Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

A Federal Judge on Biases in the Criminal Justice System.

20 Costanza 03 July 2015 03:17AM

A well-known American federal appellate judge, Alex Kozinski, has written a commentary on systemic biases and institutional myths in the criminal justice system.

The basic thrust of his criticism will be familiar to readers of the sequences and rationalists generally. Lots about cognitive biases (but some specific criticisms of fingerprints and DNA evidence as well). Still, it's interesting that a prominent federal judge -- the youngest when appointed, and later chief of the Ninth Circuit -- would treat some sacred cows of the judiciary so ruthlessly. 

This is specifically a criticism of U.S. criminal justice, but, ceteris paribus, much of it applies not only to other areas of U.S. law, but to legal practices throughout the world as well.

[link] Essay on AI Safety

11 jsteinhardt 26 June 2015 07:42AM

I recently wrote an essay about AI risk, targeted at other academics:

Long-Term and Short-Term Challenges to Ensuring the Safety of AI Systems

I think it might be interesting to some of you, so I am sharing it here. I would appreciate any feedback any of you have, especially from others who do AI / machine learning research.

The Unfriendly Superintelligence next door

38 jacob_cannell 02 July 2015 06:46PM

Markets are powerful decentralized optimization engines - it is known.  Liberals see the free market as a kind of optimizer run amuck, a dangerous superintelligence with simple non-human values that must be checked and constrained by the government - the friendly SI.  Conservatives just reverse the narrative roles.

In some domains, where the incentive structure aligns with human values, the market works well.  In our current framework, the market works best for producing gadgets. It does not work so well for pricing intangible information, and most specifically it is broken when it comes to health.

We treat health as just another gadget problem: something to be solved by pills.  Health is really a problem of knowledge; it is a computational prediction problem.  Drugs are useful only to the extent that you can package the results of new knowledge into a pill and patent it.  If you can't patent it, you can't profit from it.

So the market is constrained to solve human health by coming up with new patentable designs for mass-producible physical objects which go into human bodies.  Why did we add that constraint - thou should solve health, but thou shalt only use pills?  (Ok technically the solutions don't have to be ingestible, but that's a detail.)

The gadget model works for gadgets because we know how gadgets work - we built them, after all.  The central problem with health is that we do not completely understand how the human body works - we did not build it.  Thus we should be using the market to figure out how the body works - completely - and arguably we should be allocating trillions of dollars towards that problem.

The market optimizer analogy runs deeper when we consider the complexity of instilling values into a market.  Lawmakers cannot program the market with goals directly, so instead they attempt to engineer desireable behavior by ever more layers and layers of constraints.  Lawmakers are deontologists.

As an example, consider the regulations on drug advertising.  Big pharma is unsafe - its profit function does not encode anything like "maximize human health and happiness" (which of course itself is an oversimplification).  If allowed to its own devices, there are strong incentives to sell subtly addictive drugs, to create elaborate hyped false advertising campaigns, etc.  Thus all the deontological injunctions.  I take that as a strong indicator of a poor solution - a value alignment failure.

What would healthcare look like in a world where we solved the alignment problem?

To solve the alignment problem, the market's profit function must encode long term human health and happiness.  This really is a mechanism design problem - its not something lawmakers are even remotely trained or qualified for.  A full solution is naturally beyond the scope of a little blog post, but I will sketch out the general idea.

To encode health into a market utility function, first we create financial contracts with an expected value which captures long-term health.  We can accomplish this with a long-term contract that generates positive cash flow when a human is healthy, and negative when unhealthy - basically an insurance contract.  There is naturally much complexity in getting those contracts right, so that they measure what we really want.  But assuming that is accomplished, the next step is pretty simple - we allow those contracts to trade freely on an open market.

There are some interesting failure modes and considerations that are mostly beyond scope but worth briefly mentioning.  This system probably needs to be asymmetric.  The transfers on poor health outcomes should partially go to cover medical payments, but it may be best to have a portion of the wealth simply go to nobody/everybody - just destroyed.

In this new framework, designing and patenting new drugs can still be profitable, but it is now put on even footing with preventive medicine.  More importantly, the market can now actually allocate the correct resources towards long term research.

To make all this concrete, let's use an example of a trillion dollar health question - one that our current system is especially ill-posed to solve:

What are the long-term health effects of abnormally low levels of solar radiation?  What levels of sun exposure are ideal for human health?

This is a big important question, and you've probably read some of the hoopla and debate about vitamin D.  I'm going to soon briefly summarize a general abstract theory, one that I would bet heavily on if we lived in a more rational world where such bets were possible.

In a sane world where health is solved by a proper computational market, I could make enormous - ridiculous really - amounts of money if I happened to be an early researcher who discovered the full health effects of sunlight.  I would bet on my theory simply by buying up contracts for individuals/demographics who had the most health to gain by correcting their sunlight deficiency.  I would then publicize the theory and evidence, and perhaps even raise a heap pile of money to create a strong marketing engine to help ensure that my investments - my patients - were taking the necessary actions to correct their sunlight deficiency.  Naturally I would use complex machine learning models to guide the trading strategy.

Now, just as an example, here is the brief 'pitch' for sunlight.

If we go back and look across all of time, there is a mountain of evidence which more or less screams - proper sunlight is important to health.  Heliotherapy has a long history.

Humans, like most mammals, and most other earth organisms in general, evolved under the sun.  A priori we should expect that organisms will have some 'genetic programs' which take approximate measures of incident sunlight as an input.  The serotonin -> melatonin mediated blue-light pathway is an example of one such light detecting circuit which is useful for regulating the 24 hour circadian rhythm.

The vitamin D pathway has existed since the time of algae such as the Coccolithophore.  It is a multi-stage pathway that can measure solar radiation over a range of temporal frequencies.  It starts with synthesis of fat soluble cholecalciferiol which has a very long half life measured in months. [1] [2]

The rough pathway is:

  • Cholecalciferiol (HL ~ months) becomes 
  • 25(OH)D (HL ~ 15 days) which finally becomes 
  • 1,25(OH)2 D (HL ~ 15 hours)

The main recognized role for this pathway in regards to human health - at least according to the current Wikipedia entry - is to enhance "the internal absorption of calcium, iron, magnesium, phosphate, and zinc".  Ponder that for a moment.

Interestingly, this pathway still works as a general solar clock and radiation detector for carnivores - as they can simply eat the precomputed measurement in their diet.

So, what is a long term sunlight detector useful for?  One potential application could be deciding appropriate resource allocation towards DNA repair.  Every time an organism is in the sun it is accumulating potentially catastrophic DNA damage that must be repaired when the cell next divides.  We should expect that genetic programs would allocate resources to DNA repair and various related activities dependent upon estimates of solar radiation.

I should point out - just in case it isn't obvious - that this general idea does not imply that cranking up the sunlight hormone to insane levels will lead to much better DNA/cellular repair.  There are always tradeoffs, etc.

One other obvious use of a long term sunlight detector is to regulate general strategic metabolic decisions that depend on the seasonal clock - especially for organisms living far from the equator.  During the summer when food is plentiful, the body can expect easy calories.  As winter approaches calories become scarce and frugal strategies are expected.

So first off we'd expect to see a huge range of complex effects showing up as correlations between low vit D levels and various illnesses, and specifically illnesses connected to DNA damage (such as cancer) and or BMI.  

Now it turns out that BMI itself is also strongly correlated with a huge range of health issues.  So the first key question to focus on is the relationship between vit D and BMI.  And - perhaps not surprisingly - there is pretty good evidence for such a correlation [3][4] , and this has been known for a while.

Now we get into the real debate.  Numerous vit D supplement intervention studies have now been run, and the results are controversial.  In general the vit D experts (such as my father, who started the vit D council, and publishes some related research[5]) say that the only studies that matter are those that supplement at high doses sufficient to elevate vit D levels into a 'proper' range which substitutes for sunlight, which in general requires 5000 IU day on average - depending completely on genetics and lifestyle (to the point that any one-size-fits all recommendation is probably terrible).

The mainstream basically ignores all that and funds studies at tiny RDA doses - say 400 IU or less - and then they do meta-analysis over those studies and conclude that their big meta-analysis, unsurprisingly, doesn't show a statistically significant effect.  However, these studies still show small effects.  Often the meta-analysis is corrected for BMI, which of course also tends to remove any vit D effect, to the extent that low vit D/sunlight is a cause of both weight gain and a bunch of other stuff.

So let's look at two studies for vit D and weight loss.

First, this recent 2015 study of 400 overweight Italians (sorry the actual paper doesn't appear to be available yet) tested vit D supplementation for weight loss.  The 3 groups were (0 IU/day, ~1,000 IU / day, ~3,000 IU/day).  The observed average weight loss was (1 kg, 3.8 kg, 5.4 kg). I don't know if the 0 IU group received a placebo.  Regardless, it looks promising.

On the other hand, this 2013 meta-analysis of 9 studies with 1651 adults total (mainly women) supposedly found no significant weight loss effect for vit D.  However, the studies used between 200 IU/day to 1,100 IU/day, with most between 200 to 400 IU.  Five studies used calcium, five also showed weight loss (not necessarily the same - unclear).  This does not show - at all - what the study claims in its abstract.

In general, medical researchers should not be doing statistics.  That is a job for the tech industry.

Now the vit D and sunlight issue is complex, and it will take much research to really work out all of what is going on.  The current medical system does not appear to be handling this well - why?  Because there is insufficient financial motivation.

Is Big Pharma interested in the sunlight/vit D question?  Well yes - but only to the extent that they can create a patentable analogue!  The various vit D analogue drugs developed or in development is evidence that Big Pharma is at least paying attention.  But assuming that the sunlight hypothesis is mainly correct, there is very little profit in actually fixing the real problem.

There is probably more to sunlight that just vit D and serotonin/melatonin.  Consider the interesting correlation between birth month and a number of disease conditions[6].  Perhaps there is a little grain of truth to astrology after all.

Thus concludes my little vit D pitch.  

In a more sane world I would have already bet on the general theory.  In a really sane world it would have been solved well before I would expect to make any profitable trade.  In that rational world you could actually trust health advertising, because you'd know that health advertisers are strongly financially motivated to convince you of things actually truly important for your health.

Instead of charging by the hour or per treatment, like a mechanic, doctors and healthcare companies should literally invest in their patients long-term health, and profit from improvements to long term outcomes.  The sunlight health connection is a trillion dollar question in terms of medical value, but not in terms of exploitable profits in today's reality.  In a properly constructed market, there would be enormous resources allocated to answer these questions, flowing into legions of profit motivated startups that could generate billions trading on computational health financial markets, all without selling any gadgets.

So in conclusion: the market could solve health, but only if we allowed it to and only if we setup appropriate financial mechanisms to encode the correct value function.  This is the UFAI problem next door.


In praise of gullibility?

21 ahbwramc 18 June 2015 04:52AM

I was recently re-reading a piece by Yvain/Scott Alexander called Epistemic Learned Helplessness. It's a very insightful post, as is typical for Scott, and I recommend giving it a read if you haven't already. In it he writes:

When I was young I used to read pseudohistory books; Immanuel Velikovsky's Ages in Chaos is a good example of the best this genre has to offer. I read it and it seemed so obviously correct, so perfect, that I could barely bring myself to bother to search out rebuttals.

And then I read the rebuttals, and they were so obviously correct, so devastating, that I couldn't believe I had ever been so dumb as to believe Velikovsky.

And then I read the rebuttals to the rebuttals, and they were so obviously correct that I felt silly for ever doubting.

And so on for several more iterations, until the labyrinth of doubt seemed inescapable.

He goes on to conclude that the skill of taking ideas seriously - often considered one of the most important traits a rationalist can have - is a dangerous one. After all, it's very easy for arguments to sound convincing even when they're not, and if you're too easily swayed by argument you can end up with some very absurd beliefs (like that Venus is a comet, say).

This post really resonated with me. I've had several experiences similar to what Scott describes, of being trapped between two debaters who both had a convincingness that exceeded my ability to discern truth. And my reaction in those situations was similar to his: eventually, after going through the endless chain of rebuttals and counter-rebuttals, changing my mind at each turn, I was forced to throw up my hands and admit that I probably wasn't going to be able to determine the truth of the matter - at least, not without spending a lot more time investigating the different claims than I was willing to. And so in many cases I ended up adopting a sort of semi-principled stance of agnosticism: unless it was a really really important question (in which case I was sort of obligated to do the hard work of investigating the matter to actually figure out the truth), I would just say I don't know when asked for my opinion.

[Non-exhaustive list of areas in which I am currently epistemically helpless: geopolitics (in particular the Israel/Palestine situation), anthropics, nutrition science, population ethics]

All of which is to say: I think Scott is basically right here, in many cases we shouldn't have too strong of an opinion on complicated matters. But when I re-read the piece recently I was struck by the fact that his whole argument could be summed up much more succinctly (albeit much more pithily) as:

"Don't be gullible."

Huh. Sounds a lot more obvious that way.

Now, don't get me wrong: this is still good advice. I think people should endeavour to not be gullible if at all possible. But it makes you wonder: why did Scott feel the need to write a post denouncing gullibility? After all, most people kind of already think being gullible is bad - who exactly is he arguing against here?

Well, recall that he wrote the post in response to the notion that people should believe arguments and take ideas seriously. These sound like good, LW-approved ideas, but note that unless you're already exceptionally smart or exceptionally well-informed, believing arguments and taking ideas seriously is tantamount to...well, to being gullible. In fact, you could probably think of gullibility as a kind of extreme and pathological form of lightness; a willingness to be swept away by the winds of evidence, no matter how strong (or weak) they may be.

There seems to be some tension here. On the one hand we have an intuitive belief that gullibility is bad; that the proper response to any new claim should be skepticism. But on the other hand we also have some epistemic norms here at LW that are - well, maybe they don't endorse being gullible, but they don't exactly not endorse it either. I'd say the LW memeplex is at least mildly friendly towards the notion that one should believe conclusions that come from convincing-sounding arguments, even if they seem absurd. A core tenet of LW is that we change our mind too little, not too much, and we're certainly all in favour of lightness as a virtue.

Anyway, I thought about this tension for a while and came to the conclusion that I had probably just lost sight of my purpose. The goal of (epistemic) rationality isn't to not be gullible or not be skeptical - the goal is to form correct beliefs, full stop. Terms like gullibility and skepticism are useful to the extent that people tend to be systematically overly accepting or dismissive of new arguments - individual beliefs themselves are simply either right or wrong. So, for example, if we do studies and find out that people tend to accept new ideas too easily on average, then we can write posts explaining why we should all be less gullible, and give tips on how to accomplish this. And if on the other hand it turns out that people actually accept far too few new ideas on average, then we can start talking about how we're all much too skeptical and how we can combat that. But in the end, in terms of becoming less wrong, there's no sense in which gullibility would be intrinsically better or worse than skepticism - they're both just words we use to describe deviations from the ideal, which is accepting only true ideas and rejecting only false ones.

This answer basically wrapped the matter up to my satisfaction, and resolved the sense of tension I was feeling. But afterwards I was left with an additional interesting thought: might gullibility be, if not a desirable end point, then an easier starting point on the path to rationality?

That is: no one should aspire to be gullible, obviously. That would be aspiring towards imperfection. But if you were setting out on a journey to become more rational, and you were forced to choose between starting off too gullible or too skeptical, could gullibility be an easier initial condition?

I think it might be. It strikes me that if you start off too gullible you begin with an important skill: you already know how to change your mind. In fact, changing your mind is in some ways your default setting if you're gullible. And considering that like half the freakin sequences were devoted to learning how to actually change your mind, starting off with some practice in that department could be a very good thing.

I consider myself to be...well, maybe not more gullible than average in absolute terms - I don't get sucked into pyramid scams or send money to Nigerian princes or anything like that. But I'm probably more gullible than average for my intelligence level. There's an old discussion post I wrote a few years back that serves as a perfect demonstration of this (I won't link to it out of embarrassment, but I'm sure you could find it if you looked). And again, this isn't a good thing - to the extent that I'm overly gullible, I aspire to become less gullible (Tsuyoku Naritai!). I'm not trying to excuse any of my past behaviour. But when I look back on my still-ongoing journey towards rationality, I can see that my ability to abandon old ideas at the (relative) drop of a hat has been tremendously useful so far, and I do attribute that ability in part to years of practice at...well, at believing things that people told me, and sometimes gullibly believing things that people told me. Call it epistemic deferentiality, or something - the tacit belief that other people know better than you (especially if they're speaking confidently) and that you should listen to them. It's certainly not a character trait you're going to want to keep as a rationalist, and I'm still trying to do what I can to get rid of it - but as a starting point? You could do worse I think.

Now, I don't pretend that the above is anything more than a plausibility argument, and maybe not a strong one at that. For one I'm not sure how well this idea carves reality at its joints - after all, gullibility isn't quite the same thing as lightness, even if they're closely related. For another, if the above were true, you would probably expect LWer's to be more gullible than average. But that doesn't seem quite right - while LW is admirably willing to engage with new ideas, no matter how absurd they might seem, the default attitude towards a new idea on this site is still one of intense skepticism. Post something half-baked on LW and you will be torn to shreds. Which is great, of course, and I wouldn't have it any other way - but it doesn't really sound like the behaviour of a website full of gullible people.

(Of course, on the other hand it could be that LWer's really are more gullible than average, but they're just smart enough to compensate for it)

Anyway, I'm not sure what to make of this idea, but it seemed interesting and worth a discussion post at least. I'm curious to hear what people think: does any of the above ring true to you? How helpful do you think gullibility is, if it is at all? Can you be "light" without being gullible? And for the sake of collecting information: do you consider yourself to be more or less gullible than average for someone of your intelligence level?

Michigan Meetup Feedback and Planning

7 Zubon 18 June 2015 02:06AM

Our meetup last weekend was at the downtown Ann Arbor Public Library. There were several comments, requests, and discussion items. This discussion topic goes out to attendees, people who might have wanted to attend but didn't, and members of other meetup groups who have suggestions.

1. Several people mentioned having trouble commenting here on the Less Wrong forums. Some functions are restricted by karma, and if you cannot comment, you cannot accummulate karma.

  • Have you verified your e-mail address? This is a common stumbling point.
  • Please try to comment on this post. Restrictions on comments are (moderate certainty) looser than comments on starting posts.
  • If that does not work, please try to comment on a comment on this post. I will add one specifically for this purpose. Restrictions on comments on comments may be (low certainty) looser than starting new comment threads.
  • If someone has already troubleshot new users' problems with commenting, please link.

2. Some people felt intimidated about attending. Prominent community members include programmers, physicists, psychiatrists, philosophy professors, Ph.D.s, and other impressive folks who do not start with P like fanfiction writers. Will I be laughed out of the room if I have not read all the Sequences?

No. Not only is there no minimum requirement to attend, as a group, we are *very excited* about explaining things to people. Our writing can be informationally dense, but our habit of linking to long essays is (often) meant to provide context, not to say, "You must read all the dependencies before you are allowed to talk."

And frankly, we are not that intimidating. Being really impressive makes it easy to become prominent, which via availability bias makes us all look impressive, but our average is way lower than that. And the really impressive people will welcome you to the discussion.

So how can we express this in meetup announcements? I promised to draft a phrasing. Please critique and edit in comments.

Everyone is welcome. There is no minimum in terms of age, education, or reading history. There is no minimum contribution to the community nor requirement to speak. You need not be this tall to ride. If you can read this and are interested in the meetup, we want you to come to the meetup.

3. As part of signalling "be comfortable, you are welcome here," I bought some stim toys from Stimtastic and put them out for whoever might need them. They seemed popular. Comforting, distracting, how did that go for folks? They seemed good for some folks who wanted to do something with their hands, but I was worried that we had a bit much "play" at some points.

Your recommendations on accommodating access needs are welcome. (But I'm not buying chewable stim toys to share; you get to bring your own on those.)

4. The location was sub-optimal. It is a fine meeting space, but the library is under construction, has poor parking options, and does not allow food or drink. Attendees requested somewhere more comfortable, with snacking options. Our previous meeting was at a restaurant, which offers much of that but has more background noise and seemed less socially optimal in terms of coordinating discussion. Prior to that, Michigan meetups had been at Yvain's home.

We moved to Ann Arbor from Livonia because (1) Yvain had been hosting and moved to Ann Arbor, (2) half the Livonia attendees seemed to be Ann Arbor-area folks, and (3) I knew the library had a free meeting room.

Recommendations and volunteers for a meeting site in the area are welcome. I'm in Lansing and not well set up for a group of our size.

5. We had 17 people, although not all at once. It was suggested that we break up into two or more groups for part of the discussion. This is probably a good idea, and it would give more people a chance to participate.

6. Many groups have pre-defined topics or projects. No one leaped at that idea, but we can discuss on here.

7. Rationalist game night game night was another suggestion. I like it. Again, volunteers for hosts are welcome. Many public locations like restaurants are problematic for game nights.

Autism, or early isolation?

17 JonahSinick 17 June 2015 08:52AM

I've often heard LWers describe themselves as having autism, or Asperger's Syndrome (which is no longer considered a valid construct, and was removed from the Diagnostic and Statistical Manual of Mental Disorders two years ago.) This is given as an explanation for various forms of social dysfunction. The suggestion is that such people have a genetic disorder.

I've come to think that the issues are seldom genetic in origin. There's a simpler explanation. LWers are often intellectually gifted. This is conducive to early isolation. In The Outsiders Grady Towers writes:

The single greatest adjustment problem faced by the intellectually gifted, however, is their tendency to become isolated from the rest of humanity. Hollingworth points out that the exceptionally gifted do not deliberately choose isolation, but are forced into it against their wills. These children are not unfriendly or ungregarious by nature. Typically they strive to play with others but their efforts are defeated by the difficulties of the case... Other children do not share their interests, their vocabulary, or their desire to organize activities. [...] Forms of solitary play develop, and these, becoming fixed as habits, may explain the fact that many highly intellectual adults are shy, ungregarious, and unmindful of human relationships, or even misanthropic and uncomfortable in ordinary social intercourse.

Most people pick up a huge amount of tacit social knowledge as children and adolescents, through very frequent interaction with many peers. This is often not true of intellectually gifted people, who usually grew up in relative isolation on account of lack of peers who shared their interests.

They often have the chance to meet others similar to themselves later on in life. One might think that this would resolve the issue. But in many cases intellectually gifted people simply never learn how beneficial it can be to interact with others. For example, the great mathematician Robert Langlands wrote:

Bochner pointed out my existence to Selberg and he invited me over to speak with him at the Institute. I have known Selberg for more than 40 years. We are on cordial terms and our offices have been essentially adjacent for more than 20 years.This is nevertheless the only mathematical conversation I ever had with him. It was a revelation.

At first blush, this seems very strange: much of Langlands' work involves generalizations of Selberg's trace formula. It seems obvious that it would be fruitful for Langlands to have spoken with Selberg about math more than once, especially given that the one conversation that he had was very fruitful! But if one thinks about what their early life experiences must have been like, as a couple of the most brilliant people in the world, it sort of makes sense: they plausibly had essentially nobody to talk to about their interests for many years, and if you go for many years without having substantive conversations with people, you might never get into the habit.

When intellectually gifted people do interact, one often sees cultural clashes, because such people created their own cultures as a substitute for usual cultural acclimation, and share no common background culture. From the inside, one sees other intellectually gifted people, recognizes that they're very odd by mainstream standards, and thinks "these people are freaks!" But at the same time, the people who one sees as freaks see one in the same light, and one is often blind to how unusual one's own behavior is, only in different ways. Thus, one gets trainwreck scenarios, as when I inadvertently offended dozens of people when I made strong criticisms of MIRI and Eliezer back in 2010, just after I joined the LW community.

Grady Towers concludes the essay by writing:

The tragedy is that none of the super high IQ societies created thus far have been able to meet those needs, and the reason for this is simple. None of these groups is willing to acknowledge or come to terms with the fact that much of their membership belong to the psychological walking wounded. This alone is enough to explain the constant schisms that develop, the frequent vendettas, and the mediocre level of their publications. But those are not immutable facts; they can be changed. And the first step in doing so is to see ourselves as we are.

Surprising examples of non-human optimization

18 Jan_Rzymkowski 14 June 2015 05:05PM

I am very much interested in examples of non-human optimization processes producing working, but surprising solutions. What is most fascinating is how they show human approach is often not the only one and much more alien solutions can be found, which humans are just not capable of conceiving. It is very probable, that more and more such solutions will arise and will slowly make big part of technology ununderstandable by humans.

I present following examples and ask for linking more in comments:

1. Nick Bostrom describes efforts in evolving circuits that would produce oscilloscope and frequency discriminator, that yielded very unorthodox designs:
http://www.damninteresting.com/on-the-origin-of-circuits/
http://homepage.ntlworld.com/r.stow1/jb/publications/Bird_CEC2002.pdf (IV. B. Oscillator Experiments; also C. and D. in that section)

2. Algorithms learns to play NES games with some eerie strategies:
https://youtu.be/qXXZLoq2zFc?t=361 (description by Vsause)
http://hackaday.com/2013/04/14/teaching-a-computer-to-play-mario-seemingly-through-voodoo/ (more info)

3. Eurisko finding unexpected way of winning Traveller TCS stratedy game:
http://aliciapatterson.org/stories/eurisko-computer-mind-its-own
http://www.therpgsite.com/showthread.php?t=14095

Philosophical differences

18 ahbwramc 13 June 2015 01:16AM

[Many people have been complaining about the lack of new content on LessWrong lately, so I thought I'd cross-post my latest blog post here in discussion. Feel free to critique the content as much as you like, but please do keep in mind that I wrote this for my personal blog and not with LW in mind specifically, so some parts might not be up to LW standards, whereas others might be obvious to everyone here. In other words...well, be gentle]

---------------------------

You know what’s scarier than having enemy soldiers at your border?

Having sleeper agents within your borders.

Enemy soldiers are malevolent, but they are at least visibly malevolent. You can see what they’re doing; you can fight back against them or set up defenses to stop them. Sleeper agents on the other hand are malevolent and invisible. They are a threat and you don’t know that they’re a threat. So when a sleeper agent decides that it’s time to wake up and smell the gunpowder, not only will you be unable to stop them, but they’ll be in a position to do far more damage than a lone soldier ever could. A single well-placed sleeper agent can take down an entire power grid, or bring a key supply route to a grinding halt, or – in the worst case – kill thousands with an act of terrorism, all without the slightest warning.

Okay, so imagine that your country is in wartime, and that a small group of vigilant citizens has uncovered an enemy sleeper cell in your city. They’ve shown you convincing evidence for the existence of the cell, and demonstrated that the cell is actively planning to commit some large-scale act of violence – perhaps not imminently, but certainly in the near-to-mid-future. Worse, the cell seems to have even more nefarious plots in the offing, possibly involving nuclear or biological weapons.

Now imagine that when you go to investigate further, you find to your surprise and frustration that no one seems to be particularly concerned about any of this. Oh sure, they acknowledge that in theory a sleeper cell could do some damage, and that the whole matter is probably worthy of further study. But by and large they just hear you out and then shrug and go about their day. And when you, alarmed, point out that this is not just a theory – that you have proof that a real sleeper cell is actually operating and making plans right now – they still remain remarkably blase. You show them the evidence, but they either don’t find it convincing, or simply misunderstand it at a very basic level (“A wiretap? But sleeper agents use cellphones, and cellphones are wireless!”). Some people listen but dismiss the idea out of hand, claiming that sleeper cell attacks are “something that only happen in the movies”. Strangest of all, at least to your mind, are the people who acknowledge that the evidence is convincing, but say they still aren’t concerned because the cell isn’t planning to commit any acts of violence imminently, and therefore won’t be a threat for a while. In the end, all of your attempts to raise the alarm are to no avail, and you’re left feeling kind of doubly scared – scared first because you know the sleeper cell is out there, plotting some heinous act, and scared second because you know you won’t be able to convince anyone of that fact before it’s too late to do anything about it.

This is roughly how I feel about AI risk.

You see, I think artificial intelligence is probably the most significant existential threat facing humanity right now. This, to put it mildly, is something of a fringe position in most intellectual circles (although that’s becoming less and less true as time goes on), and I’ll grant that it sounds kind of absurd. But regardless of whether or not you think I’m right to be scared of AI, you can imagine how the fact that AI risk is really hard to explain would make me even more scared about it. Threats like nuclear war or an asteroid impact, while terrifying, at least have the virtue of being simple to understand – it’s not exactly hard to sell people on the notion that a 2km hunk of rock colliding with the planet might be a bad thing. As a result people are aware of these threats and take them (sort of) seriously, and various organizations are (sort of) taking steps to stop them.

AI is different, though. AI is more like the sleeper agents I described above – frighteningly invisible. The idea that AI could be a significant risk is not really on many people’s radar at the moment, and worse, it’s an idea that resists attempts to put it on more people’s radar, because it’s so bloody confusing a topic even at the best of times. Our civilization is effectively blind to this threat, and meanwhile AI research is making progress all the time. We’re on the Titanic steaming through the North Atlantic, unaware that there’s an iceberg out there with our name on it – and the captain is ordering full-speed ahead.

(That’s right, not one but two ominous metaphors. Can you see that I’m serious?)

But I’m getting ahead of myself. I should probably back up a bit and explain where I’m coming from.

Artificial intelligence has been in the news lately. In particular, various big names like Elon Musk, Bill Gates, and Stephen Hawking have all been sounding the alarm in regards to AI, describing it as the greatest threat that our species faces in the 21st century. They (and others) think it could spell the end of humanity – Musk said, “If I had to guess what our biggest existential threat is, it’s probably [AI]”, and Gates said, “I…don’t understand why some people are not concerned [about AI]”.

Of course, others are not so convinced – machine learning expert Andrew Ng said that “I don’t work on not turning AI evil today for the same reason I don’t worry about the problem of overpopulation on the planet Mars”.

In this case I happen to agree with the Musks and Gates of the world – I think AI is a tremendous threat that we need to focus much of our attention on it in the future. In fact I’ve thought this for several years, and I’m kind of glad that the big-name intellectuals are finally catching up.

Why do I think this? Well, that’s a complicated subject. It’s a topic I could probably spend a dozen blog posts on and still not get to the bottom of. And maybe I should spend those dozen-or-so blog posts on it at some point – it could be worth it. But for now I’m kind of left with this big inferential gap that I can’t easily cross. It would take a lot of explaining to explain my position in detail. So instead of talking about AI risk per se in this post, I thought I’d go off in a more meta-direction – as I so often do – and talk about philosophical differences in general. I figured if I couldn’t make the case for AI being a threat, I could at least make the case for making the case for AI being a threat.

(If you’re still confused, and still wondering what the whole deal is with this AI risk thing, you can read a not-too-terrible popular introduction to the subject here, or check out Nick Bostrom’s TED Talk on the topic. Bostrom also has a bestselling book out called Superintelligence. The one sentence summary of the problem would be: how do we get a superintelligent entity to want what we want it to want?)

(Trust me, this is much much harder than it sounds)

So: why then am I so meta-concerned about AI risk? After all, based on the previous couple paragraphs it seems like the topic actually has pretty decent awareness: there are popular internet articles and TED talks and celebrity intellectual endorsements and even bestselling books! And it’s true, there’s no doubt that a ton of progress has been made lately. But we still have a very long way to go. If you had seen the same number of online discussions about AI that I’ve seen, you might share my despair. Such discussions are filled with replies that betray a fundamental misunderstanding of the problem at a very basic level. I constantly see people saying things like “Won’t the AI just figure out what we want?”, or “If the AI gets dangerous why can’t we just unplug it?”, or “The AI can’t have free will like humans, it just follows its programming”, or “lol so you’re scared of Skynet?”, or “Why not just program it to maximize happiness?”.

Having read a lot about AI, these misunderstandings are frustrating to me. This is not that unusual, of course – pretty much any complex topic is going to have people misunderstanding it, and misunderstandings often frustrate me. But there is something unique about the confusions that surround AI, and that’s the extent to which the confusions are philosophical in nature.

Why philosophical? Well, artificial intelligence and philosophy might seem very distinct at first glance, but look closer and you’ll see that they’re connected to one another at a very deep level. Take almost any topic of interest to philosophers – free will, consciousness, epistemology, decision theory, metaethics – and you’ll find an AI researcher looking into the same questions. In fact I would go further and say that those AI researchers are usually doing a better job of approaching the questions. Daniel Dennet said that “AI makes philosophy honest”, and I think there’s a lot of truth to that idea. You can’t write fuzzy, ill-defined concepts into computer code. Thinking in terms of having to program something that actually works takes your head out of the philosophical clouds, and puts you in a mindset of actually answering questions.

All of which is well and good. But the problem with looking at philosophy through the lens of AI is that it’s a two-way street – it means that when you try to introduce someone to the concepts of AI and AI risk, they’re going to be hauling all of their philosophical baggage along with them.

And make no mistake, there’s a lot of baggage. Philosophy is a discipline that’s notorious for many things, but probably first among them is a lack of consensus (I wouldn’t be surprised if there’s not even a consensus among philosophers about how much consensus there is among philosophers). And the result of this lack of consensus has been a kind of grab-bag approach to philosophy among the general public – people see that even the experts are divided, and think that that means they can just choose whatever philosophical position they want.

Want. That’s the key word here. People treat philosophical beliefs not as things that are either true or false, but as choices – things to be selected based on their personal preferences, like picking out a new set of curtains. They say “I prefer to believe in a soul”, or “I don’t like the idea that we’re all just atoms moving around”. And why shouldn’t they say things like that? There’s no one to contradict them, no philosopher out there who can say “actually, we settled this question a while ago and here’s the answer”, because philosophy doesn’t settle things. It’s just not set up to do that. Of course, to be fair people seem to treat a lot of their non-philosophical beliefs as choices as well (which frustrates me to no end) but the problem is particularly pronounced in philosophy. And the result is that people wind up running around with a lot of bad philosophy in their heads.

(Oh, and if that last sentence bothered you, if you’d rather I said something less judgmental like “philosophy I disagree with” or “philosophy I don’t personally happen to hold”, well – the notion that there’s no such thing as bad philosophy is exactly the kind of bad philosophy I’m talking about)

(he said, only 80% seriously)

Anyway, I find this whole situation pretty concerning. Because if you had said to me that in order to convince people of the significance of the AI threat, all we had to do was explain to them some science, I would say: no problem. We can do that. Our society has gotten pretty good at explaining science; so far the Great Didactic Project has been far more successful than it had any right to be. We may not have gotten explaining science down to a science, but we’re at least making progress. I myself have been known to explain scientific concepts to people every now and again, and fancy myself not half-bad at it.

Philosophy, though? Different story. Explaining philosophy is really, really hard. It’s hard enough that when I encounter someone who has philosophical views I consider to be utterly wrong or deeply confused, I usually don’t even bother trying to explain myself – even if it’s someone I otherwise have a great deal of respect for! Instead I just disengage from the conversation. The times I’ve done otherwise, with a few notable exceptions, have only ended in frustration – there’s just too much of a gap to cross in one conversation. And up until now that hasn’t really bothered me. After all, if we’re being honest, most philosophical views that people hold aren’t that important in grand scheme of things. People don’t really use their philosophical views to inform their actions – in fact, probably the main thing that people use philosophy for is to sound impressive at parties.

AI risk, though, has impressed upon me an urgency in regards to philosophy that I’ve never felt before. All of a sudden it’s important that everyone have sensible notions of free will or consciousness; all of a sudden I can’t let people get away with being utterly confused about metaethics.

All of a sudden, in other words, philosophy matters.

I’m not sure what to do about this. I mean, I guess I could just quit complaining, buckle down, and do the hard work of getting better at explaining philosophy. It’s difficult, sure, but it’s not infinitely difficult. I could write blogs posts and talk to people at parties, and see what works and what doesn’t, and maybe gradually start changing a few people’s minds. But this would be a long and difficult process, and in the end I’d probably only be able to affect – what, a few dozen people? A hundred?

And it would be frustrating. Arguments about philosophy are so hard precisely because the questions being debated are foundational. Philosophical beliefs form the bedrock upon which all other beliefs are built; they are the premises from which all arguments start. As such it’s hard enough to even notice that they’re there, let alone begin to question them. And when you do notice them, they often seem too self-evident to be worth stating.

Take math, for example – do you think the number 5 exists, as a number?

Yes? Okay, how about 700? 3 billion? Do you think it’s obvious that numbers just keep existing, even when they get really big?

Well, guess what – some philosophers debate this!

It’s actually surprisingly hard to find an uncontroversial position in philosophy. Pretty much everything is debated. And of course this usually doesn’t matter – you don’t need philosophy to fill out a tax return or drive the kids to school, after all. But when you hold some foundational beliefs that seem self-evident, and you’re in a discussion with someone else who holds different foundational beliefs, which they also think are self-evident, problems start to arise. Philosophical debates usually consist of little more than two people talking past one another, with each wondering how the other could be so stupid as to not understand the sheer obviousness of what they’re saying. And the annoying this is, both participants are correct – in their own framework, their positions probably are obvious. The problem is, we don’t all share the same framework, and in a setting like that frustration is the default, not the exception.

This is not to say that all efforts to discuss philosophy are doomed, of course. People do sometimes have productive philosophical discussions, and the odd person even manages to change their mind, occasionally. But to do this takes a lot of effort. And when I say a lot of effort, I mean a lot of effort. To make progress philosophically you have to be willing to adopt a kind of extreme epistemic humility, where your intuitions count for very little. In fact, far from treating your intuitions as unquestionable givens, as most people do, you need to be treating them as things to be carefully examined and scrutinized with acute skepticism and even wariness. Your reaction to someone having a differing intuition from you should not be “I’m right and they’re wrong”, but rather “Huh, where does my intuition come from? Is it just a featureless feeling or can I break it down further and explain it to other people? Does it accord with my other intuitions? Why does person X have a different intuition, anyway?” And most importantly, you should be asking “Do I endorse or reject this intuition?”. In fact, you could probably say that the whole history of philosophy has been little more than an attempt by people to attain reflective equilibrium among their different intuitions – which of course can’t happen without the willingness to discard certain intuitions along the way when they conflict with others.

I guess what I’m trying to say is: when you’re discussing philosophy with someone and you have a disagreement, your foremost goal should be to try to find out exactly where your intuitions differ. And once you identify that, from there the immediate next step should be to zoom in on your intuitions – to figure out the source and content of the intuition as much as possible. Intuitions aren’t blank structureless feelings, as much as it might seem like they are. With enough introspection intuitions can be explicated and elucidated upon, and described in some detail. They can even be passed on to other people, assuming at least some kind of basic common epistemological framework, which I do think all humans share (yes, even objective-reality-denying postmodernists).

Anyway, this whole concept of zooming in on intuitions seems like an important one to me, and one that hasn’t been emphasized enough in the intellectual circles I travel in. When someone doesn’t agree with some basic foundational belief that you have, you can’t just throw up your hands in despair – you have to persevere and figure out why they don’t agree. And this takes effort, which most people aren’t willing to expend when they already see their debate opponent as someone who’s being willfully stupid anyway. But – needless to say – no one thinks of their positions as being a result of willful stupidity. Pretty much everyone holds beliefs that seem obvious within the framework of their own worldview. So if you want to change someone’s mind with respect to some philosophical question or another, you’re going to have to dig deep and engage with their worldview. And this is a difficult thing to do.

Hence, the philosophical quagmire that we find our society to be in.

It strikes me that improving our ability to explain and discuss philosophy amongst one another should be of paramount importance to most intellectually serious people. This applies to AI risk, of course, but also to many everyday topics that we all discuss: feminism, geopolitics, environmentalism, what have you – pretty much everything we talk about grounds out to philosophy eventually, if you go deep enough or meta enough. And to the extent that we can’t discuss philosophy productively right now, we can’t make progress on many of these important issues.

I think philosophers should – to some extent – be ashamed of the state of their field right now. When you compare philosophy to science it’s clear that science has made great strides in explaining the contents of its findings to the general public, whereas philosophy has not. Philosophers seem to treat their field as being almost inconsequential, as if whatever they conclude at some level won’t matter. But this clearly isn’t true – we need vastly improved discussion norms when it comes to philosophy, and we need far greater effort on the part of philosophers when it comes to explaining philosophy, and we need these things right now. Regardless of what you think about AI, the 21st century will clearly be fraught with difficult philosophical problems – from genetic engineering to the ethical treatment of animals to the problem of what to do with global poverty, it’s obvious that we will soon need philosophical answers, not just philosophical questions. Improvements in technology mean improvements in capability, and that means that things which were once merely thought experiments will be lifted into the realm of real experiments.

I think the problem that humanity faces in the 21st century is an unprecedented one. We’re faced with the task of actually solving philosophy, not just doing philosophy. And if I’m right about AI, then we have exactly one try to get it right. If we don’t, well..

Well, then the fate of humanity may literally hang in the balance.

Epistemic Trust: Clarification

18 abramdemski 13 June 2015 07:29PM

Cross-posted to my blog.


A while ago, I wrote about epistemic trust. The thrust of my argument was that rational argument is often more a function of the group dynamic, as opposed to how rational the individuals in the group are. I assigned meaning to several terms, in order to explain this:

Intellectual honesty: being up-front not just about what you believe, but also why you believe it, what your motivations are in saying it, and the degree to which you have evidence for it.

Intellectual-Honesty Culture: The norm of intellectual honesty. Calling out mistakes and immediately admitting them; feeling comfortable with giving and receiving criticism.

Face Culture: Norms associated with lack of intellectual honesty. In particular, a need to save face when one's statements turn out to be incorrect or irrelevant; the need to make everyone feel included by praising contributions and excusing mistakes.

Intellectual trust: the expectation that others in the discussion have common intellectual goals; that criticism is an attempt to help, rather than an attack. The kind of trust required to take other people's comments at face value rather than being overly concerned with ulterior motives, especially ideological motives. I hypothesized that this is caused largely by ideological common ground, and that this is the main way of achieving intellectual-honesty culture.

There are several subtleties which I did not emphasize last time.

  • Sometimes it's necessary to play at face culture. The skills which go along with face-culture are important. It is generally a good idea to try to make everyone feel included and to praise contributions even if they turn out to be incorrect. It's important to make sure that you do not offend people with criticism. Many people feel that they are under attack when engaged in critical discussion. Wanting to work against this is not an excuse for ignoring it.
  • Face culture is not the error. Being unable to play the right culture at the right time is the error. In my personal experience, I've seen that some people are unable to give up face-culture habits in more academic settings where intellectual honesty is the norm. This causes great strife and heated arguments! There is no gain in playing for face when you're in the midst of an honesty culture, unless you can do it very well and subtly. You gain a lot more face by admitting your mistakes! On the other hand, there's no honor in playing for honesty when face-culture is dominant. This also tends to cause more trouble than it's worth.
  • It's a cultural thing, but it's not just a cultural thing. Some people have personalities much better suited to one culture or the other, while other people are able to switch freely between them. I expect that groups can switch further toward intellectual honesty as a result of establishing intellectual trust, but that is not the only factor. Try to estimate the preferences of the individuals you're dealing with (while keeping in mind that people may surprise you later on).

The Joy of Bias

14 estimator 09 June 2015 07:04PM

What do you feel when you discover that your reasoning is flawed? when you find your recurring mistakes? when you find that you have been doing something wrong for quite a long time?

Many people feel bad. For example, here is a quote from a recent article on LessWrong:

By depicting the self as always flawed, and portraying the aspiring rationalist's job as seeking to find the flaws, the virtue of perfectionism is framed negatively, and is bound to result in negative reinforcement. Finding a flaw feels bad, and in many people that creates ugh fields around actually doing that search, as reported by participants at the Meetup.

But actually, when you find a serious flaw of yours, you should usually jump for joy. Here's why.

continue reading »

View more: Next