Tag Archives: artificial intelligence (AI)

An algorithm for modern quilting

Caption: Each of the blocks in this quilt were designed using an algorithm-based tool developed by Stanford researchers. Credit: Mackenzie Leake

I love the colours. This research into quilting and artificial intelligence (AI) was presented at SIGGRAPH 2021 in August. (SIGGRAPH is, also known as, ACM SIGGRAPH or ‘Association for Computing Machinery’s Special Interest Group on Computer Graphics and Interactive Techniques’.)

A June 3, 2021 news item on ScienceDaily announced the presentation,

Stanford University computer science graduate student Mackenzie Leake has been quilting since age 10, but she never imagined the craft would be the focus of her doctoral dissertation. Included in that work is new prototype software that can facilitate pattern-making for a form of quilting called foundation paper piecing, which involves using a backing made of foundation paper to lay out and sew a quilted design.

Developing a foundation paper piece quilt pattern — which looks similar to a paint-by-numbers outline — is often non-intuitive. There are few formal guidelines for patterning and those that do exist are insufficient to assure a successful result.

“Quilting has this rich tradition and people make these very personal, cherished heirlooms but paper piece quilting often requires that people work from patterns that other people designed,” said Leake, who is a member of the lab of Maneesh Agrawala, the Forest Baskett Professor of Computer Science and director of the Brown Institute for Media Innovation at Stanford. “So, we wanted to produce a digital tool that lets people design the patterns that they want to design without having to think through all of the geometry, ordering and constraints.”

A paper describing this work is published and will be presented at the computer graphics conference SIGGRAPH 2021 in August.

A June 2, 2021 Stanford University news release (also on EurekAlert), which originated the news item, provides more detail,

Respecting the craft

In describing the allure of paper piece quilts, Leake cites the modern aesthetic and high level of control and precision. The seams of the quilt are sewn through the paper pattern and, as the seaming process proceeds, the individual pieces of fabric are flipped over to form the final design. All of this “sew and flip” action means the pattern must be produced in a careful order.

Poorly executed patterns can lead to loose pieces, holes, misplaced seams and designs that are simply impossible to complete. When quilters create their own paper piecing designs, figuring out the order of the seams can take considerable time – and still lead to unsatisfactory results.

“The biggest challenge that we’re tackling is letting people focus on the creative part and offload the mental energy of figuring out whether they can use this technique or not,” said Leake, who is lead author of the SIGGRAPH paper. “It’s important to me that we’re really aware and respectful of the way that people like to create and that we aren’t over-automating that process.”

This isn’t Leake’s first foray into computer-aided quilting. She previously designed a tool for improvisational quilting, which she presented [PatchProv: Supporting Improvistiional Design Practices for Modern Quilting by Mackenzie Leake, Frances Lai, Tovi Grossman, Daniel Wigdor, and Ben Lafreniere] at the human-computer interaction conference CHI in May [2021]. [Note: Links to the May 2021 conference and paper added by me.]

Quilting theory

Developing the algorithm at the heart of this latest quilting software required a substantial theoretical foundation. With few existing guidelines to go on, the researchers had to first gain a more formal understanding of what makes a quilt paper piece-able, and then represent that mathematically.

They eventually found what they needed in a particular graph structure, called a hypergraph. While so-called “simple” graphs can only connect data points by lines, a hypergraph can accommodate overlapping relationships between many data points. (A Venn diagram is a type of hypergraph.) The researchers found that a pattern will be paper piece-able if it can be depicted by a hypergraph whose edges can be removed one at a time in a specific order – which would correspond to how the seams are sewn in the pattern.

The prototype software allows users to sketch out a design and the underlying hypergraph-based algorithm determines what paper foundation patterns could make it possible – if any. Many designs result in multiple pattern options and users can adjust their sketch until they get a pattern they like. The researchers hope to make a version of their software publicly available this summer.

“I didn’t expect to be writing my computer science dissertation on quilting when I started,” said Leake. “But I found this really rich space of problems involving design and computation and traditional crafts, so there have been lots of different pieces we’ve been able to pull off and examine in that space.”

###

Researchers from University of California, Berkeley and Cornell University are co-authors of this paper. Agrawala is also an affiliate of the Institute for Human-Centered Artificial Intelligence (HAI).

An abstract for the paper “A Mathematical Foundation for Foundation Paper Pieceable Quilts” by Mackenzie Leake, Gilbert Bernstein, Abe Davis and Maneesh Agrawala can be found here along with links to a PDF of the full paper and video on YouTube.

Afterthought: I noticed that all of the co-authors for the May 2021 paper are from the University of Toronto and most of them including Mackenzie Leake are associated with that university’s Chatham Labs.

Finishing Beethoven’s unfinished 10th Symphony

Throughout the project, Beethoven’s genius loomed. Circe Denyer

This is an artificial intelligence (AI) story set to music. Professor Ahmed Elgammal (Director of the Art & AI Lab at Rutgers University located in New Jersey, US) has a September 24, 2021 essay posted on The Conversation (and, then, in the Smithsonian Magazine online) describing the AI project and upcoming album release and performance (Note: A link has been removed),

When Ludwig van Beethoven died in 1827, he was three years removed from the completion of his Ninth Symphony, a work heralded by many as his magnum opus. He had started work on his 10th Symphony but, due to deteriorating health, wasn’t able to make much headway: All he left behind were some musical sketches.

A full recording of Beethoven’s 10th Symphony is set to be released on Oct. 9, 2021, the same day as the world premiere performance scheduled to take place in Bonn, Germany – the culmination of a two-year-plus effort.

These excerpts from the Elgammal’s September 24, 2021 essay on the The Conversation provide a summarized view of events. By the way, this isn’t the first time an attempt has been made to finish Beethoven’s 10th Symphony (Note: Links have been removed),

Around 1817, the Royal Philharmonic Society in London commissioned Beethoven to write his Ninth and 10th symphonies. Written for an orchestra, symphonies often contain four movements: the first is performed at a fast tempo, the second at a slower one, the third at a medium or fast tempo, and the last at a fast tempo.

Beethoven completed his Ninth Symphony in 1824, which concludes with the timeless “Ode to Joy.”

But when it came to the 10th Symphony, Beethoven didn’t leave much behind, other than some musical notes and a handful of ideas he had jotted down.

There have been some past attempts to reconstruct parts of Beethoven’s 10th Symphony. Most famously, in 1988, musicologist Barry Cooper ventured to complete the first and second movements. He wove together 250 bars of music from the sketches to create what was, in his view, a production of the first movement that was faithful to Beethoven’s vision.

Yet the sparseness of Beethoven’s sketches made it impossible for symphony experts to go beyond that first movement.

In early 2019, Dr. Matthias Röder, the director of the Karajan Institute, an organization in Salzburg, Austria, that promotes music technology, contacted me. He explained that he was putting together a team to complete Beethoven’s 10th Symphony in celebration of the composer’s 250th birthday. Aware of my work on AI-generated art, he wanted to know if AI would be able to help fill in the blanks left by Beethoven.

Röder then compiled a team that included Austrian composer Walter Werzowa. Famous for writing Intel’s signature bong jingle, Werzowa was tasked with putting together a new kind of composition that would integrate what Beethoven left behind with what the AI would generate. Mark Gotham, a computational music expert, led the effort to transcribe Beethoven’s sketches and process his entire body of work so the AI could be properly trained.

The team also included Robert Levin, a musicologist at Harvard University who also happens to be an incredible pianist. Levin had previously finished a number of incomplete 18th-century works by Mozart and Johann Sebastian Bach.

… We didn’t have a machine that we could feed sketches to, push a button and have it spit out a symphony. Most AI available at the time couldn’t continue an uncompleted piece of music beyond a few additional seconds.

We would need to push the boundaries of what creative AI could do by teaching the machine Beethoven’s creative process – how he would take a few bars of music and painstakingly develop them into stirring symphonies, quartets and sonatas.

Here’s Elgammal’s description of the difficulties from an AI perspective, from the September 24, 2021 essay (Note: Links have been removed),

First, and most fundamentally, we needed to figure out how to take a short phrase, or even just a motif, and use it to develop a longer, more complicated musical structure, just as Beethoven would have done. For example, the machine had to learn how Beethoven constructed the Fifth Symphony out of a basic four-note motif.

Next, because the continuation of a phrase also needs to follow a certain musical form, whether it’s a scherzo, trio or fugue, the AI needed to learn Beethoven’s process for developing these forms.

The to-do list grew: We had to teach the AI how to take a melodic line and harmonize it. The AI needed to learn how to bridge two sections of music together. And we realized the AI had to be able to compose a coda, which is a segment that brings a section of a piece of music to its conclusion.

Finally, once we had a full composition, the AI was going to have to figure out how to orchestrate it, which involves assigning different instruments for different parts.

And it had to pull off these tasks in the way Beethoven might do so.

The team tested its work, from the September 24, 2021 essay, Note: A link has been removed,

In November 2019, the team met in person again – this time, in Bonn, at the Beethoven House Museum, where the composer was born and raised.

This meeting was the litmus test for determining whether AI could complete this project. We printed musical scores that had been developed by AI and built off the sketches from Beethoven’s 10th. A pianist performed in a small concert hall in the museum before a group of journalists, music scholars and Beethoven experts.

We challenged the audience to determine where Beethoven’s phrases ended and where the AI extrapolation began. They couldn’t.

A few days later, one of these AI-generated scores was played by a string quartet in a news conference. Only those who intimately knew Beethoven’s sketches for the 10th Symphony could determine when the AI-generated parts came in.

The success of these tests told us we were on the right track. But these were just a couple of minutes of music. There was still much more work to do.

There is a preview of the finished 10th symphony,

Beethoven X: The AI Project: III Scherzo. Allegro – Trio (Official Video) | Beethoven Orchestra Bonn

Modern Recordings / BMG present as a foretaste of the album “Beethoven X – The AI Project” (release: 8.10.) the edit of the 3rd movement “Scherzo. Allegro – Trio” as a classical music video. Listen now: https://lnk.to/BeethovenX-Scherzo

Album pre-order link: https://lnk.to/BeethovenX

The Beethoven Orchestra Bonn performing with Dirk Kaftan and Walter Werzowa a great recording of world-premiere Beethoven pieces. Developed by AI and music scientists as well as composers, Beethoven’s once unfinished 10th symphony now surprises with beautiful Beethoven-like harmonics and dynamics.

For anyone who’d like to hear the October 9, 2021 performance, Sharon Kelly included some details in her August 16, 2021 article for DiscoverMusic,

The world premiere of Beethoven’s 10th Symphony on 9 October 2021 at the Telekom Forum in Bonn, performed by the Beethoven Orchestra Bonn conducted by Dirk Kaftan, will be broadcast live and free of charge on MagentaMusik 360.

Sadly, the time is not listed but MagentaMusik 360 is fairly easy to find online.

You can find out more about Professor Elgammal on his Rutgers University profile page. Elgammal has graced this blog before in an August 16, 2019 posting “AI (artificial intelligence) artist got a show at a New York City art gallery“. He’s mentioned in an excerpt about 20% of the way down the page,

Ahmed Elgammal thinks AI art can be much more than that. A Rutgers University professor of computer science, Elgammal runs an art-and-artificial-intelligence lab, where he and his colleagues develop technologies that try to understand and generate new “art” (the scare quotes are Elgammal’s) with AI—not just credible copies of existing work, like GANs do. “That’s not art, that’s just repainting,” Elgammal says of GAN-made images. “It’s what a bad artist would do.”

Elgammal calls his approach a “creative adversarial network,” or CAN. It swaps a GAN’s discerner—the part that ensures similarity—for one that introduces novelty instead. The system amounts to a theory of how art evolves: through small alterations to a known style that produce a new one. That’s a convenient take, given that any machine-learning technique has to base its work on a specific training set.

Finally, thank you to @winsontang whose tweet led me to this story.

Precision skincare

An inkjet printer for your skin—it’s an idea I’m not sure I’m ready for. Still, I’m not the target market for the product being described in Rachel Kim Raczka’s June 2, 2021 article for Fast Company (Note: Links have been removed),

… I’ve had broken capillaries, patchy spots, and enlarged pores most of my adult life. And after I turned 30, I developed a glorious strip of melasma (a “sun mustache”) across my upper lip. The delicate balance of maintaining my “good” texture—skin that looks like skin—while disguising my “bad” texture is a constant push and pull. Still, I continue to fall victim to “no makeup” makeup, the frustratingly contradictory trend that will never die. A white whale that $599 high-tech beauty printer Opte hopes to fill.

Weirdly enough, “printer” is a fair representation of what Opte is. The size and shape of an electric razor, Opte’s Precision Wand’s tiny computer claims to detect and camouflage hyperpigmentation with a series of gentle swipes. The product deposits extremely small blends of white, yellow, and red pigments to hide discoloration using a blue LED and a hypersensitive camera that scans 200 photos per second. Opte then relies on an algorithm to apply color—housed in replaceable serum cartridges, delivered through 120 thermal inkjet nozzles—only onto contrasting patches of melanin via what CEO Matt Petersen calls “the world’s smallest inkjet printer.” 

Opte is a 15-year, 500,000-R&D-hour project developed under P&G Ventures, officially launched in 2020. While targeting hyperpigmentation was an end goal, the broader mission looked at focusing on “precision skincare.” …

… You start by dropping the included 11-ingredient serum cartridge into the pod; the $129 cartridges and refills come in three shades that the company says cover 98% of skin tones and last 90 days. The handheld device very loudly refills itself and displays instructions on a tiny screen on its handle. …

… While I can’t rely on the Opte to hide a blemish or dark circles—I’ll still need concealer to achieve that level of coverage—I can’t quite describe the “glowiness” using this gadget generates. With more use, I’ve come to retrain my brain to expect Opte to work more like an eraser than a crayon; it’s skincare, not makeup. My skin looks healthier and brighter but still, without a doubt, like my skin. 

There’s more discussion of how this product works in Raczka’s June 2, 2021 article and you can find the Opte website here. I have no idea if they ship this product outside the US or what that might cost.

Memristors, it’s all about the oxides

I have one research announcement from China and another from the Netherlands, both of which concern memristors and oxides.

China

A May 17, 2021 news item on Nanowerk announces work, which suggests that memristors may not need to rely solely on oxides but could instead utilize light more gainfully,

Scientists are getting better at making neuron-like junctions for computers that mimic the human brain’s random information processing, storage and recall. Fei Zhuge of the Chinese Academy of Sciences and colleagues reviewed the latest developments in the design of these ‘memristors’ for the journal Science and Technology of Advanced Materials …

Computers apply artificial intelligence programs to recall previously learned information and make predictions. These programs are extremely energy- and time-intensive: typically, vast volumes of data must be transferred between separate memory and processing units. To solve this issue, researchers have been developing computer hardware that allows for more random and simultaneous information transfer and storage, much like the human brain.

Electronic circuits in these ‘neuromorphic’ computers include memristors that resemble the junctions between neurons called synapses. Energy flows through a material from one electrode to another, much like a neuron firing a signal across the synapse to the next neuron. Scientists are now finding ways to better tune this intermediate material so the information flow is more stable and reliable.

I had no success locating the original news release, which originated the news item, but have found this May 17, 2021 news item on eedesignit.com, which provides the remaining portion of the news release.

“Oxides are the most widely used materials in memristors,” said Zhuge. “But oxide memristors have unsatisfactory stability and reliability. Oxide-based hybrid structures can effectively improve this.”

Memristors are usually made of an oxide-based material sandwiched between two electrodes. Researchers are getting better results when they combine two or more layers of different oxide-based materials between the electrodes. When an electrical current flows through the network, it induces ions to drift within the layers. The ions’ movements ultimately change the memristor’s resistance, which is necessary to send or stop a signal through the junction.

Memristors can be tuned further by changing the compounds used for electrodes or by adjusting the intermediate oxide-based materials. Zhuge and his team are currently developing optoelectronic neuromorphic computers based on optically-controlled oxide memristors. Compared to electronic memristors, photonic ones are expected to have higher operation speeds and lower energy consumption. They could be used to construct next generation artificial visual systems with high computing efficiency.

Now for a picture that accompanied the news release, which follows,

Fig. The all-optically controlled memristor developed for optoelectronic neuromorphic computing (Image by NIMTE)

Here’s the February 7, 2021 Ningbo Institute of Materials Technology and Engineering (NIMTE) press release featuring this work and a more technical description,

A research group led by Prof. ZHUGE Fei at the Ningbo Institute of Materials Technology and Engineering (NIMTE) of the Chinese Academy of Sciences (CAS) developed an all-optically controlled (AOC) analog memristor, whose memconductance can be reversibly tuned by varying only the wavelength of the controlling light.

As the next generation of artificial intelligence (AI), neuromorphic computing (NC) emulates the neural structure and operation of the human brain at the physical level, and thus can efficiently perform multiple advanced computing tasks such as learning, recognition and cognition.

Memristors are promising candidates for NC thanks to the feasibility of high-density 3D integration and low energy consumption. Among them, the emerging optoelectronic memristors are competitive by virtue of combining the advantages of both photonics and electronics. However, the reversible tuning of memconductance depends highly on the electric excitation, which have severely limited the development and application of optoelectronic NC.

To address this issue, researchers at NIMTE proposed a bilayered oxide AOC memristor, based on the relatively mature semiconductor material InGaZnO and a memconductance tuning mechanism of light-induced electron trapping and detrapping.

The traditional electrical memristors require strong electrical stimuli to tune their memconductance, leading to high power consumption, a large amount of Joule heat, microstructural change triggered by the Joule heat, and even high crosstalk in memristor crossbars.

On the contrary, the developed AOC memristor does not involve microstructure changes, and can operate upon weak light irradiation with light power density of only 20 μW cm-2, which has provided a new approach to overcome the instability of the memristor.

Specifically, the AOC memristor can serve as an excellent synaptic emulator and thus mimic spike-timing-dependent plasticity (STDP) which is an important learning rule in the brain, indicating its potential applications in AOC spiking neural networks for high-efficiency optoelectronic NC.

Moreover, compared to purely optical computing, the optoelectronic computing using our AOC memristor showed higher practical feasibility, on account of the simple structure and fabrication process of the device.

The study may shed light on the in-depth research and practical application of optoelectronic NC, and thus promote the development of the new generation of AI.

This work was supported by the National Natural Science Foundation of China (No. 61674156 and 61874125), the Strategic Priority Research Program of Chinese Academy of Sciences (No. XDB32050204), and the Zhejiang Provincial Natural Science Foundation of China (No. LD19E020001).

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

Hybrid oxide brain-inspired neuromorphic devices for hardware implementation of artificial intelligence by Jingrui Wang, Xia Zhuge & Fei Zhuge. Science and Technology of Advanced Materials Volume 22, 2021 – Issue 1 Pages 326-344 DOI: https://doi.org/10.1080/14686996.2021.1911277 Published online:14 May 2021

This paper appears to be open access.

Netherlands

In this case, a May 18, 2021 news item on Nanowerk marries oxides to spintronics,

Classic computers use binary values (0/1) to perform. By contrast, our brain cells can use more values to operate, making them more energy-efficient than computers. This is why scientists are interested in neuromorphic (brain-like) computing.

Physicists from the University of Groningen (the Netherlands) have used a complex oxide to create elements comparable to the neurons and synapses in the brain using spins, a magnetic property of electrons.

The press release, which follows, was accompanied by this image illustrating the work,

Caption: Schematic of the proposed device structure for neuromorphic spintronic memristors. The write path is between the terminals through the top layer (black dotted line), the read path goes through the device stack (red dotted line). The right side of the figure indicates how the choice of substrate dictates whether the device will show deterministic or probabilistic behaviour. Credit: Banerjee group, University of Groningen

A May 18, 2021 University of Groningen press release (also on EurekAlert), which originated the news item, adds more ‘spin’ to the story,

Although computers can do straightforward calculations much faster than humans, our brains outperform silicon machines in tasks like object recognition. Furthermore, our brain uses less energy than computers. Part of this can be explained by the way our brain operates: whereas a computer uses a binary system (with values 0 or 1), brain cells can provide more analogue signals with a range of values.

Thin films

The operation of our brains can be simulated in computers, but the basic architecture still relies on a binary system. That is why scientist look for ways to expand this, creating hardware that is more brain-like, but will also interface with normal computers. ‘One idea is to create magnetic bits that can have intermediate states’, says Tamalika Banerjee, Professor of Spintronics of Functional Materials at the Zernike Institute for Advanced Materials, University of Groningen. She works on spintronics, which uses a magnetic property of electrons called ‘spin’ to transport, manipulate and store information.

In this study, her PhD student Anouk Goossens, first author of the paper, created thin films of a ferromagnetic metal (strontium-ruthenate oxide, SRO) grown on a substrate of strontium titanate oxide. The resulting thin film contained magnetic domains that were perpendicular to the plane of the film. ‘These can be switched more efficiently than in-plane magnetic domains’, explains Goossens. By adapting the growth conditions, it is possible to control the crystal orientation in the SRO. Previously, out-of-plane magnetic domains have been made using other techniques, but these typically require complex layer structures.

Magnetic anisotropy

The magnetic domains can be switched using a current through a platinum electrode on top of the SRO. Goossens: ‘When the magnetic domains are oriented perfectly perpendicular to the film, this switching is deterministic: the entire domain will switch.’ However, when the magnetic domains are slightly tilted, the response is probabilistic: not all the domains are the same, and intermediate values occur when only part of the crystals in the domain have switched.

By choosing variants of the substrate on which the SRO is grown, the scientists can control its magnetic anisotropy. This allows them to produce two different spintronic devices. ‘This magnetic anisotropy is exactly what we wanted’, says Goossens. ‘Probabilistic switching compares to how neurons function, while the deterministic switching is more like a synapse.’

The scientists expect that in the future, brain-like computer hardware can be created by combining these different domains in a spintronic device that can be connected to standard silicon-based circuits. Furthermore, probabilistic switching would also allow for stochastic computing, a promising technology which represents continuous values by streams of random bits. Banerjee: ‘We have found a way to control intermediate states, not just for memory but also for computing.’

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

Anisotropy and Current Control of Magnetization in SrRuO3/SrTiO3 Heterostructures for Spin-Memristors by A.S. Goossens, M.A.T. Leiviskä and T. Banerjee. Frontiers in Nanotechnology DOI: https://doi.org/10.3389/fnano.2021.680468 Published: 18 May 2021

This appears to be open access.

Speed up your reading with an interactive typeface

A May 12, 2021 news item on ScienceDaily brings news of a technology that makes reading easier,

AdaptiFont has recently been presented at CHI, the leading Conference on Human Factors in Computing.

Language is without doubt the most pervasive medium for exchanging knowledge between humans. However, spoken language or abstract text need to be made visible in order to be read, be it in print or on screen.

How does the way a text looks affect its readability, that is, how it is being read, processed, and understood? A team at TU Darmstadt’s Centre for Cognitive Science investigated this question at the intersection of perceptual science, cognitive science, and linguistics. Electronic text is even more complex. Texts are read on different devices under different external conditions. And although any digital text is formatted initially, users might resize it on screen, change brightness and contrast of the display, or even select a different font when reading text on the web.

A May 12, 2021 Technische Universitat Darmstadt (Technical University of Damstadt; Germany) press release (also on EurekAlert) provides more detail,

The team of researchers from TU Darmstadt now developed a system that leaves font design to the user’s visual system. First, they needed to come up with a way of synthesizing new fonts. This was achieved by using a machine learning algorithm, which learned the structure of fonts analysing 25 popular and classic typefaces. The system is capable of creating an infinite number of new fonts that are any intermediate form of others – for example, visually halfway between Helvetica and Times New Roman.

Since some fonts may make it more difficult to read the text, they may slow the reader down. Other fonts may help the user read more fluently. Measuring reading speed, a second algorithm can now generate more typefaces that increase the reading speed.

In a laboratory experiment, in which users read texts over one hour, the research team showed that their algorithm indeed generates new fonts that increase individual user’s reading speed. Interestingly all readers had their own personalized font that made reading especially easy for them. However: This individual favorite typeface does not necessarily fit in all situations. “AdaptiFont therefore can be understood as a system which creates fonts for an individual dynamically and continuously while reading, which maximizes the reading speed at the time of use. This may depend on the content of the text, whether you are tired, or perhaps are using different display devices,” explains Professor Constantin A. Rothkopf, Centre for Cognitive Science und head of the institute of Psychology of Information Processing at TU Darmstadt.

The AdaptiFont system was recently presented to the scientific community at the Conference on Human Factors in Computing Systems (CHI). A patent application has been filed. Future possible applications are with all electronic devices on which text is read.

There’s a 5 minute video featuring the work and narration for a researcher who speaks very quickly,

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

AdaptiFont: Increasing Individuals’ Reading Sp0eed with a Generative Font Model and Bayesian Optimization by Florian Kadner, Yannik Keller, Constantin Rothkopf. CHI ’21: Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems May 2021 Article No.: 585 Pages 1-11 DOI: https://doi.org/10.1145/3411764.3445140 Published: 06 May 2021

This paper is open access.

Artificial intelligence is not mentioned but it’s hard to believe that adaptive learning by the software is anything other than a form of AI.

Nanosensors use AI to explore the biomolecular world

EPFL scientists have developed AI-powered nanosensors that let researchers track various kinds of biological molecules without disturbing them. Courtesy: École polytechnique fédérale de Lausanne (EPFL)

If you look at the big orange dot (representing the nanosensors?), you’ll see those purplish/fuschia objects resemble musical notes (biological molecules?). I think that brainlike object to the left and in light blue is the artificial intelligence (AI) component. (If anyone wants to correct my guesses or identify the bits I can’t, please feel free to add to the Comments for this blog.)

Getting back to my topic, keep the ‘musical notes’ in mind as you read about some of the latest research from l’École polytechnique fédérale de Lausanne (EPFL) in an April 7, 2021 news item on Nanowerk,

The tiny world of biomolecules is rich in fascinating interactions between a plethora of different agents such as intricate nanomachines (proteins), shape-shifting vessels (lipid complexes), chains of vital information (DNA) and energy fuel (carbohydrates). Yet the ways in which biomolecules meet and interact to define the symphony of life is exceedingly complex.

Scientists at the Bionanophotonic Systems Laboratory in EPFL’s School of Engineering have now developed a new biosensor that can be used to observe all major biomolecule classes of the nanoworld without disturbing them. Their innovative technique uses nanotechnology, metasurfaces, infrared light and artificial intelligence.

To each molecule its own melody

In this nano-sized symphony, perfect orchestration makes physiological wonders such as vision and taste possible, while slight dissonances can amplify into horrendous cacophonies leading to pathologies such as cancer and neurodegeneration.

An April 7, 2021 EPFL press release, which originated the news item, provides more detail,

“Tuning into this tiny world and being able to differentiate between proteins, lipids, nucleic acids and carbohydrates without disturbing their interactions is of fundamental importance for understanding life processes and disease mechanisms,” says Hatice Altug, the head of the Bionanophotonic Systems Laboratory. 

Light, and more specifically infrared light, is at the core of the biosensor developed by Altug’s team. Humans cannot see infrared light, which is beyond the visible light spectrum that ranges from blue to red. However, we can feel it in the form of heat in our bodies, as our molecules vibrate under the infrared light excitation.

Molecules consist of atoms bonded to each other and – depending on the mass of the atoms and the arrangement and stiffness of their bonds – vibrate at specific frequencies. This is similar to the strings on a musical instrument that vibrate at specific frequencies depending on their length. These resonant frequencies are molecule-specific, and they mostly occur in the infrared frequency range of the electromagnetic spectrum. 

“If you imagine audio frequencies instead of infrared frequencies, it’s as if each molecule has its own characteristic melody,” says Aurélian John-Herpin, a doctoral assistant at Altug’s lab and the first author of the publication. “However, tuning into these melodies is very challenging because without amplification, they are mere whispers in a sea of sounds. To make matters worse, their melodies can present very similar motifs making it hard to tell them apart.” 

Metasurfaces and artificial intelligence

The scientists solved these two issues using metasurfaces and AI. Metasurfaces are man-made materials with outstanding light manipulation capabilities at the nano scale, thereby enabling functions beyond what is otherwise seen in nature. Here, their precisely engineered meta-atoms made out of gold nanorods act like amplifiers of light-matter interactions by tapping into the plasmonic excitations resulting from the collective oscillations of free electrons in metals. “In our analogy, these enhanced interactions make the whispered molecule melodies more audible,” says John-Herpin.

AI is a powerful tool that can be fed with more data than humans can handle in the same amount of time and that can quickly develop the ability to recognize complex patterns from the data. John-Herpin explains, “AI can be imagined as a complete beginner musician who listens to the different amplified melodies and develops a perfect ear after just a few minutes and can tell the melodies apart, even when they are played together – like in an orchestra featuring many instruments simultaneously.” 

The first biosensor of its kind

When the scientists’ infrared metasurfaces are augmented with AI, the new sensor can be used to analyze biological assays featuring multiple analytes simultaneously from the major biomolecule classes and resolving their dynamic interactions. 

“We looked in particular at lipid vesicle-based nanoparticles and monitored their breakage through the insertion of a toxin peptide and the subsequent release of vesicle cargos of nucleotides and carbohydrates, as well as the formation of supported lipid bilayer patches on the metasurface,” says Altug.

This pioneering AI-powered, metasurface-based biosensor will open up exciting perspectives for studying and unraveling inherently complex biological processes, such as intercellular communication via exosomesand the interaction of nucleic acids and carbohydrates with proteins in gene regulation and neurodegeneration. 

“We imagine that our technology will have applications in the fields of biology, bioanalytics and pharmacology – from fundamental research and disease diagnostics to drug development,” says Altug. 

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

Infrared Metasurface Augmented by Deep Learning for Monitoring Dynamics between All Major Classes of Biomolecules by Aurelian John‐Herpin, Deepthy Kavungal. Lea von Mücke, Hatice Altug. Advanced Materials Volume 33, Issue 14 April 8, 2021 2006054 DOI: https://doi.org/10.1002/adma.202006054 First published: 22 February 2021

This paper is open access.

Art, sound, AI, & the Metacreation Lab’s Spring 2021 newsletter

The Metacreation Lab’s Spring 2021 newsletter (received via email) features a number of events either currently taking place or about to take place.

2021 AI Song Contest

2021 marks the 2nd year for this international event, an artificial intelligence/AI Song Contest 2021. The folks at Simon Fraser University’s (SFU) Metacreation Lab have an entry for the 2021 event, A song about the weekend (and you can do whatever you want). Should you click on the song entry, you will find an audio file, a survey/vote consisting of four questions and, if you keep scrolling down, more information about the creative, team, the song and more,

Driven by collaborations involving scientists, experts in artificial intelligence, cognitive sciences, designers, and artists, the Metacreation Lab for Creative AI is at the forefront of the development of generative systems, whether these are embedded in interactive experiences or automating workflows integrated into cutting-edge creative software.

Team:

Cale Plut (Composer and musician) is a PhD Student in the Metacreation lab, researching AI music applications in video games.

Philippe Pasquier (Producer and supervisor) is an Associate Professor, and leads the Metacreation Lab. 

Jeff Ens (AI programmer) is a PhD Candidate in the Metacreation lab, researching AI models for music generation.

Renaud Tchemeube (Producer and interaction designer) is a PhD Student in the Metacreation Lab, researching interaction software design for creativity.

Tara Jadidi (Research Assistant) is an undergraduate student at FUM, Iran, working with the Metacreation lab.

Dimiter Zlatkov (Research Assistant) is an undergraduate student at UBC, working with the Metacreation lab.

ABOUT THE SONG

A song about the weekend (and you can do whatever you want) explores the relationships between AI, humans, labour, and creation in a lighthearted and fun song. It is co-created with the Multi-track Music Machine (MMM)

Through the history of automation and industrialization, the relationship between the labour magnification power of automation and the recipients of the benefits of that magnification have been in contention. While increasing levels of automation are often accompanied by promises of future leisure increases, this rarely materializes for the workers whose labour is multiplied. By primarily using automated methods to create a “fun” song about leisure, we highlight both the promise of AI-human cooperation as well as the disparities in its real-world deployment. 

As for the competition itself, here’s more from the FAQs (frequently asked questions),

What is the AI Song Contest?

AI Song Contest is an international creative AI contest. Teams from all over the world try to create a 4-minute pop song with the help of artificial intelligence.

When and where does it take place?

Between June 1, 2021 and July 1, 2021 voting is open for the international public. On July 6 there will be multiple online panel sessions, and the winner of the AI Song Contest 2021 will be announced in an online award ceremony. All sessions on July 6 are organised in collaboration with Wallifornia MusicTech.

How is the winner determined?

Each participating team will be awarded two sets of points: one a public vote by the contest’s international audience, the other the determination of an expert jury.

Anyone can evaluate as many songs as they like: from one, up to all thirty-eight. Every song can be evaluated only once. Even though it won’t count in the grand total, lyrics can be evaluated too; we do like to determine which team wrote the best accoring to the audience.

Can I vote multiple times for the same team?

No, votes are controlled by IP address. So only one of your votes will count.

Is this the first time the contest is organised?

This is the second time the AI Song Contest is organised. The contest was first initiated in 2020 by Dutch public broadcaster VPRO together with NPO Innovation and NPO 3FM. Teams from Europe and Australia tried to create a Eurovision kind of song with the help of AI. Team Uncanny Valley from Australia won the first edition with their song Beautiful the World. The 2021 edition is organised independently.

What is the definition of artificial intelligence in this contest?

Artificial intelligence is a very broad concept. For this contest it will mean that teams can use techniques such as -but not limited to- machine learning, such as deep learning, natural language processing, algorithmic composition or combining rule-based approaches with neural networks for the creation of their songs. Teams can create their own AI tools, or use existing models and algorithms.  

What are possible challenges?

Read here about the challenges teams from last year’s contest faced.

As an AI researcher, can I collaborate with musicians?

Yes – this is strongly encouraged!

For the 2020 edition, all songs had to be Eurovision-style. Is that also the intention for 2021 entries?

Last year, the first year the contest was organized, it was indeed all about Eurovision. For this year’s competition, we are trying to expand geographically, culturally, and musically. Teams from all over the world can compete, and songs in all genres can be submitted.

If you’re not familiar with Eurovision-style, you can find a compilation video with brief excerpts from the 26 finalists for Eurovision 2021 here (Bill Young’s May 23, 2021 posting on tellyspotting.kera.org; the video runs under 10 mins.). There’s also the “Eurovision Song Contest: The Story of Fire Saga” 2020 movie starring Rachel McAdams, Will Ferrell, and Dan Stevens. It’s intended as a gentle parody but the style is all there.

ART MACHINES 2: International Symposium on Machine Learning and Art 2021

The symposium, Art Machines 2, started yesterday (June 10, 2021 and runs to June 14, 2021) in Hong Kong and SFU’s Metacreation Lab will be represented (from the Spring 2021 newsletter received via email),

On Sunday, June 13 [2021] at 21:45 Hong Kong Standard Time (UTC +8) as part of the Sound Art Paper Session chaired by Ryo Ikeshiro, the Metacreation Lab’s Mahsoo Salimi and Philippe Pasquier will present their paper, Exploiting Swarm Aesthetics in Sound Art. We’ve included a more detailed preview of the paper in this newsletter below.

Concurrent with ART MACHINES 2 is the launch of two exhibitions – Constructing Contexts and System Dreams. Constructing Contexts, curated by Tobias Klein and Rodrigo Guzman-Serrano, will bring together 27 works with unique approaches to the question of contexts as applied by generative adversarial networks. System Dreams highlights work from the latest MFA talent from the School of Creative Media. While the exhibitions take place in Hong Kong, the participating artists and artwork are well documented online.

Liminal Tones: Swarm Aesthetics in Sound Art

Applications of swarm aesthetics in music composition are not new and have already resulted in volumes of complex soundscapes and musical compositions. Using an experimental approach, Mahsoo Salimi and Philippe Pasquier create a series of sound textures know as Liminal Tones (B/ Rain Dream) based on swarming behaviours

Findings of the Liminal Tones project will be presented in papers for the Art Machines 2: International Symposium on Machine Learning (June 10-14 [2021]) and the International Conference on Swarm Intelligence (July 17-21 [2021]).

Talk about Creative AI at the University of British Columbia

This is the last item I’m excerpting from the newsletter. (Should you be curious about what else is listed, you can go to the Metacreation Lab’s contact page and sign up for the newsletter there.) On June 22, 2021 at 2:00 PM PDT, there will be this event,

Creative AI: on the partial or complete automation of creative tasks @ CAIDA

Philippe Pasquier will be giving a talk on creative applications of AI at CAIDA: UBC ICICS Centre for Artificial Intelligence Decision-making and Action. Overviewing the state of the art of computer-assisted creativity and embedded systems and their various applications, the talk will survey the design, deployment, and evaluation of generative systems.

Free registration for the talk is available at the link below.

Register for Creative AI @ CAIDA

Remember, if you want to see the rest of the newsletter, you can sign up at the Metacreation Lab’s contact page.

US Army researchers’ vision for artificial intelligence and ethics

The US Army peeks into a near future where humans and some forms of artificial intelligence (AI) work together in battle and elsewhere. From a February 3, 2021 U.S. Army Research Laboratory news release (also on EurekAlert but published on February 16, 2021),

The Army of the future will involve humans and autonomous machines working together to accomplish the mission. According to Army researchers, this vision will only succeed if artificial intelligence is perceived to be ethical.

Researchers, based at the U.S. Army Combat Capabilities Development Command, now known as DEVCOM, Army Research Laboratory, Northeastern University and the University of Southern California, expanded existing research to cover moral dilemmas and decision making that has not been pursued elsewhere.

This research, featured in Frontiers in Robotics and AI, tackles the fundamental challenge of developing ethical artificial intelligence, which, according to the researchers, is still mostly understudied.

“Autonomous machines, such as automated vehicles and robots, are poised to become pervasive in the Army,” said DEVCOM ARL researcher Dr. Celso de Melo, who is located at the laboratory’s ARL West regional site in Playa Vista, California. “These machines will inevitably face moral dilemmas where they must make decisions that could very well injure humans.”

For example, de Melo said, imagine that an automated vehicle is driving in a tunnel and suddenly five pedestrians cross the street; the vehicle must decide whether to continue moving forward injuring the pedestrians or swerve towards the wall risking the driver.

What should the automated vehicle do in this situation?

Prior work has framed these dilemmas in starkly simple terms, framing decisions as life and death, de Melo said, neglecting the influence of risk of injury to the involved parties on the outcome.

“By expanding the study of moral dilemmas to consider the risk profile of the situation, we significantly expanded the space of acceptable solutions for these dilemmas,” de Melo said. “In so doing, we contributed to the development of autonomous technology that abides to acceptable moral norms and thus is more likely to be adopted in practice and accepted by the general public.”

The researchers focused on this gap and presented experimental evidence that, in a moral dilemma with automated vehicles, the likelihood of making the utilitarian choice – which minimizes the overall injury risk to humans and, in this case, saves the pedestrians – was moderated by the perceived risk of injury to pedestrians and drivers.

In their study, participants were found more likely to make the utilitarian choice with decreasing risk to the driver and with increasing risk to the pedestrians. However, interestingly, most were willing to risk the driver (i.e., self-sacrifice), even if the risk to the pedestrians was lower than to the driver.

As a second contribution, the researchers also demonstrated that participants’ moral decisions were influenced by what other decision makers do – for instance, participants were less likely to make the utilitarian choice, if others often chose the non-utilitarian choice.

“This research advances the state-of-the-art in the study of moral dilemmas involving autonomous machines by shedding light on the role of risk on moral choices,” de Melo said. “Further, both of these mechanisms introduce opportunities to develop AI that will be perceived to make decisions that meet moral standards, as well as introduce an opportunity to use technology to shape human behavior and promote a more moral society.”

For the Army, this research is particularly relevant to Army modernization, de Melo said.

“As these vehicles become increasingly autonomous and operate in complex and dynamic environments, they are bound to face situations where injury to humans is unavoidable,” de Melo said. “This research informs how to navigate these moral dilemmas and make decisions that will be perceived as optimal given the circumstances; for example, minimizing overall risk to human life.”

Moving in to the future, researchers will study this type of risk-benefit analysis in Army moral dilemmas and articulate the corresponding practical implications for the development of AI systems.

“When deployed at scale, the decisions made by AI systems can be very consequential, in particular for situations involving risk to human life,” de Melo said. “It is critical that AI is able to make decisions that reflect society’s ethical standards to facilitate adoption by the Army and acceptance by the general public. This research contributes to realizing this vision by clarifying some of the key factors shaping these standards. This research is personally important because AI is expected to have considerable impact to the Army of the future; however, what kind of impact will be defined by the values reflected in that AI.”

The last time I had an item on a similar topic from the US Army Research Laboratory (ARL) it was in a March 26, 2018 posting; scroll down to the subhead, US Army (about 50% of the way down),

“As machine agents become more sophisticated and independent, it is critical for their human counterparts to understand their intent, behaviors, reasoning process behind those behaviors, and expected outcomes so the humans can properly calibrate their trust [emphasis mine] in the systems and make appropriate decisions,” explained ARL’s Dr. Jessie Chen, senior research psychologist.

This latest work also revolves around the issue of trust according to the last sentence in the 2021 study paper (link and citation to follow),

… Overall, these questions emphasize the importance of the kind of experimental work presented here, as it has the potential to shed light on people’s preferences about moral behavior in machines, inform the design of autonomous machines people are likely to trust and adopt, and, perhaps, even introduce an opportunity to promote a more moral society. [emphases mine]

From trust to adoption to a more moral society—that’s an interesting progression. For another more optimistic view of how robots and AI can have positive impacts there’s my March 29, 2021 posting, Little Lost Robot and humane visions of our technological future

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

Risk of Injury in Moral Dilemmas With Autonomous Vehicles by Celso M. de Melo, Stacy Marsella, and Jonathan Gratch. Front. Robot. AI [Frontiers in Robotics and AI], 20 January 2021 DOI: https://doi.org/10.3389/frobt.2020.572529

This paper is in an open access journal.

Artificial emotional intelligence detection

Sabotage was not my first thought on reading about artificial emotional intelligence so this February 11, 2021 Incheon National University press release (also on EurekAlert) is educational in an unexpected way (Note: A link has been removed),

With the advent of 5G communication technology and its integration with AI, we are looking at the dawn of a new era in which people, machines, objects, and devices are connected like never before. This smart era will be characterized by smart facilities and services such as self-driving cars, smart UAVs [unmanned aerial vehicle], and intelligent healthcare. This will be the aftermath of a technological revolution.

But the flip side of such technological revolution is that AI [artificial intelligence] itself can be used to attack or threaten the security of 5G-enabled systems which, in turn, can greatly compromise their reliability. It is, therefore, imperative to investigate such potential security threats and explore countermeasures before a smart world is realized.

In a recent study published in IEEE Network, a team of researchers led by Prof. Hyunbum Kim from Incheon National University, Korea, address such issues in relation to an AI-based, 5G-integrated virtual emotion recognition system called 5G-I-VEmoSYS, which detects human emotions using wireless signals and body movement. “Emotions are a critical characteristic of human beings and separates humans from machines, defining daily human activity. However, some emotions can also disrupt the normal functioning of a society and put people’s lives in danger, such as those of an unstable driver. Emotion detection technology thus has great potential for recognizing any disruptive emotion and in tandem with 5G and beyond-5G communication, warning others of potential dangers,” explains Prof. Kim. “For instance, in the case of the unstable driver, the AI enabled driver system of the car can inform the nearest network towers, from where nearby pedestrians can be informed via their personal smart devices.”

The virtual emotion system developed by Prof. Kim’s team, 5G-I-VEmoSYS, can recognize at least five kinds of emotion (joy, pleasure, a neutral state, sadness, and anger) and is composed of three subsystems dealing with the detection, flow, and mapping of human emotions. The system concerned with detection is called Artificial Intelligence-Virtual Emotion Barrier, or AI-VEmoBAR, which relies on the reflection of wireless signals from a human subject to detect emotions. This emotion information is then handled by the system concerned with flow, called Artificial Intelligence-Virtual Emotion Flow, or AI-VEmoFLOW, which enables the flow of specific emotion information at a specific time to a specific area. Finally, the Artificial Intelligence-Virtual Emotion Map, or AI-VEmoMAP, utilizes a large amount of this virtual emotion data to create a virtual emotion map that can be utilized for threat detection and crime prevention.

A notable advantage of 5G-I-VEmoSYS is that it allows emotion detection without revealing the face or other private parts of the subjects, thereby protecting the privacy of citizens in public areas. Moreover, in private areas, it gives the user the choice to remain anonymous while providing information to the system. Furthermore, when a serious emotion, such as anger or fear, is detected in a public area, the information is rapidly conveyed to the nearest police department or relevant entities who can then take steps to prevent any potential crime or terrorism threats.

However, the system suffers from serious security issues such as the possibility of illegal signal tampering, abuse of anonymity, and hacking-related cyber-security threats. Further, the danger of sending false alarms to authorities remains.

While these concerns do put the system’s reliability at stake, Prof. Kim’s team are confident that they can be countered with further research. “This is only an initial study. In the future, we need to achieve rigorous information integrity and accordingly devise robust AI-based algorithms that can detect compromised or malfunctioning devices and offer protection against potential system hacks,” explains Prof. Kim, “Only then will it enable people to have safer and more convenient lives in the advanced smart cities of the future.”

Intriguing, yes? The researchers have used this image to illustrate their work,

Caption: With 5G communication technology and new AI-based systems such as emotion recognition systems, smart cities are all set to become a reality; but these systems need to be honed and security issues need to be ironed out before the smart reality can be realized. Credit: macrovector on Freepik

Before getting to the link and citation for the paper, I have a March 8, 2019 article by Meredith Somers for MIT (Massachusetts Institute of Technology) Sloan School of Management’s Ideas Made to Matter publication (Note Links have been removed),

What did you think of the last commercial you watched? Was it funny? Confusing? Would you buy the product? You might not remember or know for certain how you felt, but increasingly, machines do. New artificial intelligence technologies are learning and recognizing human emotions, and using that knowledge to improve everything from marketing campaigns to health care.

These technologies are referred to as “emotion AI.” Emotion AI is a subset of artificial intelligence (the broad term for machines replicating the way humans think) that measures, understands, simulates, and reacts to human emotions. It’s also known as affective computing, or artificial emotional intelligence. The field dates back to at least 1995, when MIT Media lab professor Rosalind Picard published “Affective Computing.”

Javier Hernandez, a research scientist with the Affective Computing Group at the MIT Media Lab, explains emotion AI as a tool that allows for a much more natural interaction between humans and machines.“Think of the way you interact with other human beings; you look at their faces, you look at their body, and you change your interaction accordingly,” Hernandez said. “How can [a machine] effectively communicate information if it doesn’t know your emotional state, if it doesn’t know how you’re feeling, it doesn’t know how you’re going to respond to specific content?”

While humans might currently have the upper hand on reading emotions, machines are gaining ground using their own strengths. Machines are very good at analyzing large amounts of data, explained MIT Sloan professor Erik Brynjolfsson. They can listen to voice inflections and start to recognize when those inflections correlate with stress or anger. Machines can analyze images and pick up subtleties in micro-expressions on humans’ faces that might happen even too fast for a person to recognize.

“We have a lot of neurons in our brain for social interactions. We’re born with some of those skills, and then we learn more. It makes sense to use technology to connect to our social brains, not just our analytical brains.” Brynjolfsson said. “Just like we can understand speech and machines can communicate in speech, we also understand and communicate with humor and other kinds of emotions. And machines that can speak that language — the language of emotions — are going to have better, more effective interactions with us. It’s great that we’ve made some progress; it’s just something that wasn’t an option 20 or 30 years ago, and now it’s on the table.”

Somers describes current uses of emotion AI (I’ve selected two from her list; Note: A link has been removed),

Call centers —Technology from Cogito, a company co-founded in 2007 by MIT Sloan alumni, helps call center agents identify the moods of customers on the phone and adjust how they handle the conversation in real time. Cogito’s voice-analytics software is based on years of human behavior research to identify voice patterns.

Mental health —  In December 2018 Cogito launched a spinoff called CompanionMx, and an accompanying mental health monitoring app. The Companion app listens to someone speaking into their phone, and analyzes the speaker’s voice and phone use for signs of anxiety and mood changes.

The app improves users’ self-awareness, and can increase coping skills including steps for stress reduction. The company has worked with the Department of Veterans Affairs, the Massachusetts General Hospital, and Brigham & Women’s Hospital in Boston.

Somers’ March 8, 2019 article was an eye-opener.

Getting back to the Korean research, here’s a link to and a citation for the paper,

Research Challenges and Security Threats to AI-Driven 5G Virtual Emotion Applications Using Autonomous Vehicles, Drones, and Smart Devices by Hyunbum Kim; Jalel Ben-Othman; Lynda Mokdad; Junggab Son; Chunguo Li. IEEE Network Volume: 34 Issue: 6 November/December 2020 Page(s): 288 – 294 DOI: 10.1109/MNET.011.2000245 Date of Publication (online): 12 October 2020

This paper is behind a paywall.

Getting to be more literate than humans

Lucinda McKnight, lecturer at Deakin University, Australia, has a February 9, 2021 essay about literacy in the coming age of artificial intelligence (AI) for The Conversation (Note 1: You can also find this essay as a February 10, 2021 news item on phys.org; Note 2: Links have been removed),

Students across Australia have started the new school year using pencils, pens and keyboards to learn to write.

In workplaces, machines are also learning to write, so effectively that within a few years they may write better than humans.

Sometimes they already do, as apps like Grammarly demonstrate. Certainly, much everyday writing humans now do may soon be done by machines with artificial intelligence (AI).

The predictive text commonly used by phone and email software is a form of AI writing that countless humans use every day.

According to an industry research organisation Gartner, AI and related technology will automate production of 30% of all content found on the internet by 2022.

Some prose, poetry, reports, newsletters, opinion articles, reviews, slogans and scripts are already being written by artificial intelligence.

Literacy increasingly means and includes interacting with and critically evaluating AI.

This means our children should no longer be taught just formulaic writing. [emphasis mine] Instead, writing education should encompass skills that go beyond the capacities of artificial intelligence.

McKnight’s focus is on how Australian education should approach the coming AI writer ‘supremacy’, from her February 9, 2021 essay (Note: Links have been removed),

In 2019, the New Yorker magazine did an experiment to see if IT company OpenAI’s natural language generator GPT-2 could write an entire article in the magazine’s distinctive style. This attempt had limited success, with the generator making many errors.

But by 2020, GPT-3, the new version of the machine, trained on even more data, wrote an article for The Guardian newspaper with the headline “A robot wrote this entire article. Are you scared yet, human?”

This latest much improved generator has implications for the future of journalism, as the Elon Musk-funded OpenAI invests ever more in research and development.

AI writing is said to have voice but no soul. Human writers, as the New Yorker’s John Seabrook says, give “color, personality and emotion to writing by bending the rules”. Students, therefore, need to learn the rules and be encouraged to break them.

Creativity and co-creativity (with machines) should be fostered. Machines are trained on a finite amount of data, to predict and replicate, not to innovate in meaningful and deliberate ways.

AI cannot yet plan and does not have a purpose. Students need to hone skills in purposeful writing that achieves their communication goals.

AI is not yet as complex as the human brain. Humans detect humor and satire. They know words can have multiple and subtle meanings. Humans are capable of perception and insight; they can make advanced evaluative judgements about good and bad writing.

There are calls for humans to become expert in sophisticated forms of writing and in editing writing created by robots as vital future skills.

… OpenAI’s managers originally refused to release GPT-3, ostensibly because they were concerned about the generator being used to create fake material, such as reviews of products or election-related commentary.

AI writing bots have no conscience and may need to be eliminated by humans, as with Microsoft’s racist Twitter prototype, Tay.

Critical, compassionate and nuanced assessment of what AI produces, management and monitoring of content, and decision-making and empathy with readers are all part of the “writing” roles of a democratic future.

It’s an interesting line of thought and McKnight’s ideas about writing education could be applicable beyond Australia., assuming you accept her basic premise.

I have a few other postings here about AI and writing:

Writing and AI or is a robot writing this blog? a July 16, 2014 posting

AI (artificial intelligence) text generator, too dangerous to release? a February 18, 2019 posting

Automated science writing? a September 16, 2019 posting

It seems I have a lot of question about the automation of any kind of writing.