usually results when
restrictions on use are removed. Such an approach triggers faster translation of ideas
among research partners and moves the boundaries of pre-competitive research to
later, applied stages of research. With research results freely accessible, companies
can focus on developing new products and processes that can be commercialized.

Two Canadian organizations exemplify the development of such models. In June
2017, Genome Canada, the Ontario government, and pharmaceutical companies
invested $33 million in the Structural Genomics Consortium (SGC) (Genome Canada,
2017). Formed in 2004, the SGC is at the forefront of the Canadian open science
movement and has contributed to many key research advancements towards new
treatments (SGC, 2018). McGill University’s Montréal Neurological Institute and
Hospital has also embraced the principles of open science. Since 2016, it has been
sharing its research results with the scientific community without restriction, with
the objective of expanding “the impact of brain research and accelerat[ing] the
discovery of ground-breaking therapies to treat patients suffering from a wide range
of devastating neurological diseases” (neuro, n.d.).

This is exciting stuff and I’m happy the panel featured it. (I wrote about the Montréal Neurological Institute initiative in a Jan. 22, 2016 posting.)

More than once, the report notes the difficulties with using bibliometric and technometric data as measures of scientific achievement and progress and open science (along with its cousins, open data and open access) are contributing to the difficulties as James Somers notes in his April 5, 2018 article ‘The Scientific Paper is Obsolete’ for The Atlantic (Note: Links have been removed),

The scientific paper—the actual form of it—was one of the enabling inventions of modernity. Before it was developed in the 1600s, results were communicated privately in letters, ephemerally in lectures, or all at once in books. There was no public forum for incremental advances. By making room for reports of single experiments or minor technical advances, journals made the chaos of science accretive. Scientists from that point forward became like the social insects: They made their progress steadily, as a buzzing mass.

The earliest papers were in some ways more readable than papers are today. They were less specialized, more direct, shorter, and far less formal. Calculus had only just been invented. Entire data sets could fit in a table on a single page. What little “computation” contributed to the results was done by hand and could be verified in the same way.

The more sophisticated science becomes, the harder it is to communicate results. Papers today are longer than ever and full of jargon and symbols. They depend on chains of computer programs that generate data, and clean up data, and plot data, and run statistical models on data. These programs tend to be both so sloppily written and so central to the results that it’s [sic] contributed to a replication crisis, or put another way, a failure of the paper to perform its most basic task: to report what you’ve actually discovered, clearly enough that someone else can discover it for themselves.

Perhaps the paper itself is to blame. Scientific methods evolve now at the speed of software; the skill most in demand among physicists, biologists, chemists, geologists, even anthropologists and research psychologists, is facility with programming languages and “data science” packages. And yet the basic means of communicating scientific results hasn’t changed for 400 years. Papers may be posted online, but they’re still text and pictures on a page.

What would you get if you designed the scientific paper from scratch today? A little while ago I spoke to Bret Victor, a researcher who worked at Apple on early user-interface prototypes for the iPad and now runs his own lab in Oakland, California, that studies the future of computing. Victor has long been convinced that scientists haven’t yet taken full advantage of the computer. “It’s not that different than looking at the printing press, and the evolution of the book,” he said. After Gutenberg, the printing press was mostly used to mimic the calligraphy in bibles. It took nearly 100 years of technical and conceptual improvements to invent the modern book. “There was this entire period where they had the new technology of printing, but they were just using it to emulate the old media.”Victor gestured at what might be possible when he redesigned a journal article by Duncan Watts and Steven Strogatz, “Collective dynamics of ‘small-world’ networks.” He chose it both because it’s one of the most highly cited papers in all of science and because it’s a model of clear exposition. (Strogatz is best known for writing the beloved “Elements of Math” column for The New York Times.)

The Watts-Strogatz paper described its key findings the way most papers do, with text, pictures, and mathematical symbols. And like most papers, these findings were still hard to swallow, despite the lucid prose. The hardest parts were the ones that described procedures or algorithms, because these required the reader to “play computer” in their head, as Victor put it, that is, to strain to maintain a fragile mental picture of what was happening with each step of the algorithm.Victor’s redesign interleaved the explanatory text with little interactive diagrams that illustrated each step. In his version, you could see the algorithm at work on an example. You could even control it yourself….

For anyone interested in the evolution of how science is conducted and communicated, Somers’ article is a fascinating and in depth look at future possibilities.

Subregional R&D

I didn’t find this quite as compelling as the last time and that may be due to the fact that there’s less information and I think the 2012 report was the first to examine the Canadian R&D scene with a subregional (in their case, provinces) lens. On a high note, this report also covers cities (!) and regions, as well as, provinces.

Here’s the conclusion (from the report released April 10, 2018),

Ontario leads Canada in R&D investment and performance. The province accounts for almost half of R&D investment and personnel, research publications and collaborations, and patents. R&D activity in Ontario produces high-quality publications in each of Canada’s five R&D strengths, reflecting both the quantity and quality of universities in the province. Quebec lags Ontario in total investment, publications, and patents, but performs as well (citations) or better (R&D intensity) by some measures. Much like Ontario, Quebec researchers produce impactful publications across most of Canada’s five R&D strengths. Although it invests an amount similar to that of Alberta, British Columbia does so at a significantly higher intensity. British Columbia also produces more highly cited publications and patents, and is involved in more international research collaborations. R&D in British Columbia and Alberta clusters around Vancouver and Calgary in areas such as physics and ICT and in clinical medicine and energy, respectively. [emphasis mine] Smaller but vibrant R&D communities exist in the Prairies and Atlantic Canada [also referred to as the Maritime provinces or Maritimes] (and, to a lesser extent, in the Territories) in natural resource industries.

Globally, as urban populations expand exponentially, cities are likely to drive innovation and wealth creation at an increasing rate in the future. In Canada, R&D activity clusters around five large cities: Toronto, Montréal, Vancouver, Ottawa, and Calgary. These five cities create patents and high-tech companies at nearly twice the rate of other Canadian cities. They also account for half of clusters in the services sector, and many in advanced manufacturing.

Many clusters relate to natural resources and long-standing areas of economic and research strength. Natural resource clusters have emerged around the location of resources, such as forestry in British Columbia, oil and gas in Alberta, agriculture in Ontario, mining in Quebec, and maritime resources in Atlantic Canada. The automotive, plastics, and steel industries have the most individual clusters as a result of their economic success in Windsor, Hamilton, and Oshawa. Advanced manufacturing industries tend to be more concentrated, often located near specialized research universities. Strong connections between academia and industry are often associated with these clusters. R&D activity is distributed across the country, varying both between and within regions. It is critical to avoid drawing the wrong conclusion from this fact. This distribution does not imply the existence of a problem that needs to be remedied. Rather, it signals the benefits of diverse innovation systems, with differentiation driven by the needs of and resources available in each province. [pp.  132-133 Print; pp. 170-171 PDF]

Intriguingly, there’s no mention that in British Columbia (BC), there are leading areas of research: Visual & Performing Arts, Psychology & Cognitive Sciences, and Clinical Medicine (according to the table on p. 117 Print, p. 153 PDF).

As I said and hinted earlier, we’ve got brains; they’re just not the kind of brains that command respect.

Final comments

My hat’s off to the expert panel and staff of the Council of Canadian Academies. Combining two previous reports into one could not have been easy. As well, kudos to their attempts to broaden the discussion by mentioning initiative such as open science and for emphasizing the problems with bibliometrics, technometrics, and other measures. I have covered only parts of this assessment, (Competing in a Global Innovation Economy: The Current State of R&D in Canada), there’s a lot more to it including a substantive list of reference materials (bibliography).

While I have argued that perhaps the situation isn’t quite as bad as the headlines and statistics may suggest, there are some concerning trends for Canadians but we have to acknowledge that many countries have stepped up their research game and that’s good for all of us. You don’t get better at anything unless you work with and play with others who are better than you are. For example, both India and Italy surpassed us in numbers of published research papers. We slipped from 7th place to 9th. Thank you, Italy and India. (And, Happy ‘Italian Research in the World Day’ on April 15, 2018, the day’s inaugural year. In Italian: Piano Straordinario “Vivere all’Italiana” – Giornata della ricerca Italiana nel mondo.)

Unfortunately, the reading is harder going than previous R&D assessments in the CCA catalogue. And in the end, I can’t help thinking we’re just a little bit like Hedy Lamarr. Not really appreciated in all of our complexities although the expert panel and staff did try from time to time. Perhaps the government needs to find better ways of asking the questions.

***ETA April 12, 2018 at 1500 PDT: Talking about missing the obvious! I’ve been ranting on about how research strength in visual and performing arts and in philosophy and theology, etc. is perfectly fine and could lead to ‘traditional’ science breakthroughs without underlining the point by noting that Antheil was a musician, Lamarr was as an actress and they set the foundation for work by electrical engineers (or people with that specialty) for their signature work leading to WiFi, etc.***

There is, by the way, a Hedy-Canada connection. In 1998, she sued Canadian software company Corel, for its unauthorized use of her image on their Corel Draw 8 product packaging. She won.

More stuff

For those who’d like to see and hear the April 10, 2017 launch for “Competing in a Global Innovation Economy: The Current State of R&D in Canada” or the Third Assessment as I think of it, go here.

The report can be found here.

For anyone curious about ‘Bombshell: The Hedy Lamarr Story’ to be broadcast on May 18, 2018 as part of PBS’s American Masters series, there’s this trailer,

For the curious, I did find out more about the Hedy Lamarr and Corel Draw. John Lettice’s December 2, 1998 article The Rgister describes the suit and her subsequent victory in less than admiring terms,

Our picture doesn’t show glamorous actress Hedy Lamarr, who yesterday [Dec. 1, 1998] came to a settlement with Corel over the use of her image on Corel’s packaging. But we suppose that following the settlement we could have used a picture of Corel’s packaging. Lamarr sued Corel earlier this year over its use of a CorelDraw image of her. The picture had been produced by John Corkery, who was 1996 Best of Show winner of the Corel World Design Contest. Corel now seems to have come to an undisclosed settlement with her, which includes a five-year exclusive (oops — maybe we can’t use the pack-shot then) licence to use “the lifelike vector illustration of Hedy Lamarr on Corel’s graphic software packaging”. Lamarr, bless ‘er, says she’s looking forward to the continued success of Corel Corporation,  …

There’s this excerpt from a Sept. 21, 2015 posting (a pictorial essay of Lamarr’s life) by Shahebaz Khan on The Blaze Blog,

6. CorelDRAW:
For several years beginning in 1997, the boxes of Corel DRAW’s software suites were graced by a large Corel-drawn image of Lamarr. The picture won Corel DRAW’s yearly software suite cover design contest in 1996. Lamarr sued Corel for using the image without her permission. Corel countered that she did not own rights to the image. The parties reached an undisclosed settlement in 1998.

There’s also a Nov. 23, 1998 Corel Draw 8 product review by Mike Gorman on mymac.com, which includes a screenshot of the packaging that precipitated the lawsuit. Once they settled, it seems Corel used her image at least one more time.

The Hedy Lamarr of international research: Canada’s Third assessment of The State of Science and Technology and Industrial Research and Development in Canada (1 of 2)

Before launching into the assessment, a brief explanation of my theme: Hedy Lamarr was considered to be one of the great beauties of her day,

“Ziegfeld Girl” Hedy Lamarr 1941 MGM *M.V.
Titles: Ziegfeld Girl
People: Hedy Lamarr
Image courtesy mptvimages.com [downloaded from https://www.imdb.com/title/tt0034415/mediaviewer/rm1566611456]

Aside from starring in Hollywood movies and, before that, movies in Europe, she was also an inventor and not just any inventor (from a Dec. 4, 2017 article by Laura Barnett for The Guardian), Note: Links have been removed,

Let’s take a moment to reflect on the mercurial brilliance of Hedy Lamarr. Not only did the Vienna-born actor flee a loveless marriage to a Nazi arms dealer to secure a seven-year, $3,000-a-week contract with MGM, and become (probably) the first Hollywood star to simulate a female orgasm on screen – she also took time out to invent a device that would eventually revolutionise mobile communications.

As described in unprecedented detail by the American journalist and historian Richard Rhodes in his new book, Hedy’s Folly, Lamarr and her business partner, the composer George Antheil, were awarded a patent in 1942 for a “secret communication system”. It was meant for radio-guided torpedoes, and the pair gave to the US Navy. It languished in their files for decades before eventually becoming a constituent part of GPS, Wi-Fi and Bluetooth technology.

(The article goes on to mention other celebrities [Marlon Brando, Barbara Cartland, Mark Twain, etc] and their inventions.)

Lamarr’s work as an inventor was largely overlooked until the 1990’s when the technology community turned her into a ‘cultish’ favourite and from there her reputation grew and acknowledgement increased culminating in Rhodes’ book and the documentary by Alexandra Dean, ‘Bombshell: The Hedy Lamarr Story (to be broadcast as part of PBS’s American Masters series on May 18, 2018).

Canada as Hedy Lamarr

There are some parallels to be drawn between Canada’s S&T and R&D (science and technology; research and development) and Ms. Lamarr. Chief amongst them, we’re not always appreciated for our brains. Not even by people who are supposed to know better such as the experts on the panel for the ‘Third assessment of The State of Science and Technology and Industrial Research and Development in Canada’ (proper title: Competing in a Global Innovation Economy: The Current State of R&D in Canada) from the Expert Panel on the State of Science and Technology and Industrial Research and Development in Canada.

A little history

Before exploring the comparison to Hedy Lamarr further, here’s a bit more about the history of this latest assessment from the Council of Canadian Academies (CCA), from the report released April 10, 2018,

This assessment of Canada’s performance indicators in science, technology, research, and innovation comes at an opportune time. The Government of Canada has expressed a renewed commitment in several tangible ways to this broad domain of activity including its Innovation and Skills Plan, the announcement of five superclusters, its appointment of a new Chief Science Advisor, and its request for the Fundamental Science Review. More specifically, the 2018 Federal Budget demonstrated the government’s strong commitment to research and innovation with historic investments in science.

The CCA has a decade-long history of conducting evidence-based assessments about Canada’s research and development activities, producing seven assessments of relevance:

The State of Science and Technology in Canada (2006) [emphasis mine]
•Innovation and Business Strategy: Why Canada Falls Short (2009)
•Catalyzing Canada’s Digital Economy (2010)
•Informing Research Choices: Indicators and Judgment (2012)
The State of Science and Technology in Canada (2012) [emphasis mine]
The State of Industrial R&D in Canada (2013) [emphasis mine]
•Paradox Lost: Explaining Canada’s Research Strength and Innovation Weakness (2013)

Using similar methods and metrics to those in The State of Science and Technology in Canada (2012) and The State of Industrial R&D in Canada (2013), this assessment tells a similar and familiar story: Canada has much to be proud of, with world-class researchers in many domains of knowledge, but the rest of the world is not standing still. Our peers are also producing high quality results, and many countries are making significant commitments to supporting research and development that will position them to better leverage their strengths to compete globally. Canada will need to take notice as it determines how best to take action. This assessment provides valuable material for that conversation to occur, whether it takes place in the lab or the legislature, the bench or the boardroom. We also hope it will be used to inform public discussion. [p. ix Print, p. 11 PDF]

This latest assessment succeeds the general 2006 and 2012 reports, which were mostly focused on academic research, and combines it with an assessment of industrial research, which was previously separate. Also, this third assessment’s title (Competing in a Global Innovation Economy: The Current State of R&D in Canada) makes what was previously quietly declared in the text, explicit from the cover onwards. It’s all about competition, despite noises such as the 2017 Naylor report (Review of fundamental research) about the importance of fundamental research.

One other quick comment, I did wonder in my July 1, 2016 posting (featuring the announcement of the third assessment) how combining two assessments would impact the size of the expert panel and the size of the final report,

Given the size of the 2012 assessment of science and technology at 232 pp. (PDF) and the 2013 assessment of industrial research and development at 220 pp. (PDF) with two expert panels, the imagination boggles at the potential size of the 2016 expert panel and of the 2016 assessment combining the two areas.

I got my answer with regard to the panel as noted in my Oct. 20, 2016 update (which featured a list of the members),

A few observations, given the size of the task, this panel is lean. As well, there are three women in a group of 13 (less than 25% representation) in 2016? It’s Ontario and Québec-dominant; only BC and Alberta rate a representative on the panel. I hope they will find ways to better balance this panel and communicate that ‘balanced story’ to the rest of us. On the plus side, the panel has representatives from the humanities, arts, and industry in addition to the expected representatives from the sciences.

The imbalance I noted then was addressed, somewhat, with the selection of the reviewers (from the report released April 10, 2018),

The CCA wishes to thank the following individuals for their review of this report:

Ronald Burnett, C.M., O.B.C., RCA, Chevalier de l’ordre des arts et des
lettres, President and Vice-Chancellor, Emily Carr University of Art and Design
(Vancouver, BC)

Michelle N. Chretien, Director, Centre for Advanced Manufacturing and Design
Technologies, Sheridan College; Former Program and Business Development
Manager, Electronic Materials, Xerox Research Centre of Canada (Brampton,
ON)

Lisa Crossley, CEO, Reliq Health Technologies, Inc. (Ancaster, ON)
Natalie Dakers, Founding President and CEO, Accel-Rx Health Sciences
Accelerator (Vancouver, BC)

Fred Gault, Professorial Fellow, United Nations University-MERIT (Maastricht,
Netherlands)

Patrick D. Germain, Principal Engineering Specialist, Advanced Aerodynamics,
Bombardier Aerospace (Montréal, QC)

Robert Brian Haynes, O.C., FRSC, FCAHS, Professor Emeritus, DeGroote
School of Medicine, McMaster University (Hamilton, ON)

Susan Holt, Chief, Innovation and Business Relationships, Government of
New Brunswick (Fredericton, NB)

Pierre A. Mohnen, Professor, United Nations University-MERIT and Maastricht
University (Maastricht, Netherlands)

Peter J. M. Nicholson, C.M., Retired; Former and Founding President and
CEO, Council of Canadian Academies (Annapolis Royal, NS)

Raymond G. Siemens, Distinguished Professor, English and Computer Science
and Former Canada Research Chair in Humanities Computing, University of
Victoria (Victoria, BC) [pp. xii- xiv Print; pp. 15-16 PDF]

The proportion of women to men as reviewers jumped up to about 36% (4 of 11 reviewers) and there are two reviewers from the Maritime provinces. As usual, reviewers external to Canada were from Europe. Although this time, they came from Dutch institutions rather than UK or German institutions. Interestingly and unusually, there was no one from a US institution. When will they start using reviewers from other parts of the world?

As for the report itself, it is 244 pp. (PDF). (For the really curious, I have a  December 15, 2016 post featuring my comments on the preliminary data for the third assessment.)

To sum up, they had a lean expert panel tasked with bringing together two inquiries and two reports. I imagine that was daunting. Good on them for finding a way to make it manageable.

Bibliometrics, patents, and a survey

I wish more attention had been paid to some of the issues around open science, open access, and open data, which are changing how science is being conducted. (I have more about this from an April 5, 2018 article by James Somers for The Atlantic but more about that later.) If I understand rightly, they may not have been possible due to the nature of the questions posed by the government when requested the assessment.

As was done for the second assessment, there is an acknowledgement that the standard measures/metrics (bibliometrics [no. of papers published, which journals published them; number of times papers were cited] and technometrics [no. of patent applications, etc.] of scientific accomplishment and progress are not the best and new approaches need to be developed and adopted (from the report released April 10, 2018),

It is also worth noting that the Panel itself recognized the limits that come from using traditional historic metrics. Additional approaches will be needed the next time this assessment is done. [p. ix Print; p. 11 PDF]

For the second assessment and as a means of addressing some of the problems with metrics, the panel decided to take a survey which the panel for the third assessment has also done (from the report released April 10, 2018),

The Panel relied on evidence from multiple sources to address its charge, including a literature review and data extracted from statistical agencies and organizations such as Statistics Canada and the OECD. For international comparisons, the Panel focused on OECD countries along with developing countries that are among the top 20 producers of peer-reviewed research publications (e.g., China, India, Brazil, Iran, Turkey). In addition to the literature review, two primary research approaches informed the Panel’s assessment:
•a comprehensive bibliometric and technometric analysis of Canadian research publications and patents; and,
•a survey of top-cited researchers around the world.

Despite best efforts to collect and analyze up-to-date information, one of the Panel’s findings is that data limitations continue to constrain the assessment of R&D activity and excellence in Canada. This is particularly the case with industrial R&D and in the social sciences, arts, and humanities. Data on industrial R&D activity continue to suffer from time lags for some measures, such as internationally comparable data on R&D intensity by sector and industry. These data also rely on industrial categories (i.e., NAICS and ISIC codes) that can obscure important trends, particularly in the services sector, though Statistics Canada’s recent revisions to how this data is reported have improved this situation. There is also a lack of internationally comparable metrics relating to R&D outcomes and impacts, aside from those based on patents.

For the social sciences, arts, and humanities, metrics based on journal articles and other indexed publications provide an incomplete and uneven picture of research contributions. The expansion of bibliometric databases and methodological improvements such as greater use of web-based metrics, including paper views/downloads and social media references, will support ongoing, incremental improvements in the availability and accuracy of data. However, future assessments of R&D in Canada may benefit from more substantive integration of expert review, capable of factoring in different types of research outputs (e.g., non-indexed books) and impacts (e.g., contributions to communities or impacts on public policy). The Panel has no doubt that contributions from the humanities, arts, and social sciences are of equal importance to national prosperity. It is vital that such contributions are better measured and assessed. [p. xvii Print; p. 19 PDF]

My reading: there’s a problem and we’re not going to try and fix it this time. Good luck to those who come after us. As for this line: “The Panel has no doubt that contributions from the humanities, arts, and social sciences are of equal importance to national prosperity.” Did no one explain that when you use ‘no doubt’, you are introducing doubt? It’s a cousin to ‘don’t take this the wrong way’ and ‘I don’t mean to be rude but …’ .

Good news

This is somewhat encouraging (from the report released April 10, 2018),

Canada’s international reputation for its capacity to participate in cutting-edge R&D is strong, with 60% of top-cited researchers surveyed internationally indicating that Canada hosts world-leading infrastructure or programs in their fields. This share increased by four percentage points between 2012 and 2017. Canada continues to benefit from a highly educated population and deep pools of research skills and talent. Its population has the highest level of educational attainment in the OECD in the proportion of the population with
a post-secondary education. However, among younger cohorts (aged 25 to 34), Canada has fallen behind Japan and South Korea. The number of researchers per capita in Canada is on a par with that of other developed countries, andincreased modestly between 2004 and 2012. Canada’s output of PhD graduates has also grown in recent years, though it remains low in per capita terms relative to many OECD countries. [pp. xvii-xviii; pp. 19-20]

Don’t let your head get too big

Most of the report observes that our international standing is slipping in various ways such as this (from the report released April 10, 2018),

In contrast, the number of R&D personnel employed in Canadian businesses
dropped by 20% between 2008 and 2013. This is likely related to sustained and
ongoing decline in business R&D investment across the country. R&D as a share
of gross domestic product (GDP) has steadily declined in Canada since 2001,
and now stands well below the OECD average (Figure 1). As one of few OECD
countries with virtually no growth in total national R&D expenditures between
2006 and 2015, Canada would now need to more than double expenditures to
achieve an R&D intensity comparable to that of leading countries.

Low and declining business R&D expenditures are the dominant driver of this
trend; however, R&D spending in all sectors is implicated. Government R&D
expenditures declined, in real terms, over the same period. Expenditures in the
higher education sector (an indicator on which Canada has traditionally ranked
highly) are also increasing more slowly than the OECD average. Significant
erosion of Canada’s international competitiveness and capacity to participate
in R&D and innovation is likely to occur if this decline and underinvestment
continue.

Between 2009 and 2014, Canada produced 3.8% of the world’s research
publications, ranking ninth in the world. This is down from seventh place for
the 2003–2008 period. India and Italy have overtaken Canada although the
difference between Italy and Canada is small. Publication output in Canada grew
by 26% between 2003 and 2014, a growth rate greater than many developed
countries (including United States, France, Germany, United Kingdom, and
Japan), but below the world average, which reflects the rapid growth in China
and other emerging economies. Research output from the federal government,
particularly the National Research Council Canada, dropped significantly
between 2009 and 2014.(emphasis mine)  [p. xviii Print; p. 20 PDF]

For anyone unfamiliar with Canadian politics,  2009 – 2014 were years during which Stephen Harper’s Conservatives formed the government. Justin Trudeau’s Liberals were elected to form the government in late 2015.

During Harper’s years in government, the Conservatives were very interested in changing how the National Research Council of Canada operated and, if memory serves, the focus was on innovation over research. Consequently, the drop in their research output is predictable.

Given my interest in nanotechnology and other emerging technologies, this popped out (from the report released April 10, 2018),

When it comes to research on most enabling and strategic technologies, however, Canada lags other countries. Bibliometric evidence suggests that, with the exception of selected subfields in Information and Communication Technologies (ICT) such as Medical Informatics and Personalized Medicine, Canada accounts for a relatively small share of the world’s research output for promising areas of technology development. This is particularly true for Biotechnology, Nanotechnology, and Materials science [emphasis mine]. Canada’s research impact, as reflected by citations, is also modest in these areas. Aside from Biotechnology, none of the other subfields in Enabling and Strategic Technologies has an ARC rank among the top five countries. Optoelectronics and photonics is the next highest ranked at 7th place, followed by Materials, and Nanoscience and Nanotechnology, both of which have a rank of 9th. Even in areas where Canadian researchers and institutions played a seminal role in early research (and retain a substantial research capacity), such as Artificial Intelligence and Regenerative Medicine, Canada has lost ground to other countries.

Arguably, our early efforts in artificial intelligence wouldn’t have garnered us much in the way of ranking and yet we managed some cutting edge work such as machine learning. I’m not suggesting the expert panel should have or could have found some way to measure these kinds of efforts but I’m wondering if there could have been some acknowledgement in the text of the report. I’m thinking a couple of sentences in a paragraph about the confounding nature of scientific research where areas that are ignored for years and even decades then become important (e.g., machine learning) but are not measured as part of scientific progress until after they are universally recognized.

Still, point taken about our diminishing returns in ’emerging’ technologies and sciences (from the report released April 10, 2018),

The impression that emerges from these data is sobering. With the exception of selected ICT subfields, such as Medical Informatics, bibliometric evidence does not suggest that Canada excels internationally in most of these research areas. In areas such as Nanotechnology and Materials science, Canada lags behind other countries in levels of research output and impact, and other countries are outpacing Canada’s publication growth in these areas — leading to declining shares of world publications. Even in research areas such as AI, where Canadian researchers and institutions played a foundational role, Canadian R&D activity is not keeping pace with that of other countries and some researchers trained in Canada have relocated to other countries (Section 4.4.1). There are isolated exceptions to these trends, but the aggregate data reviewed by this Panel suggest that Canada is not currently a world leader in research on most emerging technologies.

The Hedy Lamarr treatment

We have ‘good looks’ (arts and humanities) but not the kind of brains (physical sciences and engineering) that people admire (from the report released April 10, 2018),

Canada, relative to the world, specializes in subjects generally referred to as the
humanities and social sciences (plus health and the environment), and does
not specialize as much as others in areas traditionally referred to as the physical
sciences and engineering. Specifically, Canada has comparatively high levels
of research output in Psychology and Cognitive Sciences, Public Health and
Health Services, Philosophy and Theology, Earth and Environmental Sciences,
and Visual and Performing Arts. [emphases mine] It accounts for more than 5% of world researchin these fields. Conversely, Canada has lower research output than expected
in Chemistry, Physics and Astronomy, Enabling and Strategic Technologies,
Engineering, and Mathematics and Statistics. The comparatively low research
output in core areas of the natural sciences and engineering is concerning,
and could impair the flexibility of Canada’s research base, preventing research
institutions and researchers from being able to pivot to tomorrow’s emerging
research areas. [p. xix Print; p. 21 PDF]

Couldn’t they have used a more buoyant tone? After all, science was known as ‘natural philosophy’ up until the 19th century. As for visual and performing arts, let’s include poetry as a performing and literary art (both have been the case historically and cross-culturally) and let’s also note that one of the great physics texts, (De rerum natura by Lucretius) was a multi-volume poem (from Lucretius’ Wikipedia entry; Note: Links have been removed).

His poem De rerum natura (usually translated as “On the Nature of Things” or “On the Nature of the Universe”) transmits the ideas of Epicureanism, which includes Atomism [the concept of atoms forming materials] and psychology. Lucretius was the first writer to introduce Roman readers to Epicurean philosophy.[15] The poem, written in some 7,400 dactylic hexameters, is divided into six untitled books, and explores Epicurean physics through richly poetic language and metaphors. Lucretius presents the principles of atomism; the nature of the mind and soul; explanations of sensation and thought; the development of the world and its phenomena; and explains a variety of celestial and terrestrial phenomena. The universe described in the poem operates according to these physical principles, guided by fortuna, “chance”, and not the divine intervention of the traditional Roman deities.[16]

Should you need more proof that the arts might have something to contribute to physical sciences, there’s this in my March 7, 2018 posting,

It’s not often you see research that combines biologically inspired engineering and a molecular biophysicist with a professional animator who worked at Peter Jackson’s (Lord of the Rings film trilogy, etc.) Park Road Post film studio. An Oct. 18, 2017 news item on ScienceDaily describes the project,

Like many other scientists, Don Ingber, M.D., Ph.D., the Founding Director of the Wyss Institute, [emphasis mine] is concerned that non-scientists have become skeptical and even fearful of his field at a time when technology can offer solutions to many of the world’s greatest problems. “I feel that there’s a huge disconnect between science and the public because it’s depicted as rote memorization in schools, when by definition, if you can memorize it, it’s not science,” says Ingber, who is also the Judah Folkman Professor of Vascular Biology at Harvard Medical School and the Vascular Biology Program at Boston Children’s Hospital, and Professor of Bioengineering at the Harvard Paulson School of Engineering and Applied Sciences (SEAS). [emphasis mine] “Science is the pursuit of the unknown. We have a responsibility to reach out to the public and convey that excitement of exploration and discovery, and fortunately, the film industry is already great at doing that.”

“Not only is our physics-based simulation and animation system as good as other data-based modeling systems, it led to the new scientific insight [emphasis mine] that the limited motion of the dynein hinge focuses the energy released by ATP hydrolysis, which causes dynein’s shape change and drives microtubule sliding and axoneme motion,” says Ingber. “Additionally, while previous studies of dynein have revealed the molecule’s two different static conformations, our animation visually depicts one plausible way that the protein can transition between those shapes at atomic resolution, which is something that other simulations can’t do. The animation approach also allows us to visualize how rows of dyneins work in unison, like rowers pulling together in a boat, which is difficult using conventional scientific simulation approaches.”

It comes down to how we look at things. Yes, physical sciences and engineering are very important. If the report is to be believed we have a very highly educated population and according to PISA scores our students rank highly in mathematics, science, and reading skills. (For more information on Canada’s latest PISA scores from 2015 see this OECD page. As for PISA itself, it’s an OECD [Organization for Economic Cooperation and Development] programme where 15-year-old students from around the world are tested on their reading, mathematics, and science skills, you can get some information from my Oct. 9, 2013 posting.)

Is it really so bad that we choose to apply those skills in fields other than the physical sciences and engineering? It’s a little bit like Hedy Lamarr’s problem except instead of being judged for our looks and having our inventions dismissed, we’re being judged for not applying ourselves to physical sciences and engineering and having our work in other closely aligned fields dismissed as less important.

Canada’s Industrial R&D: an oft-told, very sad story

Bemoaning the state of Canada’s industrial research and development efforts has been a national pastime as long as I can remember. Here’s this from the report released April 10, 2018,

There has been a sustained erosion in Canada’s industrial R&D capacity and competitiveness. Canada ranks 33rd among leading countries on an index assessing the magnitude, intensity, and growth of industrial R&D expenditures. Although Canada is the 11th largest spender, its industrial R&D intensity (0.9%) is only half the OECD average and total spending is declining (−0.7%). Compared with G7 countries, the Canadian portfolio of R&D investment is more concentrated in industries that are intrinsically not as R&D intensive. Canada invests more heavily than the G7 average in oil and gas, forestry, machinery and equipment, and finance where R&D has been less central to business strategy than in many other industries. …  About 50% of Canada’s industrial R&D spending is in high-tech sectors (including industries such as ICT, aerospace, pharmaceuticals, and automotive) compared with the G7 average of 80%. Canadian Business Enterprise Expenditures on R&D (BERD) intensity is also below the OECD average in these sectors. In contrast, Canadian investment in low and medium-low tech sectors is substantially higher than the G7 average. Canada’s spending reflects both its long-standing industrial structure and patterns of economic activity.

R&D investment patterns in Canada appear to be evolving in response to global and domestic shifts. While small and medium-sized enterprises continue to perform a greater share of industrial R&D in Canada than in the United States, between 2009 and 2013, there was a shift in R&D from smaller to larger firms. Canada is an increasingly attractive place to conduct R&D. Investment by foreign-controlled firms in Canada has increased to more than 35% of total R&D investment, with the United States accounting for more than half of that. [emphasis mine]  Multinational enterprises seem to be increasingly locating some of their R&D operations outside their country of ownership, possibly to gain proximity to superior talent. Increasing foreign-controlled R&D, however, also could signal a long-term strategic loss of control over intellectual property (IP) developed in this country, ultimately undermining the government’s efforts to support high-growth firms as they scale up. [pp. xxii-xxiii Print; pp. 24-25 PDF]

Canada has been known as a ‘branch plant’ economy for decades. For anyone unfamiliar with the term, it means that companies from other countries come here, open up a branch and that’s how we get our jobs as we don’t have all that many large companies here. Increasingly, multinationals are locating R&D shops here.

While our small to medium size companies fund industrial R&D, it’s large companies (multinationals) which can afford long-term and serious investment in R&D. Luckily for companies from other countries, we have a well-educated population of people looking for jobs.

In 2017, we opened the door more widely so we can scoop up talented researchers and scientists from other countries, from a June 14, 2017 article by Beckie Smith for The PIE News,

Universities have welcomed the inclusion of the work permit exemption for academic stays of up to 120 days in the strategy, which also introduces expedited visa processing for some highly skilled professions.

Foreign researchers working on projects at a publicly funded degree-granting institution or affiliated research institution will be eligible for one 120-day stay in Canada every 12 months.

And universities will also be able to access a dedicated service channel that will support employers and provide guidance on visa applications for foreign talent.

The Global Skills Strategy, which came into force on June 12 [2017], aims to boost the Canadian economy by filling skills gaps with international talent.

As well as the short term work permit exemption, the Global Skills Strategy aims to make it easier for employers to recruit highly skilled workers in certain fields such as computer engineering.

“Employers that are making plans for job-creating investments in Canada will often need an experienced leader, dynamic researcher or an innovator with unique skills not readily available in Canada to make that investment happen,” said Ahmed Hussen, Minister of Immigration, Refugees and Citizenship.

“The Global Skills Strategy aims to give those employers confidence that when they need to hire from abroad, they’ll have faster, more reliable access to top talent.”

Coincidentally, Microsoft, Facebook, Google, etc. have announced, in 2017, new jobs and new offices in Canadian cities. There’s a also Chinese multinational telecom company Huawei Canada which has enjoyed success in Canada and continues to invest here (from a Jan. 19, 2018 article about security concerns by Matthew Braga for the Canadian Broadcasting Corporation (CBC) online news,

For the past decade, Chinese tech company Huawei has found no shortage of success in Canada. Its equipment is used in telecommunications infrastructure run by the country’s major carriers, and some have sold Huawei’s phones.

The company has struck up partnerships with Canadian universities, and say it is investing more than half a billion dollars in researching next generation cellular networks here. [emphasis mine]

While I’m not thrilled about using patents as an indicator of progress, this is interesting to note (from the report released April 10, 2018),

Canada produces about 1% of global patents, ranking 18th in the world. It lags further behind in trademark (34th) and design applications (34th). Despite relatively weak performance overall in patents, Canada excels in some technical fields such as Civil Engineering, Digital Communication, Other Special Machines, Computer Technology, and Telecommunications. [emphases mine] Canada is a net exporter of patents, which signals the R&D strength of some technology industries. It may also reflect increasing R&D investment by foreign-controlled firms. [emphasis mine] [p. xxiii Print; p. 25 PDF]

Getting back to my point, we don’t have large companies here. In fact, the dream for most of our high tech startups is to build up the company so it’s attractive to buyers, sell, and retire (hopefully before the age of 40). Strangely, the expert panel doesn’t seem to share my insight into this matter,

Canada’s combination of high performance in measures of research output and impact, and low performance on measures of industrial R&D investment and innovation (e.g., subpar productivity growth), continue to be viewed as a paradox, leading to the hypothesis that barriers are impeding the flow of Canada’s research achievements into commercial applications. The Panel’s analysis suggests the need for a more nuanced view. The process of transforming research into innovation and wealth creation is a complex multifaceted process, making it difficult to point to any definitive cause of Canada’s deficit in R&D investment and productivity growth. Based on the Panel’s interpretation of the evidence, Canada is a highly innovative nation, but significant barriers prevent the translation of innovation into wealth creation. The available evidence does point to a number of important contributing factors that are analyzed in this report. Figure 5 represents the relationships between R&D, innovation, and wealth creation.

The Panel concluded that many factors commonly identified as points of concern do not adequately explain the overall weakness in Canada’s innovation performance compared with other countries. [emphasis mine] Academia-business linkages appear relatively robust in quantitative terms given the extent of cross-sectoral R&D funding and increasing academia-industry partnerships, though the volume of academia-industry interactions does not indicate the nature or the quality of that interaction, nor the extent to which firms are capitalizing on the research conducted and the resulting IP. The educational system is high performing by international standards and there does not appear to be a widespread lack of researchers or STEM (science, technology, engineering, and mathematics) skills. IP policies differ across universities and are unlikely to explain a divergence in research commercialization activity between Canadian and U.S. institutions, though Canadian universities and governments could do more to help Canadian firms access university IP and compete in IP management and strategy. Venture capital availability in Canada has improved dramatically in recent years and is now competitive internationally, though still overshadowed by Silicon Valley. Technology start-ups and start-up ecosystems are also flourishing in many sectors and regions, demonstrating their ability to build on research advances to develop and deliver innovative products and services.

You’ll note there’s no mention of a cultural issue where start-ups are designed for sale as soon as possible and this isn’t new. Years ago, there was an accounting firm that published a series of historical maps (the last one I saw was in 2005) of technology companies in the Vancouver region. Technology companies were being developed and sold to large foreign companies from the 19th century to present day.

Part 2

Leftover 2017 memristor news bits

i have two bits of news, one from this October 2017 about using light to control a memristor’s learning properties and one from December 2017 about memristors and neural networks.

Shining a light on the memristor

Michael Berger wrote an October 30, 2017 Nanowerk Sportlight article about some of the latest work concerning memristors and light,

Memristors – or resistive memory – are nanoelectronic devices that are very promising components for next generation memory and computing devices. They are two-terminal electric elements similar to a conventional resistor – however, the electric resistance in a memristor is dependent on the charge passing through it; which means that its conductance can be precisely modulated by charge or flux through it. Its special property is that its resistance can be programmed (resistor function) and subsequently remains stored (memory function).

In this sense, a memristor is similar to a synapse in the human brain because it exhibits the same switching characteristics, i.e. it is able, with a high level of plasticity, to modify the efficiency of signal transfer between neurons under the influence of the transfer itself. That’s why researchers are hopeful to use memristors for the fabrication of electronic synapses for neuromorphic (i.e. brain-like) computing that mimics some of the aspects of learning and computation in human brains.

Human brains may be slow at pure number crunching but they are excellent at handling fast dynamic sensory information such as image and voice recognition. Walking is something that we take for granted but this is quite challenging for robots, especially over uneven terrain.

“Memristors present an opportunity to make new types of computers that are different from existing von Neumann architectures, which traditional computers are based upon,” Dr Neil T. Kemp, a Lecturer in Physics at the University of Hull [UK], tells Nanowerk. “Our team at the University of Hull is focussed on making memristor devices dynamically reconfigurable and adaptive – we believe this is the route to making a new generation of artificial intelligence systems that are smarter and can exhibit complex behavior. Such systems would also have the advantage of memristors, high density integration and lower power usage, so these systems would be more lightweight, portable and not need re-charging so often – which is something really needed for robots etc.”

In their new paper in Nanoscale (“Reversible Optical Switching Memristors with Tunable STDP Synaptic Plasticity: A Route to Hierarchical Control in Artificial Intelligent Systems”), Kemp and his team demonstrate the ability to reversibly control the learning properties of memristors via optical means.

The reversibility is achieved by changing the polarization of light. The researchers have used this effect to demonstrate tuneable learning in a memristor. One way this is achieved is through something called Spike Timing Dependent Plasticity (STDP), which is an effect known to occur in human brains and is linked with sensory perception, spatial reasoning, language and conscious thought in the neocortex.

STDP learning is based upon differences in the arrival time of signals from two adjacent neurons. The University of Hull team has shown that they can modulate the synaptic plasticity via optical means which enables the devices to have tuneable learning.

“Our research findings are important because it demonstrates that light can be used to control the learning properties of a memristor,” Kemp points out. “We have shown that light can be used in a reversible manner to change the connection strength (or conductivity) of artificial memristor synapses and as well control their ability to forget i.e. we can dynamically change device to have short-term or long-term memory.”

According to the team, there are many potential applications, such as adaptive electronic circuits controllable via light, or in more complex systems, such as neuromorphic computing, the development of optically reconfigurable neural networks.

Having optically controllable memristors can also facilitate the implementation of hierarchical control in larger artificial-brain like systems, whereby some of the key processes that are carried out by biological molecules in human brains can be emulated in solid-state devices through patterning with light.

Some of these processes include synaptic pruning, conversion of short term memory to long term memory, erasing of certain memories that are no longer needed or changing the sensitivity of synapses to be more adept at learning new information.

“The ability to control this dynamically, both spatially and temporally, is particularly interesting since it would allow neural networks to be reconfigurable on the fly through either spatial patterning or by adjusting the intensity of the light source,” notes Kemp.

In their new paper in Nanoscale Currently, the devices are more suited to neuromorphic computing applications, which do not need to be as fast. Optical control of memristors opens the route to dynamically tuneable and reprogrammable synaptic circuits as well the ability (via optical patterning) to have hierarchical control in larger and more complex artificial intelligent systems.

“Artificial Intelligence is really starting to come on strong in many areas, especially in the areas of voice/image recognition and autonomous systems – we could even say that this is the next revolution, similarly to what the industrial revolution was to farming and production processes,” concludes Kemp. “There are many challenges to overcome though. …

That excerpt should give you the gist of Berger’s article and, for those who need more information, there’s Berger’s article and, also, a link to and a citation for the paper,

Reversible optical switching memristors with tunable STDP synaptic plasticity: a route to hierarchical control in artificial intelligent systems by Ayoub H. Jaafar, Robert J. Gray, Emanuele Verrelli, Mary O’Neill, Stephen. M. Kelly, and Neil T. Kemp. Nanoscale, 2017,9, 17091-17098 DOI: 10.1039/C7NR06138B First published on 24 Oct 2017

This paper is behind a paywall.

The memristor and the neural network

It would seem machine learning could experience a significant upgrade if the work in Wei Lu’s University of Michigan laboratory can be scaled for general use. From a December 22, 2017 news item on ScienceDaily,

A new type of neural network made with memristors can dramatically improve the efficiency of teaching machines to think like humans.

The network, called a reservoir computing system, could predict words before they are said during conversation, and help predict future outcomes based on the present.

The research team that created the reservoir computing system, led by Wei Lu, professor of electrical engineering and computer science at the University of Michigan, recently published their work in Nature Communications.

A December 19, 2017 University of Michigan news release (also on EurekAlert) by Dan Newman, which originated the news item, expands on the theme,

Reservoir computing systems, which improve on a typical neural network’s capacity and reduce the required training time, have been created in the past with larger optical components. However, the U-M group created their system using memristors, which require less space and can be integrated more easily into existing silicon-based electronics.

Memristors are a special type of resistive device that can both perform logic and store data. This contrasts with typical computer systems, where processors perform logic separate from memory modules. In this study, Lu’s team used a special memristor that memorizes events only in the near history.

Inspired by brains, neural networks are composed of neurons, or nodes, and synapses, the connections between nodes.

To train a neural network for a task, a neural network takes in a large set of questions and the answers to those questions. In this process of what’s called supervised learning, the connections between nodes are weighted more heavily or lightly to minimize the amount of error in achieving the correct answer.

Once trained, a neural network can then be tested without knowing the answer. For example, a system can process a new photo and correctly identify a human face, because it has learned the features of human faces from other photos in its training set.

“A lot of times, it takes days or months to train a network,” says Lu. “It is very expensive.”

Image recognition is also a relatively simple problem, as it doesn’t require any information apart from a static image. More complex tasks, such as speech recognition, can depend highly on context and require neural networks to have knowledge of what has just occurred, or what has just been said.

“When transcribing speech to text or translating languages, a word’s meaning and even pronunciation will differ depending on the previous syllables,” says Lu.

This requires a recurrent neural network, which incorporates loops within the network that give the network a memory effect. However, training these recurrent neural networks is especially expensive, Lu says.

Reservoir computing systems built with memristors, however, can skip most of the expensive training process and still provide the network the capability to remember. This is because the most critical component of the system – the reservoir – does not require training.

When a set of data is inputted into the reservoir, the reservoir identifies important time-related features of the data, and hands it off in a simpler format to a second network. This second network then only needs training like simpler neural networks, changing weights of the features and outputs that the first network passed on until it achieves an acceptable level of error.

Enlargereservoir computing system

IMAGE:  Schematic of a reservoir computing system, showing the reservoir with internal dynamics and the simpler output. Only the simpler output needs to be trained, allowing for quicker and lower-cost training. Courtesy Wei Lu.

 

“The beauty of reservoir computing is that while we design it, we don’t have to train it,” says Lu.

The team proved the reservoir computing concept using a test of handwriting recognition, a common benchmark among neural networks. Numerals were broken up into rows of pixels, and fed into the computer with voltages like Morse code, with zero volts for a dark pixel and a little over one volt for a white pixel.

Using only 88 memristors as nodes to identify handwritten versions of numerals, compared to a conventional network that would require thousands of nodes for the task, the reservoir achieved 91% accuracy.

Reservoir computing systems are especially adept at handling data that varies with time, like a stream of data or words, or a function depending on past results.

To demonstrate this, the team tested a complex function that depended on multiple past results, which is common in engineering fields. The reservoir computing system was able to model the complex function with minimal error.

Lu plans on exploring two future paths with this research: speech recognition and predictive analysis.

“We can make predictions on natural spoken language, so you don’t even have to say the full word,” explains Lu.

“We could actually predict what you plan to say next.”

In predictive analysis, Lu hopes to use the system to take in signals with noise, like static from far-off radio stations, and produce a cleaner stream of data. “It could also predict and generate an output signal even if the input stopped,” he says.

EnlargeWei Lu

IMAGE:  Wei Lu, Professor of Electrical Engineering & Computer Science at the University of Michigan holds a memristor he created. Photo: Marcin Szczepanski.

 

The work was published in Nature Communications in the article, “Reservoir computing using dynamic memristors for temporal information processing”, with authors Chao Du, Fuxi Cai, Mohammed Zidan, Wen Ma, Seung Hwan Lee, and Prof. Wei Lu.

The research is part of a $6.9 million DARPA [US Defense Advanced Research Projects Agency] project, called “Sparse Adaptive Local Learning for Sensing and Analytics [also known as SALLSA],” that aims to build a computer chip based on self-organizing, adaptive neural networks. The memristor networks are fabricated at Michigan’s Lurie Nanofabrication Facility.

Lu and his team previously used memristors in implementing “sparse coding,” which used a 32-by-32 array of memristors to efficiently analyze and recreate images.

Here’s a link to and a citation for the paper,

Reservoir computing using dynamic memristors for temporal information processing by Chao Du, Fuxi Cai, Mohammed A. Zidan, Wen Ma, Seung Hwan Lee & Wei D. Lu. Nature Communications 8, Article number: 2204 (2017) doi:10.1038/s41467-017-02337-y Published online: 19 December 2017

This is an open access paper.

Machine learning software and quantum computers that think

A Sept. 14, 2017 news item on phys.org sets the stage for quantum machine learning by explaining a few basics first,

Language acquisition in young children is apparently connected with their ability to detect patterns. In their learning process, they search for patterns in the data set that help them identify and optimize grammar structures in order to properly acquire the language. Likewise, online translators use algorithms through machine learning techniques to optimize their translation engines to produce well-rounded and understandable outcomes. Even though many translations did not make much sense at all at the beginning, in these past years we have been able to see major improvements thanks to machine learning.

Machine learning techniques use mathematical algorithms and tools to search for patterns in data. These techniques have become powerful tools for many different applications, which can range from biomedical uses such as in cancer reconnaissance, in genetics and genomics, in autism monitoring and diagnosis and even plastic surgery, to pure applied physics, for studying the nature of materials, matter or even complex quantum systems.

Capable of adapting and changing when exposed to a new set of data, machine learning can identify patterns, often outperforming humans in accuracy. Although machine learning is a powerful tool, certain application domains remain out of reach due to complexity or other aspects that rule out the use of the predictions that learning algorithms provide.

Thus, in recent years, quantum machine learning has become a matter of interest because of is vast potential as a possible solution to these unresolvable challenges and quantum computers show to be the right tool for its solution.

A Sept. 14, 2017 Institute of Photonic Sciences ([Catalan] Institut de Ciències Fotòniques] ICFO) press release, which originated the news item, goes on to detail a recently published overview of the state of quantum machine learning,

In a recent study, published in Nature, an international team of researchers integrated by Jacob Biamonte from Skoltech/IQC, Peter Wittek from ICFO, Nicola Pancotti from MPQ, Patrick Rebentrost from MIT, Nathan Wiebe from Microsoft Research, and Seth Lloyd from MIT have reviewed the actual status of classical machine learning and quantum machine learning. In their review, they have thoroughly addressed different scenarios dealing with classical and quantum machine learning. In their study, they have considered different possible combinations: the conventional method of using classical machine learning to analyse classical data, using quantum machine learning to analyse both classical and quantum data, and finally, using classical machine learning to analyse quantum data.

Firstly, they set out to give an in-depth view of the status of current supervised and unsupervised learning protocols in classical machine learning by stating all applied methods. They introduce quantum machine learning and provide an extensive approach on how this technique could be used to analyse both classical and quantum data, emphasizing that quantum machines could accelerate processing timescales thanks to the use of quantum annealers and universal quantum computers. Quantum annealing technology has better scalability, but more limited use cases. For instance, the latest iteration of D-Wave’s [emphasis mine] superconducting chip integrates two thousand qubits, and it is used for solving certain hard optimization problems and for efficient sampling. On the other hand, universal (also called gate-based) quantum computers are harder to scale up, but they are able to perform arbitrary unitary operations on qubits by sequences of quantum logic gates. This resembles how digital computers can perform arbitrary logical operations on classical bits.

However, they address the fact that controlling a quantum system is very complex and analyzing classical data with quantum resources is not as straightforward as one may think, mainly due to the challenge of building quantum interface devices that allow classical information to be encoded into a quantum mechanical form. Difficulties, such as the “input” or “output” problems appear to be the major technical challenge that needs to be overcome.

The ultimate goal is to find the most optimized method that is able to read, comprehend and obtain the best outcomes of a data set, be it classical or quantum. Quantum machine learning is definitely aimed at revolutionizing the field of computer sciences, not only because it will be able to control quantum computers, speed up the information processing rates far beyond current classical velocities, but also because it is capable of carrying out innovative functions, such quantum deep learning, that could not only recognize counter-intuitive patterns in data, invisible to both classical machine learning and to the human eye, but also reproduce them.

As Peter Wittek [emphasis mine] finally states, “Writing this paper was quite a challenge: we had a committee of six co-authors with different ideas about what the field is, where it is now, and where it is going. We rewrote the paper from scratch three times. The final version could not have been completed without the dedication of our editor, to whom we are indebted.”

It was a bit of a surprise to see local (Vancouver, Canada) company D-Wave Systems mentioned but i notice that one of the paper’s authors (Peter Wittek) is mentioned in a May 22, 2017 D-Wave news release announcing a new partnership to foster quantum machine learning,

Today [May 22, 2017] D-Wave Systems Inc., the leader in quantum computing systems and software, announced a new initiative with the Creative Destruction Lab (CDL) at the University of Toronto’s Rotman School of Management. D-Wave will work with CDL, as a CDL Partner, to create a new track to foster startups focused on quantum machine learning. The new track will complement CDL’s successful existing track in machine learning. Applicants selected for the intensive one-year program will go through an introductory boot camp led by Dr. Peter Wittek [emphasis mine], author of Quantum Machine Learning: What Quantum Computing means to Data Mining, with instruction and technical support from D-Wave experts, access to a D-Wave 2000Q™ quantum computer, and the opportunity to use a D-Wave sampling service to enable machine learning computations and applications. D-Wave staff will be a part of the committee selecting up to 40 individuals for the program, which begins in September 2017.

For anyone interested in the paper, here’s a link to and a citation,

Quantum machine learning by Jacob Biamonte, Peter Wittek, Nicola Pancotti, Patrick Rebentrost, Nathan Wiebe, & Seth Lloyd. Nature 549, 195–202 (14 September 2017) doi:10.1038/nature23474 Published online 13 September 2017

This paper is behind a paywall.