My book Everyday Chaos launched last week. Yay! As part of the launch, I gave some talks and interviews. Here are three of the conversations, three three great interviewers:

Leonard Lopate, WBAI

Hidden Forces podcast

Berkman Klein book talk, and conversation with Joi Ito:

Be the first to comment »

In 1948 when Claude Shannon was inventing information science [pdf] (and, I’d say, information itself), he took as an explanatory example a simple algorithm for predicting the element of a sentence. For example, treating each letter as equiprobable, he came up with sentences such as:

XFOML RXKHRJFFJUJ ZLPWCFWKCYJ FFJEYVKCQSGHYD QPAAMKBZAACIBZLHJQD.

If you instead use the average frequency of each letter, you instead come up with sentences that seem more language-like:

OCRO HLI RGWR NMIELWIS EU LL NBNESEBYA TH EEI ALHENHTTPA OOBTTVA NAH BRL.

At least that one has a reasonable number of vowels.

If you then consider the frequency of letters following other letters—U follows a Q far more frequently than X does—you are practically writing nonsense Latin:

ON IE ANTSOUTINYS ARE T INCTORE ST BE S DEAMY ACHIN D ILONASIVE TUCOOWE AT TEASONARE FUSO TIZIN ANDY TOBE SEACE CTISBE.

Looking not at pairs of letters but triplets Shannon got:

IN NO IST LAT WHEY CRATICT FROURE BIRS GROCID PONDENOME OF DEMONSTURES OF THE REPTAGIN IS REGOACTIONA OF CRE.

Then Shannon changes his units from triplets of letters to triplets of words, and gets:

THE HEAD AND IN FRONTAL ATTACK ON AN ENGLISH WRITER THAT THE CHARACTER OF THIS POINT IS THEREFORE ANOTHER METHOD FOR THE LETTERS THAT THE TIME OF WHO EVER TOLD THE PROBLEM FOR AN UNEXPECTED.

Pretty good! But still gibberish.

Now jump ahead seventy years and try to figure out which pieces of the following story were written by humans and which were generated by a computer:

In a shocking finding, scientist discovered a herd of unicorns living in a remote, previously unexplored valley, in the Andes Mountains. Even more surprising to the researchers was the fact that the unicorns spoke perfect English.

The scientist named the population, after their distinctive horn, Ovid’s Unicorn. These four-horned, silver-white unicorns were previously unknown to science.

Now, after almost two centuries, the mystery of what sparked this odd phenomenon is finally solved.

Dr. Jorge Pérez, an evolutionary biologist from the University of La Paz, and several companions, were exploring the Andes Mountains when they found a small valley, with no other animals or humans. Pérez noticed that the valley had what appeared to be a natural fountain, surrounded by two peaks of rock and silver snow.

Pérez and the others then ventured further into the valley. “By the time we reached the top of one peak, the water looked blue, with some crystals on top,” said Pérez.

“Pérez and his friends were astonished to see the unicorn herd”Pérez and his friends were astonished to see the unicorn herd. These creatures could be seen from the air without having to move too much to see them – they were so close they could touch their horns.

While examining these bizarre creatures the scientists discovered that the creatures also spoke some fairly regular English. Pérez stated, “We can see, for example, that they have a common ‘language,’ something like a dialect or dialectic.”

Dr. Pérez believes that the unicorns may have originated in Argentina, where the animals were believed to be descendants of a lost race of people who lived there before the arrival of humans in those parts of South America.

While their origins are still unclear, some believe that perhaps the creatures were created when a human and a unicorn met each other in a time before human civilization. According to Pérez, “In South America, such incidents seem to be quite common.”

However, Pérez also pointed out that it is likely that the only way of knowing for sure if unicorns are indeed the descendants of a lost alien race is through DNA. “But they seem to be able to communicate in English quite well, which I believe is a sign of evolution, or at least a change in social organization,” said the scientist.

The answer: The first paragraph was written by a human being. The rest was generated by a machine learning system trained on a huge body of text. You can read about it in a fascinating article (pdf of the research paper) by its creators at OpenAI. (Those creators are: Alec Radford, Jeffrey Wu, Rewon Child, David Luan, Dario Amodei, and Ilya Sutskever.)

There are two key differences between this approach and Shannon’s.

First, the new approach analyzed a very large body of documents from the Web. It ingested 45 million pages linked in Reddit comments that got more than three upvotes. After removing duplicates and some other cleanup, the data set was reduced to 8 million Web pages. That is a lot of pages. Of course the use of Reddit, or any one site, can bias the dataset. But one of the aims was to compare this new, huge, dataset to the results from existing sets of text-based data. For that reason, the developers also removed Wikipedia pages from the mix since so many existing datasets rely on those pages, which would smudge the comparisons.

(By the way, a quick google search for any page from before December 2018 mentioning both “Jorge Pérez” and “University of La Paz” turned up nothing. “The AI is constructing, not copy-pasting.”The AI is constructing, not copy-pasting.)

The second distinction from Shannon’s method: the developers used machine learning (ML) to create a neural network, rather than relying on a table of frequencies of words in triplet sequences. ML creates a far, far more complex model that can assess the probability of the next word based on the entire context of its prior uses.

The results can be astounding. While the developers freely acknowledge that the examples they feature are somewhat cherry-picked, they say:

When prompted with topics that are highly represented in the data (Brexit, Miley Cyrus, Lord of the Rings, and so on), it seems to be capable of generating reasonable samples about 50% of the time. The opposite is also true: on highly technical or esoteric types of content, the model can perform poorly.

There are obviously things to worry about as this technology advances. For example, fake news could become the Earth’s most abundant resource. For fear of its abuse, its developers are not releasing the full dataset or model weights. Good!

Nevertheless, the possibilities for research are amazing. And, perhaps most important in the longterm, one by one the human capabilities that we take as unique and distinctive are being shown to be replicable without an engine powered by a miracle.

That may be a false conclusion. Human speech does not consist simply of the utterances we make but the complex intentional and social systems in which those utterances are more than just flavored wind. But ML intends nothing and appreciates nothing. “Nothing matters to ML.”Nothing matters to ML. Nevertheless, knowing that sufficient silicon can duplicate the human miracle should shake our confidence in our species’ special place in the order of things.

(FWIW, my personal theology says that when human specialness is taken as conferring special privilege, any blow to it is a good thing. When that specialness is taken as placing special obligations on us, then at its very worst it’s a helpful illusion.)

6 Comments »

I’ve written before about the need to teach The Kids (also: all of us) not only how to think critically so we can see what we should not believe, but also how to come to belief. That piece, which I now cannot locate, was prompted by danah boyd’s excellent post on the problem with media literacy. Robert Berkman, Outreach, Business Librarian at the University of Rochester and Editor of The Information Advisor’s Guide to Internet Research, asked me how one can go about teaching people how to come to belief. Here’s an edited version of my reply:

I’m afraid I don’t have a good answer. I actually haven’t thought much about how to teach people how to come to belief, beyond arguing for doing this as a social process (the ol’ “knowledge is a network” argument :) I have a pretty good sense of how *not* to do it: the way philosophy teachers relentlessly show how every proposed position can be torn down.

I wonder what we’d learn by taking a literature course as a model — not one that is concerned primarily with critical method, but one that is trying to teach students how to appreciate literature. Or art. The teacher tries to get the students to engage with one another to find what’s worthwhile in a work. Formally, you implicitly teach the value of consistency, elegance of explanation, internal coherence, how well a work clarifies one’s own experience, etc. Those are useful touchstones for coming to belief.

I wouldn’t want to leave students feeling that it’s up to them to come up with an understanding on their own. I’d want them to value the history of interpretation, bringing their critical skills to it. The last thing we need is to make people feel yet more unmoored.

I’m also fond of the orthodox Jewish way of coming to belief, as I, as a non-observant Jew, understand it. You have an unchanging and inerrant text that means nothing until humans interpret it. To interpret it means to be conversant with the scholarly opinions of the great Rabbis, who disagree with one another, often diametrically. Formulating a belief in this context means bringing contemporary intelligence to a question while finding support in the old Rabbis…and always always talking respectfully about those other old Rabbis who disagree with your interpretation. No interpretations are final. Learned contradiction is embraced.

That process has the elements I personally like (being moored to a tradition, respecting those with whom one disagrees, acceptance of the finitude of beliefs, acceptance that they result from a social process), but it’s not going to be very practical outside of Jewish communities if only because it rests on the acceptance of a sacred document, even though it’s one that literally cannot be taken literally; it always requires interpretation.

My point: We do have traditions that aim at enabling us to come to belief. Science is one of them. But there are others. We should learn from them.

TL;DR: I dunno.

2 Comments »

Google PAIR (People + AI Research) has just posted my attempt to explain what fairness looks like when it’s operationalized for a machine learning system. It’s pegged around five “fairness buttons” on the new Google What-If tool, a resource for developers who want to try to figure out what factors (“features” in machine learning talk) are affecting an outcome.

Note that there are far more than five ways to operationalize fairness. The point of the article is that once we are forced to decide exactly what we’re going to count as fair — exactly enough that a machine learning system can implement it — we realize just how freaking complex fairness is. OMG. I broke my brain trying to figure out how to explain some of those ideas, and it took several Google developers (especially James Wexler) and a fine mist of vegetarian broth to restore it even incompletely. Even so, my explanations are less clear than I (or you, I’m sure) would like. But at least there’s no math in them :)

I’ll write more about this at some point, but for me the big take-away is that fairness has had value as a moral concept so far because it is vague enough to allow our intuition to guide us. Machine learning is going to force us to get very specific about it. But we are not yet adept enough at it — e.g., we don’t have a vocabulary for talking about the various varieties — plus we don’t agree about them enough to be able to navigate the shoals. It’s going to be a big mess, but something we have to work through. When we do, we’ll be better at being fair.

Now, about the fact that I am a writer-in-residence at Google. Well, yes I am, and have been for about 6 weeks. It’s a 6 month part-time experiment. My role is to try to explain some of machine learning to people who, like me, lack the technical competence to actually understand it. I’m also supposed to be reflecting in public on what the implications of machine learning might be on our ideas. I am expected to be an independent voice, an outsider on the inside.

So far, it’s been an amazing experience. I’m attached to PAIR, which has developers working on very interesting projects. They are, of course, super-smart, but they have not yet tired of me asking dumb questions that do not seem to be getting smarter over time. So, selfishly, it’s been great for me. And isn’t that all that really matters, hmmm?

Comments Off on Five types of AI fairness

Primavera De Filippi is an expert in blockchain-based tech. She is giving a ThursdAI talk on Plantoid, an event held by Harvard’s Berkman Klein Center for Internet & Society and the MIT Media Lab. Her talk is officially on operational autonomy vs. decisional autonomy, but it’s really about how weird things become when you build a computerized flower that merges AI and the blockchain. For me, a central question of her talk was: Can we have autonomous robots that have legal rights and can own and spend assets, without having to resort to conferring personhood on them the way we have with corporations?

“If a lion could talk, we could not understand him.”
— Ludwig Wittgenstein, Philosophical Investigations, 1953.

“If an algorithm could talk, we could not understand it.”
— Deep learning, Now.

1 Comment »

An earlier draft of Descartes’ Meditations has been discovered, which will inevitably lead to a new round of unfunny jokes under the rubric of “Descartes’ First Draft.” I can’t wait :(

The draft is a big discovery. Camilla Shumaker at Research Frontiers reports that Jeremy Hyman, a philosophy instructor at the University of Arkansas, came across a reference to the manuscript and hied off to a municipal library in Toulouse … a gamble, but he apparently felt he had nothing left Toulouse.

And so it begins…

Comments Off on Here comes a new round of "I think, therefore I am" philosophical Dad jokes

Robert Epstein argues in Aeon against the dominant assumption that the brain is a computer, that it processes information, stores and retrieves memories, etc. That we assume so comes from what I think of as the informationalizing of everything.

The strongest part of his argument is that computers operate on symbolic information, but brains do not. There is no evidence (that I know of, but I’m no expert. On anything) that the brain decomposes visual images into pixels and those pixels into on-offs in a code that represents colors.

In the second half, Epstein tries to prove that the brain isn’t a computer through some simple experiments, such as drawing a dollar bill from memory and while looking at it. Someone committed to the idea that the brain is a computer would probably just conclude that the brain just isn’t a very good computer. But judge for yourself. There’s more to it than I’m presenting here.

Back to Epstein’s first point…

It is of the essence of information that it is independent of its medium: you can encode it into voltage levels of transistors, magnetized dust on tape, or holes in punch cards, and it’s the same information. Therefore, a representation of a brain’s states in another medium should also be conscious. Epstein doesn’t make the following argument, but I will (and I believe I am cribbing it from someone else but I don’t remember who).

Because information is independent of its medium, we could encode it in dust particles swirling clockwise or counter-clockwise; clockwise is an on, and counter is an off. In fact, imagine there’s a dust cloud somewhere in the universe that has 86 billion motes, the number of neurons in the human brain. Imagine the direction of those motes exactly matches the on-offs of your neurons when you first spied the love of your life across the room. Imagine those spins shift but happen to match how your neural states shifted over the next ten seconds of your life. That dust cloud is thus perfectly representing the informational state of your brain as you fell in love. It is therefore experiencing your feelings and thinking your thoughts.

That by itself is absurd. But perhaps you say it is just hard to imagine. Ok, then let’s change it. Same dust cloud. Same spins. But this time we say that clockwise is an off, and the other is an on. Now that dust cloud no longer represents your brain states. It therefore is both experiencing your thoughts and feeling and is not experiencing them at the same time. Aristotle would tell us that that is logically impossible: a thing cannot simultaneously be something and its opposite.

Anyway…

Toward the end of the article, Epstein gets to a crucial point that I was very glad to see him bring up: Thinking is not a brain activity, but the activity of a body engaged in the world. (He cites Anthony Chemero’s Radical Embodied Cognitive Science (2009) which I have not read. I’d trace it back further to Andy Clark, David Chalmers, Eleanor Rosch, Heidegger…). Reducing it to a brain function, and further stripping the brain of its materiality to focus on its “processing” of “information” is reductive without being clarifying.

I came into this debate many years ago already made skeptical of the most recent claims about the causes of consciousness by having some awareness of the series of failed metaphors we have used over the past couple of thousands of years. Epstein puts this well, citing another book I have not read (and another book I’ve consequently just ordered):

In his book In Our Own Image (2015), the artificial intelligence expert George Zarkadakis describes six different metaphors people have employed over the past 2,000 years to try to explain human intelligence.

In the earliest one, eventually preserved in the Bible, humans were formed from clay or dirt, which an intelligent god then infused with its spirit. That spirit ‘explained’ our intelligence – grammatically, at least.

The invention of hydraulic engineering in the 3rd century BCE led to the popularity of a hydraulic model of human intelligence, the idea that the flow of different fluids in the body – the ‘humours’ – accounted for both our physical and mental functioning. The hydraulic metaphor persisted for more than 1,600 years, handicapping medical practice all the while.

By the 1500s, automata powered by springs and gears had been devised, eventually inspiring leading thinkers such as René Descartes to assert that humans are complex machines. In the 1600s, the British philosopher Thomas Hobbes suggested that thinking arose from small mechanical motions in the brain. By the 1700s, discoveries about electricity and chemistry led to new theories of human intelligence – again, largely metaphorical in nature. In the mid-1800s, inspired by recent advances in communications, the German physicist Hermann von Helmholtz compared the brain to a telegraph.

Maybe this time our tech-based metaphor has happened to get it right. But history says we should assume not. We should be very alert to the disanologies, which Epstein helps us with.

Getting this right, or at least not getting it wrong, matters. The most pressing problem with the informationalizing of thought is not that it applies a metaphor, or even that the metaphor is inapt. Rather it’s that this metaphor leads us to a seriously diminished understanding of what it means to be a living, caring creature.

I think.

Hat tip to @JenniferSertl for pointing out the Aeon article.

Comments Off on The brain is not a computer and the world is not information

I’ve been at a two-day workshop sponsored by the Michigan State Uiversity and the National Science Foundation: “Workshop on Trustworthy Algorithmic Decision-Making.” After multiple rounds of rotating through workgroups iterating on five different questions, each group presented its findings — questions, insights, areas of future research.

Seriously, I cannot capture all of this.

Conduct of Data Science

What are the problems?

• Who defines and how do we ensure good practice in data science and machine learning?

Why is the topic important? Because algorithms are important. And they have important real-world effects on people’s lives.

Why is the problem difficult?

• Wrong incentives.

• It can be difficult to generalize practices.

• Best practices may be good for one goal but not another, e.g., efficiency but not social good. Also: Lack of shared concepts and vocabulary.

How to mitigate the problems?

• Change incentives

• Increase communication via vocabularies, translations

• Education through MOOCS, meetups, professional organizations

• Enable and encourage resource sharing: an open source lesson about bias, code sharing, data set sharing

Accountability group

The problem: How to integratively assess the impact of an algorithmic system on the public good? “Integrative” = the impact may be positive and negative and affect systems in complex ways. The impacts may be distributed differently across a population, so you have to think about disparities. These impacts may well change over time

We aim to encourage work that is:

• Aspirationally casual: measuring outcomes causally but not always through randomized control trials.

• The goal is not to shut down algorithms to to make positive contributions that generat solutions.

This is a difficult problem because:

• Lack of variation in accountability, enforcements, and interventions.

• It’s unclear what outcomes should be measure and how. This is context-dependent

• It’s unclear which interventions are the highest priority

Why progress is possible: There’s a lot of good activity in this space. And it’s early in the topic so there’s an ability to significantly influence the field.

What are the barriers for success?

• Incomplete understanding of contexts. So, think it in terms of socio-cultural approaches, and make it interdisciplinary.

• The topic lies between disciplines. So, develop a common language.

• High-level triangulation is difficult. Examine the issues at multiple scales, multiple levels of abstraction. Where you assess accountability may vary depending on what level/aspect you’re looking at.

Handling Uncertainty

The problem: How might we holistically treat and attribute uncertainty through data analysis and decisions systems. Uncertainty exists everywhere in these systems, so we need to consider how it moves through a system. This runs from choosing data sources to presenting results to decision-makers and people impacted by these results, and beyond that its incorporation into risk analysis and contingency planning. It’s always good to know where the uncertainty is coming from so you can address it.

Why difficult:

• Uncertainty arises from many places

• Recognizing and addressing uncertainties is a cyclical process

• End users are bad at evaluating uncertain info and incorporating uncertainty in their thinking.

• Many existing solutions are too computationally expensive to run on large data sets

Progress is possible:

• We have sampling-based solutions that provide a framework.

• Some app communities are recognizing that ignoring uncertainty is reducing the quality of their work

How to evaluate and recognize success?

• A/B testing can show that decision making is better after incorporating uncertainty into analysis

• Statistical/mathematical analysis

Barriers to success

• Cognition: Train users.

• It may be difficult to break this problem into small pieces and solve them individually

• Gaps in theory: many of the problems cannot currently be solved algorithmically.

The presentation ends with a note: “In some cases, uncertainty is a useful tool.” E.g., it can make the system harder to game.

Adversaries, workarounds, and feedback loops

Adversarial examples: add a perturbation to a sample and it disrupts the classification. An adversary tries to find those perturbations to wreck your model. Sometimes this is used not to hack the system so much as to prevent the system from, for example, recognizing your face during a protest.

Feedback loops: A recidivism prediction system says you’re likely to commit further crimes, which sends you to prison, which increases the likelihood that you’ll commit further crimes.

What is the problem: How should a trustworthy algorithm account for adversaries, workarounds, and feedback loops?

Who are the stakeholders?

System designers, users, non-users, and perhaps adversaries.

Why is this a difficult problem?

• It’s hard to define the boundaries of the system

• From whose vantage point do we define adversarial behavior, workarounds, and feedback loops.

Unsolved problems

• How do we reason about the incentives users and non-users have when interacting with systems in unintended ways.

• How do we think about oversight and revision in algorithms with respect to feedback mechanisms

• How do we monitor changes, assess anomalies, and implement safeguards?

• How do we account for stakeholders while preserving rights?

How to recognize progress?

• Mathematical model of how people use the system

• Define goals

• Find stable metrics and monitor them closely

• Proximal metrics. Causality?

• Establish methodologies and see them used

• See a taxonomy of adversarial behavior used in practice

Likely approaches

• Security methodology to anticipating and unintended behaviors and adversarial interactions’. Monitor and measure

• Record and taxonomize adversarial behavior in different domains

• Test . Try to break things.

Barriers

• Hard to anticipate unanticipated behavior

• Hard to define the problem in particular cases.

• Goodhardt’s Law

• Systems are born brittle

• What constitutes adversarial behavior vs. a workaround is subjective.

• Dynamic problem

Algorithms and trust

How do you define and operationalize trust.

The problem: What are the processes through which different stakeholders come to trust an algorithm?

Multiple processes lead to trust.

• Procedural vs. substantive trust: are you looking at the weights of the algorithms (e.g.), or what were the steps to get you there?

• Social vs personal: did you see the algorithm at work, or are you relying on peers?

These pathways are not necessarily predictive of each other.

Stakeholders build truth through multiple lenses and priorities

• the builders of the algorithms

• the people who are affected

• those who oversee the outcomes

Mini case study: a child services agency that does not want to be identified. [All of the following is 100% subject to my injection of errors.]

• The agency uses a predictive algorithm. The stakeholders range from the children needing a family, to NYers as a whole. The agency knew what into the model. “We didn’t buy our algorithm from a black-box vendor.” They trusted the algorithm because they staffed a technical team who had credentials and had experience with ethics…and who they trusted intuitively as good people. Few of these are the quantitative metrics that devs spend their time on. Note that FAT (fairness, accountability, transparency) metrics were not what led to trust.

Temporality:

• Processes that build trust happen over time.

• Trust can change or maybe be repaired over time. “

• The timescales to build social trust are outside the scope of traditional experiments,” although you can perhaps find natural experiments.

Barriers:

• Assumption of reducibility or transfer from subcomponents

• Access to internal stakeholders for interviews and process understanding

• Some elements are very long term

What’s next for this workshop

We generated a lot of scribbles, post-it notes, flip charts, Slack conversations, slide decks, etc. They’re going to put together a whitepaper that goes through the major issues, organizing them, and tries to capture the complexity while helping to make sense of it.

There are weak or no incentives to set appropriate levels of trust

Key takeways:

• Trust is irreducible to FAT metrics alone

• Trust is built over time and should be defined in terms of the temporal process

• Isolating the algorithm as an instantiation misses the socio-technical factors in trust.

Comments Off on [liveblog] Conclusion of Workshop on Trustworthy Algorithmic Decision-Making

I’m at a two-day inter-disciplinary workshop on “Trustworthy Algorithmic Decision-Making” put on by the National Science Foundation and Michigan State University. The 2-page whitepapers
from the participants are online. (Here’s mine.) I may do some live-blogging of the workshops.

Goals:

– Key problems and critical qustionos?

– What to tell pol;icy-makers and others about the impact of these systems?

– Product approaches?

– What ideas, people, training, infrastructure are needed for these approaches?

Excellent diversity of backgrounds: CS, policy, law, library science, a philosopher, more. Good diversity in gender and race. As the least qualified person here, I’m greatly looking forward to the conversations.

Comments Off on Workshop: Trustworthy Algorithmic Decision-Making