Zbetna Fvapynver [rot13]
Grue_Slinky
Yes, perhaps I should’ve been more clear. Learning certain distance functions is a practical solution to some things, so maybe the phrase “distance functions are hard” is too simplistic. What I meant to say is more like
Fully-specified distance functions are hard, over and above the difficulty of formally specifying most things, and it’s often hard to notice this difficulty
This is mostly applicable to Agent Foundations-like research, where we are trying to give a formal model of (some aspect of) how agents work. Sometimes, we can reduce our problem to defining the appropriate distance function, and it can feel like we’ve made some progress, but we haven’t actually gotten anywhere (the first two examples in the post are like this).
The 3rd example, where we are trying to formally verify an ML model against adversarial examples, is a bit different now that I think of it. Here we apparently need transparent, formally-specified distance function if we have any hope of absolutely proving the absence of adversarial examples. And in formal verification, the specification problem often is just philosophically hard like this. So I suppose this example is less insightful, except insofar as it lends extra intuitions for the other class of examples.
Is this open thread not going to be a monthly thing?
FWIW I liked reading the comment threads here, and would be inclined to participate in the future. But that’s just my opinion. I’m curious if more senior people had reasons for not liking the idea?
I often hear about deepfakes—pictures/videos that can be entirely synthesized by a deep learning model and made to look real—and how this could greatly amplify the “fake news” phenomenon and really undermine the ability of the public to actually evaluate evidence.
And this sounds like a well-founded worry, but then I was just thinking, what about Photoshop? That’s existed for over a decade, and for all that time it’s been possible to doctor images to look real. So why should deepfakes be any scarier?
Part of it could be that we can fake videos, not just images, but that can’t be all of it.
I suspect the main reason is that in the future, deepfakes will also be able to fool experts. This does seem like an important threshold.
This raises another question: is it, in fact, impossible to fool experts with Photoshop? Are there fundamental limitations on it that prevent it from being this potent, and this was always understood so people weren’t particularly fearful of it? (FWIW when I learned about Photoshop as a kid I freaked out with Orwellian visions even worse than people have with deepfakes now, and pretty much only relaxed out of conformity. I remain ignorant about the technical details of Photoshop and its capabilities)
But even if deepfakes are bound to cross this threshold (not that it’s a fine line) in a way Photoshop never could, aren’t there also plenty of things which experts have had and do have trouble classifying as real/fake? Wikipedia’s list of hoaxes is extensive, albeit most of those fooled the public rather than experts. But I feel like there are plenty of hoaxes that lasted hundreds of years before being debunked (Shroud of Turin, or maybe fake fossils?).
I guess we’re just used to seeing less hoaxes in modern times. Like, in the past hoaxes abounded, and there often weren’t the proper experts around to debunk them, so probably those times warranted a greater degree of epistemic learned helplessness or something. But since the last century, our forgery-spotting techniques have gotten a lot better while the corresponding forgeries just haven’t kept up, so we just happen to live in a time where the “offense” is relatively weaker than the “defense”, but there’s no particular reason it should stay that way.
I’m really not sure how worried I should be about deepfakes, but having just thought through all that, it does seem like the existence of “evidence” in political discourse is not an all-or-nothing phenomenon. Images/videos will likely come to be trusted less, maybe other things as well if deep learning contributes in other ways to the “offense” more than the “defense”. And maybe things will reach a not-so-much-worse equilibrium. Or maybe not, but the deepfake phenomenon certainly does not seem completely new.
K-complexity: The minimum description length of something (relative to some fixed description language)
Cake-complexity: The minimum description length of something, where the only noun you can use is “cake”
To be clear I unendorsed the idea about a minute after posting because it felt like more of a low-effort shitpost than a constructive idea for understanding the world (and I don’t want to make that a norm on shortform). That said I had in mind that you’re describing the thing to someone who you can’t communicate with beforehand, except there’s common knowledge that you’re forbidden any nouns besides “cake”. In practice I feel like it degenerates to putting all the meaning on adjectives to construct the nouns you’d want to use. E.g. your own “speaking cake” to denote a person, “flat, vertical, compartmentalizing cakes” to denote walls. Of course you’d have to ban any “-like” and “-esque” constructions and similar things, but it’s not clear to me if the boundaries there are too fuzzy to make a good rule set.
Actually, maybe this could be a board game similar to charades. You get a random word such as “elephant”, and you write down a description of it with this constraint. Then the description is gradually read off, and your team tries to guess the word based on the description. It’s inverse to charades in that the reading is monotonous and w/o body language (and could even be done by the other team).
That all seems pretty fair.
If a system is trying to align with idealized reflectively-endorsed values (similar to CEV), then one might expect such values to be coherent.
That’s why I distinguished between the hypotheses of “human utility” and CEV. It is my vague understanding (and I could be wrong) that some alignment researchers see it as their task to align AGI with current humans and their values, thinking the “extrapolation” less important or that it will take care of itself, while others consider extrapolation an important part of the alignment problem. For the former group, human utility is more salient, while the latter probably cares more about the CEV hypothesis (and the arguments you list in favor of it).
Arguably, you can’t fully align with inconsistent preferences
My intuitions tend to agree, but I’m also inclined to ask “why not?” e.g. even if my preferences are absurdly cyclical, but we get AGI to imitate me perfectly (or me + faster thinking + more information), under what sense of the word is it “unaligned” with me? More generally, what is it about these other coherence conditions that prevent meaningful “alignment”? (Maybe it takes a big discursive can of worms, but I actually haven’t seen this discussed on a serious level so I’m quite happy to just read references).
Essentially, I think one should either stick to a more-or-less utility-theoretic framework, or resort to taking a much more empirical approach where human preferences are learned in all their inconsistent detail (without a background assumption such as prospect theory).
That’s still a false dichotomy, but I think it is an appropriate response to many critiques of utility theory.
Hadn’t thought about it this way. Partially updated (but still unsure what I think).
For reference, LeCun discussed his atheoretic/experimentalist views in more depth in this FB debate with Ali Rahimi and also this lecture. But maybe we should distinguish some distinct axes of the experimentalist/theorist divide in DL:
1) Experimentalism/theorism is a more appropriate paradigm for thinking about AI safety
2) Experimentalism/theorism is a more appropriate paradigm for making progress in AI capabilities
Where the LeCun/Russell debate is about (1) and LeCun/Rahimi is about (2). And maybe this is oversimplifying things, since “theorism” may be an overly broad way of describing Russell/Rahimi’s views on safety/capabilities, but I suspect LeCun is “seeing the same ghost”, or in his words (to Rahimi), seeing the same:
kind of attitude that lead the ML community to abandon neural nets for over 10 years, *despite* ample empirical evidence that they worked very well in many situations.
And whether or not Rahimi should be lumped into that “kind of attitude”, I think LeCun is right (from a certain perspective) to want to push back against that attitude.
I’d even go further: given that LeCun has been more successful than Rahimi/Russell in AI research this century, all else equal I would weight the former’s intuitions on research progress more. (I think the best counterargument is that while experimentalism might be better in the short-term, theorism has better payoff in the long-term, but I’m not sure about this.)
In fact, one of my major fears is that LeCun is right about this, because even if he is right about (2), I don’t think that’s good evidence he’s right about (1) since these seem pretty orthogonal. But they don’t look orthogonal until you spend a lot of time reading/thinking about AI safety, which you’re not inclined to do if you already know a lot about AI and assume that knowledge transfers to AI safety.
In other words, the “correct” intuitions (on experimentalism/theorism) for modern AI research might be the opposite of the “correct” intuitions for AI safety. (I would, for instance, predict that if Superintelligence were published during the era of GOFAI, all else equal it would’ve made a bigger splash because AI researchers then were more receptive to abstract theorizing.)
This question also has a negative answer, as witnessed by the example of an ant colony—agent-like behavior without agent-like architecture, produced by a “non-agenty” optimization process of evolution. Nonetheless, a general version of the question remains: If some X exhibits agent-like behavior, does it follow that there exists some interesting physical structure causally upstream of X?
Neat example! But for my part, I’m confused about this last sentence, even after reading the footnote:
An example of such “interesting physical structure” would be an implementation of an optimization architecture.
For one thing, I’m not sure I have much intuition about what is meant by “optimization architecture”. For instance, I would not know how to begin answering the question:
Does optimization behavior imply optimization architecture?
And I have even less of a clue what is intended by “interesting physical structure” (perhaps facetiously, any process that causes agent-like behavior to arise sounds “interesting” for that reason alone).
In your ant colony example, is evolution the “interesting physical structure”, and if so, how is it a physical structure?
I have a bunch of half-baked ideas, most of which are mediocre in expectation and probably not worth investing my time and other’s attention writing up. Some of them probably are decent, but I’m not sure which ones, and the user base is probably as good as any for feedback.
So I’m just going to post them all as replies to this comment. Upvote if they seem promising, downvote if not. Comments encouraged. I reserve the “right” to maintain my inside view, but I wouldn’t make this poll if I didn’t put substantial weight on this community’s opinions.
- EA Hotel Fundraiser 6: Concrete outputs after 17 months by 31 Oct 2019 21:39 UTC; 78 points) (EA Forum;
- EA Hotel Fundraiser 7: Pitch focusing on case studies with counterfactuals by 22 Nov 2019 11:25 UTC; 44 points) (EA Forum;
- 11 Oct 2019 14:03 UTC; 18 points) 's comment on Long-Term Future Fund: August 2019 grant recommendations by (EA Forum;
Upvote this comment (and downvote the others as appropriate) if most of the other ideas don’t seem that fruitful.
By default, I’d mostly take this as a signal of “my time would be better spent working on someone else’s agenda or existing problems that people have posed” but I suppose other alternatives exist, if so comment below.
(1)
A classification of some of the vulnerabilities/issues we might expect AGIs to face because they are potentially open-source, and generally more “transparent” to potential adversaries. For instance, they could face adversarial examples, open-source game theory problems, Dutch books, or weird threats that humans don’t have to deal with. Also, there’s a spectrum from “extreme black box” to “extreme white box” with quite a few plausible milestones along the way, that makes for a certain transparency hierarchy, and it may be helpful to analyze this (or at least take a stab at formulating it).
(2)
[I probably need a better term for this] “Wide-open-source game theory”: Where other agents can not only simulate you, but also figure out “why” you made a given decision. There’s a Standard Objection to this: it’s unfair to compare algorithms in environments where they are judged not only by their actions, but on arbitrary features of their code; to which I say, this isn’t an arbitrary feature. I was thinking about this in the context of how, even if an AGI makes the right decision, we care “why” it did so, i.e. because it’s optimizing for what we want vs. optimizing for human approval for instrumental reasons). I doubt we’ll formalize this “why” anytime soon (see e.g. section 5 of this), but I think semi-formal things can be said about it upon some effort. [I thought of this independently from (1), but I think every level of the “transparency hierarchy” could have its own kind of game theory, much like the “open-source” level clearly does]
(3)
“When and why should we be worried about robustness to distributional shift?”: When reading that section of Concrete Problems, there’s a temptation to just say “this isn’t relevant long-term, since an AGI by definition would have solved that problem”. But adversarial examples and the human safety problems (to the extent we worry about them) both say that in some circumstances we don’t expect this to be solved by default. I’d like to think more about when the naïve “AGI will be smart” intuition applies and when it breaks.
(4)
A post discussing my confusions about Goodhart and Garrabrant’s taxonomy of it. I find myself not completely satisfied with it:
1) “adversarial” seems too broad to be that useful as a category
2) It doesn’t clarify what phenomenon is meant by “Goodhart”; in particular, “regressional” doesn’t feel like something the original law was talking about, and any natural definition of “Goodhart” that includes it seems really broad
3) Whereas “regressional” and “extremal” (and perhaps “causal”) are defined statistically, “adversarial” is defined in terms of agents, and this may have downsides (I’m less sure about this objection)
But I’m also not sure how I’d reclassify it and that task seems hard. Which partially updates me in favor of the Taxonomy being good, but at the very least I feel there’s more to say about it.
(5)
A skeptical take on Part I of “What failure looks like” (3 objections, to summarize briefly: not much evidence so far, not much precedent historically, and “why this, of all the possible axes of differential progress?”) [Unsure if these objections will stand up if written out more fully]
(6)
An analysis of what kinds of differential progress we can expect from stronger ML. Actually, I don’t feel like writing this post, but I just don’t understand why Dai and Christiano, respectively, are particularly concerned about differential progress on the polynomial hierarchy and what’s easy-to-measure vs. hard-to-measure. My gut reaction is “maybe, but why privilege that axis of differential progress of all things”, and I can’t resolve that in my mind without doing a comprehensive analysis of potential “differential progresses” that ML could precipitate. Which, argh, sounds like an exhausting task, but someone should do it?
(7)
A critique of MIRI’s “Fixed Points” paradigm, expanding on some points I made on MIRIxDiscord a while ago (which would take a full post to properly articulate). Main issue is, I’m unsure if it’s still guiding anyone’s research and/or who outside MIRI would care.
(8)
In light of the “Fixed Points” critique, a set of exercises that seem more useful/reflective of MIRI’s research than those exercises. What I have in mind is taking some of the classic success stories of formalized philosophy (e.g. Turing machines, Kolmogorov complexity, Shannon information, Pearlian causality, etc., but this could also be done for reflective oracles and logical induction), introducing the problems they were meant to solve, and giving some stepping stones that guide one to have the intuitions and thoughts that (presumably) had to be developed to make the finished product. I get that this will be hard, but I think this can be feasibly done for some of the (mostly easier) concepts, and if done really well, it could even be a better way for people to learn those concepts than actually reading about them.
*begins drafting longer proposal*
Yeah, this is definitely more high-risk, high-reward than the others, and the fact that there’s potentially some very substantial spillover effects if successful makes me both excited and nervous about the concept. I’m thinking of Arbital as an example of “trying to solve way too many problems at once”, so I want to manage expectations and just try to make some exercises that inspire people to think about the art of mathematizing certain fuzzy philosophical concepts. (Running title is “Formalization Exercises”, but I’m not sure if there’s a better pithy name that captures it).
In any case, I appreciate the feedback, Mr. Entworth.
Yes, here: https://www.lesswrong.com/posts/QePFiEKZ4R2KnxMkW/posts-i-repent-of