Joho the BlogSeptember 2017 - Joho the Blog

September 26, 2017

[liveblog][pair] Blaise Agüera y Arcas on the source of bias

At the PAIR Symposium, Google’s Blaise Agüera y Arcas is providing some intellectual and historical perspective on AI issues.

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

[Note: This is a talk tough to live-blog because it is carefully structured intellectually. My apologies.]

He says neural networks have been part of the computing environment from the beginning. E.g., he thinks that the loop at the end of the logic gate symbol in fact comes from a 1943 symbolization of biological neural networks. There are indications of neural networks in Turing’s early papers. So these ideas go way back. Blaise thinks that the majority of computing processes in a few years will be running on processors designed for running neural networks.

ML has raised anxiety reminiscent of Walter Benjamin’s concern — he cites The Work of Art in the Age of Mechanical Reproduction — about the mass reproduction of art that strips it of its aura. Now there’s the same kind of moral panic about art and human exceptionalism and existence. (Cf. Nick Bostrom’s SuperIntelligence). It reminds him of Jakob Mohr’s 1910 The Influencing Machine in which schizophrenics believe they’re being influenced by an external machine. (They always thought men were managing the machine.) He points to what he calls Bostrom’s ultimate colonialism, in which we are able to populate the universe with 10^58 human minds. [Sorry, but I didn’t get this. My fault.] He ties this to Bacon’s reverence for the domination of nature. Blaise prefers a feminist view, citing Kember & Zylinksa’s Life After New Media.

Many say we have a value alignment problem, he says: how do we make AI that embeds human values? But AI systems do have human values because they’re trained on human data. The problem is that our human values are off. He references a paper on judging criminality based on faces. The paper claims it’s free of human biases. But it’s based on data that is biased. Nevertheless, this sort of tech is being commercialized. E.g., Faception claims to classify people based on their faces: High IQ, Pedophile, etc.

Also, there’s the recent paper about a ML system classifies one’s gender preferences based on faces. Blaise ran a test on Mechanical Turk asking about some of the features in the composite gay and straight faces in that paper. He found that people attracted to the same sex were more likely to wear glasses. There were also significant differences in facial hair, use of makeup, and face tan, features also in the composite faces. Thus, the ML system might have been using social markers, not physiognomy, “There are a lot of tells.”

In conclusion, none of these are arguments against ML. On the contrary. The biases and prejudices, and the social signalling, are things ML lets us hold a mirror up to.

2 Comments »

[liveblog][pair] Golan Levin

At the PAIR Symposium, Golan Levin of CMU is talking about ML and art.

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

The use of computers for serendipitous creativity has been a theme of computer science since its beginning, Golan says. The job of AI should be serendipity and creativity. He gives examples of his projects.

Put your hand up to a scanner and it shows you hand with an extra finger. Or with extra hands at the end of your fingers.

Augmented Hand Series (v.2), Live Screen Recordings from Golan Levin on Vimeo.

[He talks very very quickly. I’ll have to let the project videos talk for themselves. Sorry.]

Terrapattern provides orbital info about us. It’s an open source neural network tool which offers similar-image search for satellite imagery. It’s especially good at finding “soft” structures often not noted on maps. E.g., click on a tennis court and it will find you all of them in the area. Click on crossroads, same thing.

Terrapattern (Overview & Demo) from STUDIO for Creative Inquiry on Vimeo.

This is, he says, an absurdist tool of serendipity. But it also democratizes satellite intelligence. His favorite example: finding all the rusty boats floating in NYC harbor.

Next he talks about our obsession with “masterpieces.” Will a computer ever be able to create masterpiece, he keeps getting asked. But artworks are not in-themselves. They exist in relationship to their audience. (He recommends When the Machine Made Art by Grant D. Taylor.)

Optical illusions get us to see things that aren’t there. “Print on paper beats brain.” We see faces in faucets and life in tree trunks. “This is us deep dreaming.” The people who understand this best are animators. See The illusion of Life, a Disney book about how to make things seem alive.

The observer is not separate from the object observed. Artificial intelligence occurs in the mind as well as in the machine.

He announces a digression: “Some of the best AI-enabled art is being made by engineers,” as computer art was made by early computer engineers.

He points to the color names ML-generated by Janelle Shane. And Gabriel Goh’s synthetic porn. It uses Yahoo’s porn detector and basically runs it in reverse starting with white noise. “This is conceptual art of the highest order.”

“I’m frankly worried, y’all,” he says. People use awful things using imaging technology. E.g., face tracking can be abused by governments and others. These apps are developed to make decisions. And those are the thoughtless explicit abuses, not to mention implicit biases like HP’s face scanning software that doesn’t recognize black faces. He references Zeynep Tufecki’s warnings.

A partial, tiny, and cost-effective solution: integrate artists into your research community. [He lists sensible reasons too fast for me to type.]

Be the first to comment »

[liveblog][PAIR] Rebecca Fiebrink on how machines can create new things

At the PAIR symposium, Rebecca Fiebrink of Goldsmiths University of London asks how machines can create new things.

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

She works with sensors. ML can allow us to build new interactions from examples of human action and computer response. E.g., recognize my closed fist and use it to play some notes. Add more gestures. This is a conventional suprvised training framework. But suppose you want to build a new gesture recognizer?

The first problem is the data set: there isn’t an obvious one to use. Also, would a 99% recognition rate be great or not so much? It depends on what was happening. IF it goes wrong, you modify the training examples.

She gives a live demo — the Wekinator — using a very low-res camera (10×10 pixels maybe) image of her face to control a drum machine. It learns to play stuff based on whether she is leaning to the left or right, and immediately learns to change if she holds up her hand. She then complicates it, starting from scratch again, training it to play based on her hand position. Very impressive.

Ten years ago Rebecca began with the thought that ML can help unlock the interactive potential of sensors. She plays an early piece by Anne Hege using Playstation golf controllers to make music:

Others make music with instruments that don’t look normal. E.g., Laetitia Sonami uses springs as instruments.

She gives other examples. E.g., a facial expression to meme system.

Beyond building new things, what are the consequences, she asks?

First, faster creation means more prototyping and wider exploration, she says.

Second, ML opens up new creative roles for humans. For example, Sonami says, playing an instrument now can be a bit wild, like riding a bull.

Third, ML lets more people be creators and use their own data.

Rebecca teaches a free MOC on Kadenze
: Machine learning for artists and musicians.

Be the first to comment »

[liveblog][PAIR] Doug Eck on creativity

At the PAIR Symposium, Doug Eck, a research scientist at Google Magenta, begins by playing a video:

Douglas Eck – Transforming Technology into Art from Future Of StoryTelling on Vimeo.

Magenta is part of Google Brain that explores creativity.
By the way:

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

He talks about three ideas Magenta has come to for “building a new kind of artist.”

1. Get the right type of data. It’s important to get artists to share and work with them, he says.

Magenta has been trying to get neural networks to compose music. They’ve learned that rather than trying to model musical scores, it’s better to model performances captured as MIDI. They have tens of thousands of performances. From this they were able to build a model that tries to predict the piano roll view of the music. At any moment, should the AI stay at the same time, stacking up notes into chords, or move forward? What are the next notes? Etc. They are not yet capturing much of the “geometry” of, say, Chopin: the piano-roll-ish vision of the score. (He plays music created by ML trained on scores and one trained on performances. The score-based on is clipped. The other is far more fluid and expressive.)

He talks about training ML to draw based on human drawings. He thinks running human artists’ work through ML could point out interesting facets of them.

He points to the playfulness in the drawings created by ML from simple human drawings. ML trained on pig drawings interpreted a drawing of a truck as pig-like.

2. Interfaces that work. Guitar pedals are the perfect interface: they’re indestructible, clear, etc. We should do that for AI musical interfaces, but the sw is so complex technically. He points to the NSyth sound maker and AI duet from Google Creative Lab. (He also touts deeplearn.js.)

3. Learning from users. Can we use feedback from users to improve these systems?

He ends by pointing to the blog, datasets, discussion list, and code at g.co/magenta.

Be the first to comment »

[liveblog][pair] John Zimmerman on UI for AI…and making AI the new UI

At the PAIR symposium, John Zimmerman is giving a great talk on UX for AI. But it relies on graphics that I can’t capture, and I’m about to run out of battery. Sorry :(

Be the first to comment »

[lliveblog][PAIR] Antonio Torralba on machine vision, human vision

At the PAIR Symposium, Antonio Torralba asks why image identification has traditionally gone so wrong.

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

If we train our data on Google Images of bedrooms, we’re training on idealized photos, not real world. It’s a biased set. Likewise for mugs, where the handles in images are almost all on the right side, not the left.

Another issue: The CANNY edge detector (for example) detects edges and throws a black and white reduction to the next level. “All the information is gone!” he says, showing that a messy set of white lines on black is in fact an image of a palace. [Maybe the White House?] (A different example of edge detection:)

/div>

Deep neural networks work well, and can be trained to recognize places in images, e.g., beach. hotel room, street. You train your neural net and it becomes a black box. E.g., how can it recognize that a bedroom is in fact a hotel room? Maybe it’s the lamp? But you trained it to recognize places, not objects. It works but we don’t know how.

When training a system on place detection, we found some units in some layers were in fact doing object detection. It was finding the lamps. Another unit was detecting cars, another detected roads. This lets us interpret the neural networks’ work. In this case, you could put names to more than half of the units.

How to quantify this? How is the representation being built? For this: Network dissection. This shows that when training a network on places, objects emerges. “The network may be doing something more interesting than your task.”The network may be doing something more interesting than your task: object detection is harder than place detection.

We currently train systems by gathering labeled data. But small children learn without labels. Children are self-supervised systems. So, take in the rgb values of frames of a movie, and have the system predict the sounds. When you train a system this way, it kind of works. If you want to predict the ambient sounds of a scene, you have to be able to recognize the objects, e.g., the sound of a car. To solve this, the network has to do object detection. That’s what they found when they looked into the system. It was doing face detection without having been trained to do that. It also detects baby faces, which make a different type of sound. It detects waves. All through self-supervision.

Other examples: On the basis of one segment, predict the next in the sequence. Colorize images. Fill in an empty part of an image. These systems work, and do so by detecting objects without having been trained to do so.

Conclusions: 1. Neural networks build represntations that are sometimes interpretatble. 2. The rep might solve a task that’s evem ore interesting than the primary task. 3. Understanding how these reps are built might allow new approaches for unsupervised or self-supervised training.

Be the first to comment »

[liveblog][PAIR] Maya Gupta on controlling machine learning

At the PAIR symposium. Maya Gupta runs Glass Box at Google, which looks at black box issues. She is talking about how we can control machine learning to do what we want

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

The core idea of machine learning are its role models, i.e., its training data. That’s the best way to control machine learning. She’s going to address by looking at the goals of controlling machine learning.

A simple example of monotinicity. Let’s say we’re tring to recommend nearby coffee shops. So we use data about the happiness of customers and distance from the shop. We can fit the model ot a linear model. Or we can fit it to a curve, which works better for nearby shops but goes wrong for distant shops. That’s fine for Tokyo but terrible for Montana because it’ll be sending people many miles away. A montonic example says we don’t want to do that. This controls ML to make it more useful. Conclusion: the best ML has the right examples and the right kinds of flexibility. [Hard to blog this without her graphics. Sorry.] See “Deep Lattice Networks for Learning Partial Monotonic Models,” NIPS 2017; it will soon by on the TensorFlow site.

“The best way to do things for practitioners is to work next to them”The best way to do things for practitioners is to work next to them.

A fairness goal: e.g., we want to make sure that accuracy in India is the same as accuracy in the US. So, add a constraint that says what accuracy levels we want. Math lets us do that.

Another fairness goal: the rate of positive classifications should be the same in India as in the US, e.g., rate of students being accepted to a college. In one example, there is an accuracy trade-off in order to get fairness. Her attitude: Just tell us what you want and we’ll do it

Fairness isn’t always relative. E.g., E.g., minimize classification errors differently for different regions. You can’t always get what you want, but you sometimes can or can get close. [paraphrase!] See fatml.org

It can be hard to state what we want, but we can look at examples. E.g., someone hand-labels 100 examples. That’s not enough as training date, but we can train the system so that it classifies those 100 at something like 95% accuracy.

Sometimes you want to improve an existing ML system. You don’t want to retrain because you like the old results. So, you can add in a constraint such as keep the differences from the original classifications to less than 2%.

You can put all of the above together. See “Satisfying Real-World Goals with Dataset Constraints,” NIPS, 2016. Look for tools coming to TensorFlow.

Some caveats about this approach.

First, to get results that are the same for men and women, the data needs to come with labels. But sometimes there are privacy issues about that. “Can we make these fairness goals work without labels? ”Can we make these fairness goals work without labels? Research so far says the answer is messy. E.g., if we make ML more fair for gender (because you have gender labels), it may also make it fairer for race.

Second, this approach relies on categories, but individuals don’t always fit into categories. But, usually if you get things right on categories, it usually works out well in the blended examples.

Maya is an optimist about ML. “But we need more work on the steering wheel.” We’re not always sure we want to go with this technology. And we need more human-usable controls.

Be the first to comment »

[liveblog][PAIR] Hae Won Park on living with AI

At the PAIR conference, Hae Won Park of the MIT Media Lab is talking abiut personal social robots for home.

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

Home robots no longer look like R2D2. She shows a 2014 Jibo video.

What sold people is the social dynamic: Jibo is engaged with family members.

She wants to talk about the effect of AI on people’s lives at home.

For example, Google Home changed her morning routine, how she purchases goods, and how she controls her home environment. She shows a 2008 robot called Autom, a weight management coach.

A studied showed that the robot kept people at it longer than using paper or a computer program, and people had the strongest “working alliance” with the robot. They also had emotional engagement with it, personalizing it, giving them names, etc. These users understand it’s just a machine. Why?

She shows a video of Leonardo, a social robot that exhibits bodily cues of emotion. We seem to share a mental model.

They studied how children tell stories and listen to each other. Jin Joo Lee developed a model in which the robot appears to be very attentive as the child tells a story. It notes cues about whether the speaker is engaged. The children were indeed engaged by this reactive behavior.

Researchers have found that social robots activate social thinking, lighting up the social thinking part of the brain. Social modeling occurs between humans and robots too.

Working with children aged 4-6, they studied “growth mindset”: the belief that you can get better if you try hard. Parents and teachers have been shown to affect this. They created a growth mindset robot that plays a game with the child. The robot encourages the child at times determined by a “Boltzmann Machine”en.wikipedia.org/wiki/Boltzmann_machine”>Boltzmann Machine. [Over my head.]

Their researc showed that playing puzzles with a growth-mindset robot fosters that mindset in children. For example, the children tried harder over time.

They also studied early literacy education using personalized robot tutors. In multiple studies of about 120 children. The robot, among other things, encourages the child to tell stories. Over four weeks, they found children more effectively learn vocabulary, and when the robot provided more expressive story telling (rather than speaking in an affect-less TTY voice) the children retained more and would mimic that expressiveness.

Now they’re studying fully automonmous storytelling robots. The robot uses the child’s responses to further engage the child. The children respond more, tell longer stories, and stayed engaged over longer periods across sessions.

We are headed toward a time when robots are more human-centered rather than task focused. So we need to think about making AI not just human-like but humanistic. We hope to make AI that make us better people.

Be the first to comment »

[liveblog][PAIR] Karrie Karahalios

At the Google PAIR conference, Karrie Karahalios is going to talk about how people make sense of their world and lives online. (This is an information-rich talk, and Karrie talks quickly, so this post is extra special unreliable. Sorry. But she’s great. Google her work.)

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

Today, she says, people want to understand how the information they see comes to them. Why does it vary? “Why do you get different answers depending on your wifi network? ”Why do you get different answers depending on your wifi network? These algorithms also affect our personal feeds, e.g., Instagram and Twitter; Twitter articulates it, but doesn’t tell you how it decides what you will see

In 2012, Christian Sandvig and [missed first name] Holbrook were wondering why they were getting odd personalized ads in their feeds. Most people were unaware that their feeds are curated: only 38% were aware of this in 2012. Thsoe who were aware became aware through “folk theories”: non-authoritative explanations that let them make sense of their feed. Four theories:

1. Personal engagement theory: If you like and click on someone, the more of that person you’ll see in your feed. Some people were liking their friends’ baby photos, but got tired of it.

2. Global population theory: If lots of people like, it will show up on more people’s feeds.

3. Narcissist: You’ll see more from people who are like you.

4. Format theory: Some types of things get shared more, e.g., photos or movies. But people didn’t get

Kempton studied thermostats in the 1980s. People either thought of it as a switch or feedback, or as a valve. He looked at their usage patterns. Regardless of which theory, they made it work for them.

She shows an Orbitz page that spits out flights. You see nothing under the hood. But someone found out that if you use a Mac, your prices were higher. People started using designs that shows the seams. So, Karrie’s group created a view that showed the feed and all the content from their network, which was three times bigger than what they saw. For many, this was like awakening from the Matrix. More important, they realized that their friends weren’t “liking” or commenting because the algorithm had kept their friends from seeing what they posted.

Another tool shows who you are seeing posts from and who you are not. This was upsetting for many people.

After going through this process people came up with new folk theories. E.g., they thought it must be FB’s wisdom in stripping out material that’s uninteresting one way or another. [paraphrasing].

They let them configure who they saw, which led many people to say that FB’s algorithm is actually pretty good; there was little to change.

Are these folk theories useful? Only two: personal engagement and control panel, because these let you do something. But there are poor tweaking tools.

How to embrace folk theories: 1. Algorithm probes, to poke and prod. “It would be great, Karrie says, to have open APIs so people could create tools”(It would be great to have open APIs so people could create tools. FB deprecated it.) 2. Seamful interfaces to geneate actionable folk theories. Tuning to revert of borrow?

Another control panel UI, built by Eric Gilbert, uses design to expose the algorithms.

She ends with a wuote form Richard Dyer: “All technolgoies are at once technical and also always social…”

Be the first to comment »

[liveblog][PAIR] Jess Holbrook

I’m at the PAIR conference at Google. Jess Holbrook is UX lead for AI. He’s talking about human-centered machine learning.

NOTE: Live-blogging. Getting things wrong. Missing points. Omitting key information. Introducing artificial choppiness. Over-emphasizing small matters. Paraphrasing badly. Not running a spellpchecker. Mangling other people’s ideas and words. You are warned, people.

“We want to put AI into the maker toolkit, to help you solve real problems.” One of the goals of this: “How do we democratize AI and change what it means to be an expert in this space?” He refers to a blog post he did with Josh Lovejoy about human-centered ML. He emphasizes that we are right at the beginning of figuring this stuff out.

Today, someone finds a data set, and finds a problem that that set could solve. You train a model and look at its performance, and decided if it’s good enough. And then you launch “The world’s first smart X. Next step: profit.” But what if you could do this in a human-centered way?

Human-centered design means: 1. Staying proximate. Know your users. 2. Inclusive divergence: reach out and bring in the right people. 3. Shared definition of success: what does it mean to be done? 4. Make early and often: lots of prototyping. 5. Iterate, test, throw it away.

So, what would a human-centered approach to ML look like? He gives some examples.

Instead of trying to find an application for data, human-centered ML finds a problem and then finds a data set appropriate for that problem. E.g., diagnosis plant diseases. Assemble tagged photos of plants. Or, use ML to personalize a “balancing spoon” for people with Parkinsons.

Today, we find bias in data sets after a problem is discoered. E.g., ProPublica’s article exposing the bias in ML recidivism predictions. Instead, proactively inspect for bias, as per JG’s prior talk.

Today, models personalize experiences, e.g., keyboards that adapt to you. With human-centered ML, people can personalize their models. E.g., someone here created a raccoon detector that uses images he himself took and uploaded, personalized to his particular pet raccoon.

Today, we have to centralize data to get results. “With human-centered ML we’d also have decentralized, federated learning”With human-centered ML we’d also have decentralized, federated learning, getting the benefits while maintaining privacy.

Today there’s a small group of ML experts. [The photo he shows are all white men, pointedly.] With human-centered ML, you get experts who have non-ML domain expertise, which leads to more makers. You can create more diverse, inclusive data sets.

Today, we have narrow training and testing. With human-centered ML, we’ll judge instead by how systems change people’s lives. E.g., ML for the blind to help them recognize things in their environment. Or real-time translation of signs.

Today, we do ML once. E.g., PicDescBot tweets out amusing misfires of image recognition. With human-centered ML we’ll combine ML and teaching. E.g., a human draws an example, and the neural net generates alternatives. In another example, ML improved on landscapes taken by StreetView, where it learned what is an improvement from a data set of professional photos. Google auto-suggest ML also learns from human input. He also shows a video of Simone Giertz, “Queen of the Shitty Robots.”

He references Amanda Case: “Expanding people’s definion of normal” is almost always a gradual process.

[The photo of his team is awesomely diverse.]

Be the first to comment »

Next Page »