I genuinely don’t know what you want elaboration of. Reacts are nice for what they are, but saying something out loud about what you want to hear more about / what’s confusing / what you did and didn’t understand/agree with, is more helpful.
Re/ “to whom not...”, I’m asking Wei: what groups of people would not be described by the list of 6 “underestimating the difficult of philosophy” things? It seems to me that broadly, EAs and “AI alignment” people tend to favor somewhat too concrete touchpoints like “well, suppressing revolts in the past has gone like such and such, so we should try to do similar for AGI”. And broadly they don’t credit an abstract argument about why something won’t work, or would only work given substantial further philosophical insight.
Re/ “don’t think thinking …”, well, if I say “LLMs basically don’t think”, they’re like “sure it does, I can keep prompting it and it says more things, and I can even put that in a scaffold” or “what concrete behavior can you point to that it can’t do”. Like, bro, I’m saying it can’t think. That’s the tweet. What thinking is, isn’t clear, but That thinking is should be presumed, pending a forceful philosophical conceptual replacement!
That is, in fact, a helpful elaboration! When you said
Most people who “work on AI alignment” don’t even think that thinking is a thing.
my leading hypotheses for what you could mean were:
Using thought, as a tool, has not occured to most such people
Most such people have no concept whatsoever of cognition as being a thing, the way people in the year 1000 had no concept whatsoever of javascript being a thing.
Now, instead, my leading hypothesis is that you mean:
Most such people are failing to notice that there’s an important process, called “thinking”, which humans do but LLMs “basically” don’t do.
This is a bunch more precise! For one, it mentions AIs at all.
As my reacts hopefully implied, this is exactly the kind of clarification I needed—thanks!
Like, bro, I’m saying it can’t think. That’s the tweet. What thinking is, isn’t clear, but That thinking is should be presumed, pending a forceful philosophical conceptual replacement!
Sure, but you’re not preaching to the choir at that point. So surely the next step in that particular dance is to stick a knife in the crack and twist?
That is -
“OK, buddy:
Here’s property P (and if you’re good, Q and R and...) that [would have to]/[is/are obviously natural and desirable to]/[is/are pretty clearly a critical part if you want to] characterize ‘thought’ or ‘reasoning’ as distinct from whatever it is LLMs do when they read their own notes as part of a new prompt and keep chewing them up and spitting the result back as part of the new prompt for itself to read.
Here’s thing T (and if you’re good, U and V and...) that an LLM cannot actually do, even in principle, which would be trivially easy for (say) an uploaded (and sane, functional, reasonably intelligent) human H could do, even if H is denied (almost?) all of their previously consolidated memories and just working from some basic procedural memory and whatever Magical thing this ‘thinking’/‘reasoning’ thing is.”
And if neither you nor anyone else can do either of those things… maybe it’s time to give up and say that this ‘thinking’/‘reasoning’ thing is just philosophically confused? I don’t think that that’s where we’re headed, but I find it important to explicitly acknowledge the possibility; I don’t deal in more than one epiphenomenon at a time and I’m partial to Platonism already. So if this ‘reasoning’ thing isn’t meaningfully distinguishable in some observable way from what LLMs do, why shouldn’t I simply give in?
I’ve had this tweet pinned to my Twitter profile for a while, hoping to find some like-minded people, but with 13k views so far I’ve yet to get a positive answer (or find someone expressing this sentiment independently):
Among my first reactions upon hearing “artificial superintelligence” were “I can finally get answers to my favorite philosophical problems” followed by “How do I make sure the ASI actually answers them correctly?”
Anyone else reacted like this?
This aside, there are some people around LW/rationality who seem more cautious/modest/self-critical about proposing new philosophical solutions, like MIRI’s former Agent Foundations team, but perhaps partly as a result of that, they’re now out of a job!
Having worked on some of the problems myself (e.g. decision theory), I think the underlying problems are just very hard. Why do you think they could have done “so much more, much more intently, and much sooner”?
The type of fundamental problem that proper speculative philosophy is supposed to solve is the sort where streetlighting doesn’t work (or isn’t working, or isn’t working fast enough). But nearly all of the alignment field after like 2004 was still basically streetlighting. It was maybe a reasonable thing to have some hope in prospectively, but retrospectively it was too much investment in streetlighting, and retrospectively I can make arguments about why one should have maybe guessed that at the time. By 2018 IIRC, or certainly by 2019, I was vociferously arguing for that in AF team meetings—but the rest of the team either disagreed with me or didn’t understand me, and on my own I’m just not that good a thinker, and I didn’t find anyone else to try it with. I think they have good thoughts, but are nevertheless mostly streetlighting—i.e. not trying to take step after step of thinking at the level of speculative philosophy AND aimed at getting the understanding needed for alignment.
My understanding of what happened (from reading this) is that you wanted to explore in a new direction very different from the then preferred approach of the AF team, but couldn’t convince them (or someone else) to join you. To me this doesn’t clearly have much to do with streetlighting, and my current guess is that it was probably reasonable of them to not be convinced. It was also perfectly reasonable of you to want to explore a different approach, but it seems unreasonable to claim without giving any details that it would have produced better results if only they had listened to you. (I mean you can claim this, but why should I believe you?)
If you disagree (and want to explain more), maybe you could either explain the analogy more fully (e.g., what corresponds to the streetlight, why should I believe that they overexplored the lighted area, what made you able to “see in the dark” to pick out a more promising search area or did you just generally want to explore the dark more) and/or try to convince me on the object level / inside view that your approach is or was more promising?
(Also perfectly fine to stop here if you want. I’m pretty curious on both the object and meta levels about your thoughts on AF, but you may not have wanted to get into such a deep discussion when you first joined this thread.)
Ok, so, there’s this thing about AGI killing everyone. And there’s this idea of avoiding that by making AGI that’s useful like an AGI but doesn’t kill everyone and does stuff we like. And you say you’re working on that, or want to work on that. And what you’re doing day to day is {some math thing, some programming thing, something about decision theory, …}. What is the connection between these things?
and then you listen to what they say, and reask the question and interrogate their answers, IME what it very often grounds out into is something like:
Well, I don’t know what to do to make aligned AI. But it seems like X ϵ {ontology, decision, preference function, NN latent space, logical uncertainty, reasoning under uncertainty, training procedures, negotiation, coordination, interoperability, planning, …} is somehow relevant.
And, I have a formalized version of some small aspect of X in which is mathematically interesting / philosophically intriguing / amenable to testing with a program, and which seems like it’s kinda related to X writ large. So what I’m going to do, is I’m going to tinker with this formalized version for a week/month/year, and then I’m going to zoom out and think about how this relates to X, and what I have and haven’t learned, and so on.
This is a good strategy because this is how all mathematical / scientific / technological progress is made: you start with stuff you know; you expand outwards by following veins of interest, tractability, and generality/power; you keep an eye roughly towards broader goals by selecting the broad region you’re in; and you build outward. What we see historically is that this process tends to lead us to think about the central / key / important / difficult / general problems—such problems show up everywhere, so we convergently will come to address them in due time. By mostly sticking, in our day-to-day work, to things that are relatively more concrete and tractable—though continually pushing and building toward difficult things—we make forward progress, sharpen our skills, and become familiar with the landscape of concepts and questions.
So I would summarize that position as endorsing streetlighting, in a very broad sense that encompasses most math / science / technology. And this position is largely correct! My claim is that
this is probably too slow for making Friendly AI, and
maybe one could go faster by trying to more directly cleave to the core philosophical problems.
(But note that, while that essay frames things as “a proposed solution”, the solution is barely anything—more like a few guesses at pieces of methodology—and the main point is the discussion of the problem; maybe a writing mistake.)
An underemphasized point that I should maybe elaborate more on: a main claim is that there’s untapped guidance to be gotten from our partial understanding—at the philosophical level and for the philosophical level. In other words, our preliminary concepts and intuitions and propositions are, I think, already enough that there’s a lot of progress to be made by having them talk to each other, so to speak.
[2.] maybe one could go faster by trying to more directly cleave to the core philosophical problems.
...
An underemphasized point that I should maybe elaborate more on: a main claim is that there’s untapped guidance to be gotten from our partial understanding—at the philosophical level and for the philosophical level. In other words, our preliminary concepts and intuitions and propositions are, I think, already enough that there’s a lot of progress to be made by having them talk to each other, so to speak.
OK but what would this even look like?\gen
Toss away anything amenable to testing and direct empirical analysis; it’s all too concrete and model-dependent.
Toss away mathsy proofsy approaches; they’re all too formalized and over-rigid and can only prove things from starting assumptions we haven’t got yet and maybe won’t think of in time.
Toss away basically all settled philosophy, too; if there were answers to be had there rather than a few passages which ask correct questions, the Vienna Circle would have solved alignment for us.
What’s left? And what causes it to hang together? And what causes it not to vanish up its own ungrounded self-reference?
What makes you think there are any such ‘answers’, renderable in a form that you could identify?
And even if they do exist, why do you think a human being could fully grasp the explanation in finite time?
Edit: It seems quite possible that even the simplest such ‘answers’ could require many years of full time effort to understand, putting it beyond most if not all human memory capacity. i.e. By the end even those who ‘learned’ it will have forgotten many parts near the beginning.
(Upvoted since your questions seem reasonable and I’m not sure why you got downvoted.)
I see two ways to achieve some justifiable confidence in philosophical answers produced by superintelligent AI:
Solve metaphilosophy well enough that we achieve an understanding of philosophical reasoning on par with mathematical reason, and have ideas/systems analogous to formal proofs and mechanical proof checkers that we can use to check the ASI’s arguments.
We increase our own intelligence and philosophical competence until we can verify the ASI’s reasoning ourselves.
To whom does this not apply? Most people who “work on AI alignment” don’t even think that thinking is a thing.
@Nate Showell @P. @Tetraspace @Joseph Miller @Lorxus
I genuinely don’t know what you want elaboration of. Reacts are nice for what they are, but saying something out loud about what you want to hear more about / what’s confusing / what you did and didn’t understand/agree with, is more helpful.
Re/ “to whom not...”, I’m asking Wei: what groups of people would not be described by the list of 6 “underestimating the difficult of philosophy” things? It seems to me that broadly, EAs and “AI alignment” people tend to favor somewhat too concrete touchpoints like “well, suppressing revolts in the past has gone like such and such, so we should try to do similar for AGI”. And broadly they don’t credit an abstract argument about why something won’t work, or would only work given substantial further philosophical insight.
Re/ “don’t think thinking …”, well, if I say “LLMs basically don’t think”, they’re like “sure it does, I can keep prompting it and it says more things, and I can even put that in a scaffold” or “what concrete behavior can you point to that it can’t do”. Like, bro, I’m saying it can’t think. That’s the tweet. What thinking is, isn’t clear, but That thinking is should be presumed, pending a forceful philosophical conceptual replacement!
That is, in fact, a helpful elaboration! When you said
my leading hypotheses for what you could mean were:
Using thought, as a tool, has not occured to most such people
Most such people have no concept whatsoever of cognition as being a thing, the way people in the year 1000 had no concept whatsoever of javascript being a thing.
Now, instead, my leading hypothesis is that you mean:
Most such people are failing to notice that there’s an important process, called “thinking”, which humans do but LLMs “basically” don’t do.
This is a bunch more precise! For one, it mentions AIs at all.
As my reacts hopefully implied, this is exactly the kind of clarification I needed—thanks!
Sure, but you’re not preaching to the choir at that point. So surely the next step in that particular dance is to stick a knife in the crack and twist?
That is -
And if neither you nor anyone else can do either of those things… maybe it’s time to give up and say that this ‘thinking’/‘reasoning’ thing is just philosophically confused? I don’t think that that’s where we’re headed, but I find it important to explicitly acknowledge the possibility; I don’t deal in more than one epiphenomenon at a time and I’m partial to Platonism already. So if this ‘reasoning’ thing isn’t meaningfully distinguishable in some observable way from what LLMs do, why shouldn’t I simply give in?
I’ve had this tweet pinned to my Twitter profile for a while, hoping to find some like-minded people, but with 13k views so far I’ve yet to get a positive answer (or find someone expressing this sentiment independently):
This aside, there are some people around LW/rationality who seem more cautious/modest/self-critical about proposing new philosophical solutions, like MIRI’s former Agent Foundations team, but perhaps partly as a result of that, they’re now out of a job!
Yeah that was not my reaction. (More like “that’s going to be the most beautiful thing ever” and “I want to be that too”.)
No, if anything the job loss resulted from not doing so much more, much more intently, and much sooner.
Having worked on some of the problems myself (e.g. decision theory), I think the underlying problems are just very hard. Why do you think they could have done “so much more, much more intently, and much sooner”?
The type of fundamental problem that proper speculative philosophy is supposed to solve is the sort where streetlighting doesn’t work (or isn’t working, or isn’t working fast enough). But nearly all of the alignment field after like 2004 was still basically streetlighting. It was maybe a reasonable thing to have some hope in prospectively, but retrospectively it was too much investment in streetlighting, and retrospectively I can make arguments about why one should have maybe guessed that at the time. By 2018 IIRC, or certainly by 2019, I was vociferously arguing for that in AF team meetings—but the rest of the team either disagreed with me or didn’t understand me, and on my own I’m just not that good a thinker, and I didn’t find anyone else to try it with. I think they have good thoughts, but are nevertheless mostly streetlighting—i.e. not trying to take step after step of thinking at the level of speculative philosophy AND aimed at getting the understanding needed for alignment.
My understanding of what happened (from reading this) is that you wanted to explore in a new direction very different from the then preferred approach of the AF team, but couldn’t convince them (or someone else) to join you. To me this doesn’t clearly have much to do with streetlighting, and my current guess is that it was probably reasonable of them to not be convinced. It was also perfectly reasonable of you to want to explore a different approach, but it seems unreasonable to claim without giving any details that it would have produced better results if only they had listened to you. (I mean you can claim this, but why should I believe you?)
If you disagree (and want to explain more), maybe you could either explain the analogy more fully (e.g., what corresponds to the streetlight, why should I believe that they overexplored the lighted area, what made you able to “see in the dark” to pick out a more promising search area or did you just generally want to explore the dark more) and/or try to convince me on the object level / inside view that your approach is or was more promising?
(Also perfectly fine to stop here if you want. I’m pretty curious on both the object and meta levels about your thoughts on AF, but you may not have wanted to get into such a deep discussion when you first joined this thread.)
If you say to someone
and then you listen to what they say, and reask the question and interrogate their answers, IME what it very often grounds out into is something like:
So I would summarize that position as endorsing streetlighting, in a very broad sense that encompasses most math / science / technology. And this position is largely correct! My claim is that
this is probably too slow for making Friendly AI, and
maybe one could go faster by trying to more directly cleave to the core philosophical problems.
I discuss the problem more here: https://tsvibt.blogspot.com/2023/09/a-hermeneutic-net-for-agency.html
(But note that, while that essay frames things as “a proposed solution”, the solution is barely anything—more like a few guesses at pieces of methodology—and the main point is the discussion of the problem; maybe a writing mistake.)
An underemphasized point that I should maybe elaborate more on: a main claim is that there’s untapped guidance to be gotten from our partial understanding—at the philosophical level and for the philosophical level. In other words, our preliminary concepts and intuitions and propositions are, I think, already enough that there’s a lot of progress to be made by having them talk to each other, so to speak.
OK but what would this even look like?\gen
Toss away anything amenable to testing and direct empirical analysis; it’s all too concrete and model-dependent.
Toss away mathsy proofsy approaches; they’re all too formalized and over-rigid and can only prove things from starting assumptions we haven’t got yet and maybe won’t think of in time.
Toss away basically all settled philosophy, too; if there were answers to be had there rather than a few passages which ask correct questions, the Vienna Circle would have solved alignment for us.
What’s left? And what causes it to hang together? And what causes it not to vanish up its own ungrounded self-reference?
From scratch but not from scratch. https://www.lesswrong.com/posts/noxHoo3XKkzPG6s7E/most-smart-and-skilled-people-are-outside-of-the-ea?commentId=DNvmP9BAR3eNPWGBa
https://tsvibt.blogspot.com/2023/09/a-hermeneutic-net-for-agency.html
What makes you think there are any such ‘answers’, renderable in a form that you could identify?
And even if they do exist, why do you think a human being could fully grasp the explanation in finite time?
Edit: It seems quite possible that even the simplest such ‘answers’ could require many years of full time effort to understand, putting it beyond most if not all human memory capacity. i.e. By the end even those who ‘learned’ it will have forgotten many parts near the beginning.
(Upvoted since your questions seem reasonable and I’m not sure why you got downvoted.)
I see two ways to achieve some justifiable confidence in philosophical answers produced by superintelligent AI:
Solve metaphilosophy well enough that we achieve an understanding of philosophical reasoning on par with mathematical reason, and have ideas/systems analogous to formal proofs and mechanical proof checkers that we can use to check the ASI’s arguments.
We increase our own intelligence and philosophical competence until we can verify the ASI’s reasoning ourselves.