Tag Archives: Janelle Shane

Ghosts, mechanical turks, and pseudo-AI (artificial intelligence)—Is it all a con game?

There’s been more than one artificial intelligence (AI) story featured here on this blog but the ones featured in this posting are the first I’ve stumbled across that suggest the hype is even more exaggerated than even the most cynical might have thought. (BTW, the 2019 material is later as I have taken a chronological approach to this posting.)

It seems a lot of companies touting their AI algorithms and capabilities are relying on human beings to do the work, from a July 6, 2018 article by Olivia Solon for the Guardian (Note: A link has been removed),

It’s hard to build a service powered by artificial intelligence. So hard, in fact, that some startups have worked out it’s cheaper and easier to get humans to behave like robots than it is to get machines to behave like humans.

“Using a human to do the job lets you skip over a load of technical and business development challenges. It doesn’t scale, obviously, but it allows you to build something and skip the hard part early on,” said Gregory Koberger, CEO of ReadMe, who says he has come across a lot of “pseudo-AIs”.

“It’s essentially prototyping the AI with human beings,” he said.

In 2017, the business expense management app Expensify admitted that it had been using humans to transcribe at least some of the receipts it claimed to process using its “smartscan technology”. Scans of the receipts were being posted to Amazon’s Mechanical Turk crowdsourced labour tool, where low-paid workers were reading and transcribing them.

“I wonder if Expensify SmartScan users know MTurk workers enter their receipts,” said Rochelle LaPlante, a “Turker” and advocate for gig economy workers on Twitter. “I’m looking at someone’s Uber receipt with their full name, pick-up and drop-off addresses.”

Even Facebook, which has invested heavily in AI, relied on humans for its virtual assistant for Messenger, M.

In some cases, humans are used to train the AI system and improve its accuracy. …

The Turk

Fooling people with machines that seem intelligent is not new according to a Sept. 10, 2018 article by Seth Stevenson for Slate.com (Note: Links have been removed),

It’s 1783, and Paris is gripped by the prospect of a chess match. One of the contestants is François-André Philidor, who is considered the greatest chess player in Paris, and possibly the world. Everyone is so excited because Philidor is about to go head-to-head with the other biggest sensation in the chess world at the time.

But his opponent isn’t a man. And it’s not a woman, either. It’s a machine.

This story may sound a lot like Garry Kasparov taking on Deep Blue, IBM’s chess-playing supercomputer. But that was only a couple of decades ago, and this chess match in Paris happened more than 200 years ago. It doesn’t seem like a robot that can play chess would even be possible in the 1780s. This machine playing against Philidor was making an incredible technological leap—playing chess, and not only that, but beating humans at chess.

In the end, it didn’t quite beat Philidor, but the chess master called it one of his toughest matches ever. It was so hard for Philidor to get a read on his opponent, which was a carved wooden figure—slightly larger than life—wearing elaborate garments and offering a cold, mean stare.

It seems like the minds of the era would have been completely blown by a robot that could nearly beat a human chess champion. Some people back then worried that it was black magic, but many folks took the development in stride. …

Debates about the hottest topic in technology today—artificial intelligence—didn’t starts in the 1940s, with people like Alan Turing and the first computers. It turns out that the arguments about AI go back much further than you might imagine. The story of the 18th-century chess machine turns out to be one of those curious tales from history that can help us understand technology today, and where it might go tomorrow.

[In future episodes our podcast, Secret History of the Future] we’re going to look at the first cyberattack, which happened in the 1830s, and find out how the Victorians invented virtual reality.

Philidor’s opponent was known as The Turk or Mechanical Turk and that ‘machine’ was in fact a masterful hoax as The Turk held a hidden compartment from which a human being directed his moves.

People pretending to be AI agents

It seems that today’s AI has something in common with the 18th century Mechanical Turk, there are often humans lurking in the background making things work. From a Sept. 4, 2018 article by Janelle Shane for Slate.com (Note: Links have been removed),

Every day, people are paid to pretend to be bots.

In a strange twist on “robots are coming for my job,” some tech companies that boast about their artificial intelligence have found that at small scales, humans are a cheaper, easier, and more competent alternative to building an A.I. that can do the task.

Sometimes there is no A.I. at all. The “A.I.” is a mockup powered entirely by humans, in a “fake it till you make it” approach used to gauge investor interest or customer behavior. Other times, a real A.I. is combined with human employees ready to step in if the bot shows signs of struggling. These approaches are called “pseudo-A.I.” or sometimes, more optimistically, “hybrid A.I.”

Although some companies see the use of humans for “A.I.” tasks as a temporary bridge, others are embracing pseudo-A.I. as a customer service strategy that combines A.I. scalability with human competence. They’re advertising these as “hybrid A.I.” chatbots, and if they work as planned, you will never know if you were talking to a computer or a human. Every remote interaction could turn into a form of the Turing test. So how can you tell if you’re dealing with a bot pretending to be a human or a human pretending to be a bot?

One of the ways you can’t tell anymore is by looking for human imperfections like grammar mistakes or hesitations. In the past, chatbots had prewritten bits of dialogue that they could mix and match according to built-in rules. Bot speech was synonymous with precise formality. In early Turing tests, spelling mistakes were often a giveaway that the hidden speaker was a human. Today, however, many chatbots are powered by machine learning. Instead of using a programmer’s rules, these algorithms learn by example. And many training data sets come from services like Amazon’s Mechanical Turk, which lets programmers hire humans from around the world to generate examples of tasks like asking and answering questions. These data sets are usually full of casual speech, regionalisms, or other irregularities, so that’s what the algorithms learn. It’s not uncommon these days to get algorithmically generated image captions that read like text messages. And sometimes programmers deliberately add these things in, since most people don’t expect imperfections of an algorithm. In May, Google’s A.I. assistant made headlines for its ability to convincingly imitate the “ums” and “uhs” of a human speaker.

Limited computing power is the main reason that bots are usually good at just one thing at a time. Whenever programmers try to train machine learning algorithms to handle additional tasks, they usually get algorithms that can do many tasks rather badly. In other words, today’s algorithms are artificial narrow intelligence, or A.N.I., rather than artificial general intelligence, or A.G.I. For now, and for many years in the future, any algorithm or chatbot that claims A.G.I-level performance—the ability to deal sensibly with a wide range of topics—is likely to have humans behind the curtain.

Another bot giveaway is a very poor memory. …

Bringing AI to life: ghosts

Sidney Fussell’s April 15, 2019 article for The Atlantic provides more detail about the human/AI interface as found in some Amazon products such as Alexa ( a voice-control system),

… Alexa-enabled speakers can and do interpret speech, but Amazon relies on human guidance to make Alexa, well, more human—to help the software understand different accents, recognize celebrity names, and respond to more complex commands. This is true of many artificial intelligence–enabled products. They’re prototypes. They can only approximate their promised functions while humans help with what Harvard researchers have called “the paradox of automation’s last mile.” Advancements in AI, the researchers write, create temporary jobs such as tagging images or annotating clips, even as the technology is meant to supplant human labor. In the case of the Echo, gig workers are paid to improve its voice-recognition software—but then, when it’s advanced enough, it will be used to replace the hostess in a hotel lobby.

A 2016 paper by researchers at Stanford University used a computer vision system to infer, with 88 percent accuracy, the political affiliation of 22 million people based on what car they drive and where they live. Traditional polling would require a full staff, a hefty budget, and months of work. The system completed the task in two weeks. But first, it had to know what a car was. The researchers paid workers through Amazon’s Mechanical Turk [emphasis mine] platform to manually tag thousands of images of cars, so the system would learn to differentiate between shapes, styles, and colors.

It may be a rude awakening for Amazon Echo owners, but AI systems require enormous amounts of categorized data, before, during, and after product launch. ..,

Isn’t interesting that Amazon also has a crowdsourcing marketplace for its own products. Calling it ‘Mechanical Turk’ after a famous 18th century hoax would suggest a dark sense of humour somewhere in the corporation. (You can find out more about the Amazon Mechanical Turk on this Amazon website and in its Wikipedia entry.0

Anthropologist, Mary L. Gray has coined the phrase ‘ghost work’ for the work that humans perform but for which AI gets the credit. Angela Chan’s May 13, 2019 article for The Verge features Gray as she promotes her latest book with Siddarth Suri ‘Ghost Work: How to Stop Silicon Valley from Building a New Global Underclass’ (Note: A link has been removed),

“Ghost work” is anthropologist Mary L. Gray’s term for the invisible labor that powers our technology platforms. When Gray, a senior researcher at Microsoft Research, first arrived at the company, she learned that building artificial intelligence requires people to manage and clean up data to feed to the training algorithms. “I basically started asking the engineers and computer scientists around me, ‘Who are the people you pay to do this task work of labeling images and classification tasks and cleaning up databases?’” says Gray. Some people said they didn’t know. Others said they didn’t want to know and were concerned that if they looked too closely they might find unsavory working conditions.

So Gray decided to find out for herself. Who are the people, often invisible, who pick up the tasks necessary for these platforms to run? Why do they do this work, and why do they leave? What are their working conditions?

The interview that follows is interesting although it doesn’t seem to me that the question about working conditions is answered in any great detail. However, there is this rather interesting policy suggestion,

If companies want to happily use contract work because they need to constantly churn through new ideas and new aptitudes, the only way to make that a good thing for both sides of that enterprise is for people to be able to jump into that pool. And people do that when they have health care and other provisions. This is the business case for universal health care, for universal education as a public good. It’s going to benefit all enterprise.

I want to get across to people that, in a lot of ways, we’re describing work conditions. We’re not describing a particular type of work. We’re describing today’s conditions for project-based task-driven work. This can happen to everybody’s jobs, and I hate that that might be the motivation because we should have cared all along, as this has been happening to plenty of people. For me, the message of this book is: let’s make this not just manageable, but sustainable and enjoyable. Stop making our lives wrap around work, and start making work serve our lives.

Puts a different spin on AI and work, doesn’t it?

Machine learning, neural networks, and knitting

In a recent (Tuesday, March 6, 2018) live stream ‘conversation’ (‘Science in Canada; Investing in Canadian Innovation’ now published on YouTube) between Canadian Prime Minister, Justin Trudeau, and US science communicator, Bill Nye, at the University of Ottawa, they discussed, amongst many other topics, what AI (artificial intelligence) can and can’t do. They seemed to agree that AI can’t be creative, i.e., write poetry, create works of art, make jokes, etc. A conclusion which is both (in my opinion) true and not true.

There are times when I think the joke may be on us (humans). Take for example this March 6, 2018 story by Alexis Madrigal for The Atlantic magazine (Note: Links have been removed),

SkyKnit: How an AI Took Over an Adult Knitting Community

Ribald knitters teamed up with a neural-network creator to generate new types of tentacled, cozy shapes.

Janelle Shane is a humorist [Note: She describes herself as a “Research Scientist in optics. Plays with neural networks. …” in her Twitter bio.] who creates and mines her material from neural networks, the form of machine learning that has come to dominate the field of artificial intelligence over the last half-decade.

Perhaps you’ve seen the candy-heart slogans she generated for Valentine’s Day: DEAR ME, MY MY, LOVE BOT, CUTE KISS, MY BEAR, and LOVE BUN.

Or her new paint-color names: Parp Green, Shy Bather, Farty Red, and Bull Cream.

Or her neural-net-generated Halloween costumes: Punk Tree, Disco Monster, Spartan Gandalf, Starfleet Shark, and A Masked Box.

Her latest project, still ongoing, pushes the joke into a new, physical realm. Prodded by a knitter on the knitting forum Ravelry, Shane trained a type of neural network on a series of over 500 sets of knitting instructions. Then, she generated new instructions, which members of the Ravelry community have actually attempted to knit.

“The knitting project has been a particularly fun one so far just because it ended up being a dialogue between this computer program and these knitters that went over my head in a lot of ways,” Shane told me. “The computer would spit out a whole bunch of instructions that I couldn’t read and the knitters would say, this is the funniest thing I’ve ever read.”

It appears that the project evolved,

The human-machine collaboration created configurations of yarn that you probably wouldn’t give to your in-laws for Christmas, but they were interesting. The user citikas was the first to post a try at one of the earliest patterns, “reverss shawl.” It was strange, but it did have some charisma.

Shane nicknamed the whole effort “Project Hilarious Disaster.” The community called it SkyKnit.

I’m not sure what’s meant by “community” as mentioned in the previous excerpt. Are we talking about humans only, AI only, or both humans and AI?

Here’s some of what underlies Skyknit (Note: Links have been removed),

The different networks all attempt to model the data they’ve been fed by tuning a vast, funky flowchart. After you’ve created a statistical model that describes your real data, you can also roll the dice and generate new, never-before-seen data of the same kind.

How this works—like, the math behind it—is very hard to visualize because values inside the model can have hundreds of dimensions and we are humble three-dimensional creatures moving through time. But as the neural-network enthusiast Robin Sloan puts it, “So what? It turns out imaginary spaces are useful even if you can’t, in fact, imagine them.”

Out of that ferment, a new kind of art has emerged. Its practitioners use neural networks not to attain practical results, but to see what’s lurking in the these vast, opaque systems. What did the machines learn about the world as they attempted to understand the data they’d been fed? Famously, Google released DeepDream, which produced trippy visualizations that also demonstrated how that type of neural network processed the textures and objects in its source imagery.

Madrigal’s article is well worth reading if you have the time. You can also supplement Madrigal’s piece with an August 9, 2017 article about Janelle Shane’s algorithmic experiments by Jacob Brogan for slate.com.

I found some SkyKnit examples on Ravelry including this one from the Dollybird Workshop,

© Chatelaine

SkyKnit fancy addite rifopshent
by SkyKnit
Published in
Dollybird Workshop
Stitch pattern
February 2018
Suggested yarn
Yarn weight
Fingering (14 wpi) ?
24 stitches and 30 rows = 4 inches
in stockinette stitch
Needle size
US 4 – 3.5 mm


This pattern is available as a free Ravelry download

SkyKnit is a type of machine learning algorithm called an artificial neural network. Its creator, Janelle Shane of AIweirdness.com, gave it 88,000 lines of knitting instructions from Stitch-Maps.com and Ravelry, and it taught itself how to make new patterns. Join the discussion!

SkyKnit seems to have created something that has paralell columns, and is reversible. Perhaps a scarf?

Test-knitting & image courtesy of Chatelaine

Patterns may include notes from testknitters; yarn, needles, and gauge are totally at your discretion.

About the designer
SkyKnit’s favorites include lace, tentacles, and totally not the elimination of the human race.
For more information, see: http://aiweirdness.com/

Shane’s website, aiweirdness.com, is where she posts musings such as this (from a March 2, [?] 2018 posting), Note: A link has been removed,

If you’ve been on the internet today, you’ve probably interacted with a neural network. They’re a type of machine learning algorithm that’s used for everything from language translation to finance modeling. One of their specialties is image recognition. Several companies – including Google, Microsoft, IBM, and Facebook – have their own algorithms for labeling photos. But image recognition algorithms can make really bizarre mistakes.


Microsoft Azure’s computer vision API [application programming interface] added the above caption and tags. But there are no sheep in the image of above. None. I zoomed all the way in and inspected every speck.


I have become quite interested in Shane’s self descriptions such as this one from the aiweirdness.com website,



I train neural networks, a type of machine learning algorithm, to write unintentional humor as they struggle to imitate human datasets. Well, I intend the humor. The neural networks are just doing their best to understand what’s going on. Currently located on the occupied land of the Arapahoe Nation.

As for the joke being on us, I can’t help remembering the Facebook bots that developed their own language (Facebotlish), and were featured in my June 30, 2017 posting, There’s a certain eerieness to it all, which seems an appropriate response in a year celebrating the 200th anniversary of Mary Shelley’s 1818 book, Frankenstein; or, the Modern Prometheus. I’m closing with a video clip from the 1931 movie,

Happy Weekend!