This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
I feel the same way but for the opposite reason. Non technical people who don’t understand the infrastructure requirements of actually running these things are talking about them as if they’re ghosts, or spirits.
It’s not “AGI that escapes the lab and infects all the computers”, it’s: your credit card company starts using OpenAI/Microsoft products to make determinations about debt collection and there are unforeseen edge cases.
You’re not going to have an AGI that somehow worms it’s way into a nuclear computer for several reasons:
We already have actual humans trying to do the same thing.
You’d notice the semi truck loads or H200s being unloaded into your building, as well as the data center being built to house them.
Also a lot of people seem to have a fundamental misunderstanding of what LLMs actually are. The mostly accurate soundbite explanation is that they're statistical models that predict the most likely data to follow some previous data. They don't "think" (unless you're in the camp that thinks that human consciousness is basically just a really complex statistical model running on a biological computer). Hell, you can do what an LLM does with pen and paper. It would take years, but you could simulate the computations being done by AI on your own. I realize this is similar to the Chinese Room thought experiment but it means that LLMs are nothing like consciousness unless you have a very simplistic and mechanical view of what consciousness is.
The biggest threats from LLMs and other forms of "AI" aren't Skynet or paperclip maximizers. The biggest threats are social disruption due to AI eliminating lots of jobs and consolidating wealth. Or kafkaesque nightmares caused by corporations, bureaucrats, and courts blindly relying on AI (or being intentionally oblivious to its shortcomings if it allows them to do what they already want). AI won't lead to Terminator, it'll lead to Terry Gilliam's Brazil.
There's a funny story I would really like to share here but won't due to an NDA and other issues. You are correct though, and this is (among other reasons) why I remain bearish on AI.
I'm bearish on the Rationalists' view of AI, but there's plenty damage a Chinese Room AI can do to us. Just look at what we did with the Internet.
More options
Context Copy link
More options
Context Copy link
As someone vocally in that camp, I invite you to demonstrate any other model for what human consciousness could possible be. And it doesn't even matter if the AI is "conscious" if it's intelligent and capable of using that intelligence to forward ends not aligned with our goals.
I mean there have literally been hundreds if not thousands of models of consciousness proposed over the last few thousand years, so take your pick? The burden of proof is on you to show why your mechanical view of consciousness is superior to all of the others.
I think there's been a lot of foolishness in history but conflating consciousness and intelligence/formidability at solving consequentialist tasks is just too indefensible to bring up.
More options
Context Copy link
I will point to the obvious trend line where ever greater fractions of human neurobiology and cognition have received mechanistic interpretations and a firm conceptual underpinning. Are we 100% done with it? No. But we can see temporal lobe epilepsy causing visions of supernatural entities, the precise firing and wiring of our visual neurons, and plenty more.
All a mechanistic theory of consciousness truly requires is that it obeys the laws of physics, and having peered into a few brains myself, I didn't spot anything contradicting the Standard Model of Physics.
This kind of rebuttal is about as valid as claiming that modern empirical/Western medicine is unfounded because there have been plenty of models before that proved flawed, and even its adherents admit it's not 100% perfect at explaining or treating illness. That's leaving aside that a mechanistic model that doesn't rely on supernatural/preternatural influences doesn't happen to be simply better/more parsimonious by Occam's Razor. I fail to see what additional empirical evidence the alternatives provide, so it remains the default assumption even if it's incomplete.
More options
Context Copy link
More options
Context Copy link
What would you say if I told you that you are not an intelligent human being, you are simply a physical and digital expression of regression to the mean. That if the hypothetical individual behind the @self_made_human account here on theMotte were to be thanos-snapped out of existence and their online activity taken over by 'n' number of d20s no one would notice, and nothing of value would be lost.
If the above suggestion strikes you as antagonistic, uncharitable, or belittling in anyway, you've already refuted your own argument.
Now I wonder. I don't think the actual suggestion is something I'd get behind. But if we step it back a little...
Say I, or any of us, were to have some current-generation LLM trained on everything we'd ever written and tweaked as appropriate. Then, we never post on TheMotte again but instead give that LLM our account and set it up to try its best to post as we do. I wonder how long it would take for anybody to notice. How long before somebody says, man, user X's posts seem a little less interesting than usual, I wonder if something happened to them.
In hindsight it was a big mistake to think of the Turing Test as a fixed-difficulty challenge outputting a binary "yes this passes" or "no this doesn’t".
If we'd instead reified the idea of "a Turing test of length X" outputting "this passes Y% of the time", then by now we'd have graphs of "in year Z the state-of-the-art pass rate was Y(X,Z)" and a much better idea of where (and if...) our current architectures' scaling was going to plateau.
Maybe not so much a mistake, rather an idea being limited due to it being new and there not being any way to try to put it into practice yet.
I tend to think the biggest issue is the huge variance in human intelligence. There are already mental hospitals and insane asylums full of people who just can't handle the real world at all. Millions of humans can't write down their thoughts coherently. A ChatGPT-4 level model programmed to pretend to be human could probably already seem smarter and more human than some fraction of the present human population. Especially if whoever is judging has not been primed to think that the thing they're communicating with might not be an actual human being.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
It mostly strikes me as incoherent, no number of d20s can implement computation and self_made_human's output is easily distinguishable from random strings.
Granted, the d20s are an intentional Reductio ad Absurdum, but if @self_made_human's mechanistic model of consciousness is correct, there is nothing their brain (or your brain for that matter) can produce that could not be reproduced by (or replaced with) rolling dice on a sufficiently detailed random encounter table "computation" be damned.
Edit: fixed link
The Chinese Room again? Sigh.
Much as the Library of Babel is both exhaustive and utterly useless, it's the algorithmic capability of finding useful information and acting on it that matters, and not that it can theoretically be summoned. You can, if you are exceptionally stupid, try and make an LLM that works by outputting every single possible string of a given length until you get a satisfactory answer. The pitfalls are obvious.
In this particular case, you are merely frontloading the computation to a look-up table, not eliminating it entirely. For any practical configuration, someone had to go to the trouble of doing that. Human consciousness is a mess of individually stochastic neurons that in bulk can produce synchronous signals and something analogous to a clock rate, our perception of time as perfectly continuous is not representative of the underlying neuro-computation. You've just temporally shifted the cognition involved, not eliminated it, and our brain uses LUTs all the time, if you've memorized times tables, you're not doing elementary arithmetic to figure out 7*7=49. That's still maths.
So "computation be damned" is doing all the heavy lifting. Why don't you use your ML chops to run AIXI, if it's such a trivial thing to you?
The Chinese Room keeps coming up because it happens to be an apt analogy/useful illustration. See my reply down-thread.
How can you be sure that I haven't?
Given that AIXI is uncomputable, I'm eagerly awaiting the announcement of you being the recepient of the Turing award.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
Memory and response to inputs both mean the actual number of outputs would be infinite if not for mortality. As is, it's probably only one of those meaninglessly large journalist numbers like the number of atoms in the solar system or something. Not that you could in any way generate such a list of outputs without fully understanding and simulating my brain in the first place, even discounting the impossible time/space requirements of such a task.
More importantly, the computation is the entire fucking point. That this post could technically just be a meaningless random string of characters doesn't mean it is one, and you will not perceive it as one. It is very clearly chosen in a nonrandom process. Getting from your post to this reply required processing in my brain, something you can in no way skip by randomly picking one out of a list of all the possible outputs of my brain.
Precisely. But how would you go about demonstrating that?
Basic information theory would suffice, unless you want me to demonstrate the concepts of meaning or the validity of induction. In which case you've retreated from your original point to the standard 'treating solipsism as a gotcha against materialists' position. This has come up so often on this board, I should come up with a catchy enough formulation to make it my flair: either any communication happens between real minds existing in an inductive external reality (including thoughts as communication across time) or the concept of communication is nonsense. So prepend any communication ever with "Conditional on solipsism being false,".
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
9, 13, 9, 3, 7, 1, 5, 12, 7, 2.
Just to throw Hlynka a
bonedie.More options
Context Copy link
More options
Context Copy link
I don't particularly care Hlynka, if this Thanos snapping managed to take both of us, you included, I'd consider it a net positive!
But I fail to see what the difficulty of Turing-testing random pseudonymous accounts on a text-based forum has anything to with it. Last time I checked, we're both operating according to the laws of physics and biology. Your analogy of how ML works is simply painful.
That's not really what his question is about.
I've never accused him of being concise and clear, or having a point.
Am I supposed to sob in horror at the idea of replacing humans with soulless automata instead? He doesn't provide any reason to think that humans or LLMs can't both be represented as the output of statistical processes occurring on computational substrates, even if said processes and substrates are very different.
As @ArjinFerman says, this isn't about "replacing humans with soulless automata" it's about replacing you in particular. I'm asking you whether you believe that the sum of your existence (your thoughts, feelings, memories, physical existence, output here on theMotte, etc...) is meaningfully distinct from that of an arbitrarily complex random number generator in any way?
If so, why do you believe that?
Ironically for how often I get accused of not understanding how machine learning works, I suspect that I have far more practical "hands-on" experience designing, implementing, and working with machine learning algorithms than most users here.
Sure, obviously. I can only assume that you think this is a valid description of ML/LLMs/AI, which it very much is not. If it's "randomness" that has you up and at it, then set the temperature of a model to 0 to get deterministic outputs. Problem solved?
I need no justification for such atomic preferences, I just have them, both in the incredibly stupid case you wish to make, and my vain attempts at steel-manning it in the scenario you're hand-waving at modern ML. LLMs do not capture the complexity of a human, nor do they have other aspects I care about, such as the fact that I'm not talking to a machine that will immediately flush everything out of memory as soon as it's done talking to me. Then again, I think that's a valid description of certain people on this forum, so who am I to judge?
I value my existence for its own sake, but if there's a human intelligence or smarter AI out there that is capable of remembering discussions and updating on them in the future, and capable of modifying future behavior on that basis, then I'm perfectly fine talking with it at length. Even GPT does update, but only slightly so as newer conservations enter the training data for the next one, but not in the same manner as a human.
If you mean a mind upload of myself running in-silico, and not a random LLM fine tuned on me, then yes, I would accept it as a valid replacement, given my conviction that it's very likely that in internally subjective terms it has the same qualia as I do. I would obviously prefer we both co-exist, at least until my flesh fails me, but I accord such an entity every right to use the SMH name to the same extent I do.
Here I was thinking I'm a human chauvinist, and now I'm pitying an ML model. Such insanity is hardly unheard of, I happen to have an uncle who is a professor in microbiology who swears by homeopathy.
I suppose it's a sign of how streamlined the process has become, when people so utterly divorced from the theoretical underpinnings of the technology are making a living off it.
More options
Context Copy link
Please tell me you've moved on from work for the military...
More options
Context Copy link
More options
Context Copy link
Why can't we all just get along?
Well, it's less humans, and more you in particular. It's also less about sobbing in horror, and more about whether you see much of a difference between the two cases. I think the question is interesting given Rat ideas on uploading consciousness.
There's about a zero percent chance that Hlynka doesn't know about my enthusiasm for the potential of mind uploading, with me seeing such an emulation of a human mind as equivalent in every way that matters to me as the same as a biological human.
That is not the same as replacing a human with a LLM trained on the corpus of their text, with outputs indistinguishable from the human. You'd need to do way more to establish it as a high fidelity replication of the original consciousness, even if I think in principle it's doable.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
I half-expect him to agree. There isn't really a way to buy into ideas like uploading you consciousness to the cloud, without endorsing that view. Either that, or going 100% the other way, and believing in souls, and that computers can carry them, but he explicitly rejected that view.
That's the Joke. If he agrees, I'll tip my hat to him for his ideological consistency. If he doesn't, I believe I will have made my point.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link