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 -
You're leaping up a tower of assumptions here. Let's go in order:
1: A greater intelligence than humans is possible.
2: An artificial intelligence greater than human intelligence is possible.
3: Humans are able to create an artificial intelligence greater than our own.
4: Language models are a viable means by which such an intelligence might be created.
5: It is possible that humans might create such an intelligence via language models in the short to medium term future.
6: It is desirable to create such an intelligence.
7: If we fail to create such an intelligence in the short term, human civilisation will be permanently stunted or developmentally limited.
Scott asserted 7, and in return, I asserted that 7 is an insane position. You have responded to me by asserting something in the range of 2-4.
Let me make some other suggestions. It might be possible to choose not to make super-intelligences via LLMs now, but to do it in two hundred years, or five hundred years, or on any other timeline you like. Or it might be possible to make super-intelligences via some other, non-LLM means. There's no reason to think that LLMs specifically need to be used to conjure up super-intelligence right now or else the human race has no future.
I'm also inclined to challenge him just on the grounds of goals and values. He mentions 'civilisational childhood or an early grave'. Well, what would constitute 'civilisational adulthood'? What is civilisational maturity? He's making a value judgement. Suppose I don't think that technical or computational capacity is identical to civilisational maturity. That just seems implausible on the face of it - and certainly I don't believe that the worth of a civilisation or a people should be measured linearly and quantitatively like that. On the contrary, in individual humans I think that maturity involves a capacity for self-control and restraint; if that applies also on the civilisational level, then one might argue that civilisational adulthood would involve the ability to choose not to go down a path that we judge to be unwise.
In Scott's response, he says that thinking of an AI as all-powerful would be shituf, which is a sin. Firstly, that's not what shituf means, and secondly, arguably the whole point of shituf is that it's not sin for a Gentile. Shituf is an ad hoc attempt to be merciful to Gentiles who hold opinions that, in Jewish eyes, are not properly monotheistic. (I consider this to be grounded in a mistaken understanding of the Trinity, but I do appreciate the attempt to extend grace.) But if we're going to invoke the wisdom of the Jewish tradition, Scott and Pinker both being Jewish, I might suggest also considering another story in the Torah.
I cannot help but read that statement about 'civilisational childhood' and picture someone standing in front of a half-built tower reaching into the heavens and explaining, "If we stop building this tower to heaven, we are consigning ourselves to permanent civilisational childhood or an early grave."
Tolkien's reaction to the news of the atomic bombs was, in Letter #102 (9 August 1945):
Truly, there is nothing new under the sun.
Evolution has selected for intelligence in us, but has selected for intelligence only to the extent that the trade-offs are worth it. Since such trade-offs (size of birth canal, caloric requirements, etc) exist, it seems supremely unlikely that our level of intelligence is the highest one possible.
And even if von Neumann or Einstein or whichever specific individual you care to name is the absolute highest level of intelligence possible, it should be obvious that the ability to create an arbitrary number of such individuals, who do not need to sleep and who do not need to rest, would be able to produce a higher aggregate intelligence-driven output than humanity.
Human intelligence runs on physical hardware. There is nothing magic about a human mind. Building anything that occurs in nature is just an engineering problem.
A weak assumption, since we have already done so across a wide variety of domains.
The core argument really doesn't hinge on whether LLMs is the path that leads to AGI. If LLMs are a dead end, just replace "invest in LLM" with "invest in Ai research".
Nobody has said anything about the time frame.
Again, nobody said anything about time frame.
But these go together: If human level intelligence isn't the highest possible level of intelligence (and all signs point to "not"), and if there exists technologies or scientific insights that require a higher level of intelligence to discover or make proper use of, then it is trivially true that humans are locked out of such technologies.
The italicized assumption above is the only assumption that isn't trivially true and that is actually necessary for the stunting argument to be true.
You've mis-used the word 'trivial'. All those arguments you've made might be true, but none of them are trivially true. They all depend on contingent claims about the world. Any claim that is meaningfully evidence-based cannot be trivial.
At any rate, I read Scott as clearly concerned about time frame. I do not think he is talking about abstract possibilities. Yes, the word 'forever' is in that passage I quoted, but he is plainly talking about short-term technology policy. His view, as far as I can tell, is that we need to 'solve the alignment problem' while creating 'AI' in the immediate short term, or else not make 'AI' at all, and he views the latter option as some sort of permanent stunting of human potential.
More options
Context Copy link
More options
Context Copy link
I'm not making assumption 6, that's the whole crux of the issue. Concerns about the specific nature of the intelligence are the whole point in question.
1-5 are trivial.
With 7 you're bringing in 'short term'. Scott is specifically saying 'not right now' and neither am I. Scott wrote about Plan A where he explicitly asks for 15 or so years to make sure it's done right, not racing ahead. His big thing is 'let's not do it just yet, it is coming soon but we are not properly preparing or implementing it so let's slow down and do it right'. The alternative to 'we need more time to prepare' is 'never do it', which is what he is criticizing as unacceptable.
Realistically, people are not going to wait two hundred years, or five hundred years, or any 'safely beyond any need to care about it' distance of time to reap immense gains in wealth and power. This isn't wise but it's happening anyway. Non-LLM AI doesn't really work for general tasks, only LLMs have shown generality or long-horizon capabilities.
Well, you may not care for technical capacity. In some abstract sense, the Zulu empire and the British empire were equal peers. But in all real and significant senses, obviously the British were better and greater and the Maxim Gun settled the question.
The Tower of Babel is not a real thing and I don't see any need to care about it. Tolkien also writes about unreal things, things that could not be real, do not hold to principles we actually observe in the world. His genre is fantasy.
Science fiction actually relates to reality in a certain sense, that's the key distinction here. Things start off in sci-fi (like the atomic bomb) and have been made real, often by creators inspired by the fiction. You can't criticize sci-fi by appealing to fantasy, Jewish or otherwise.
I don't actually think 1-5 are all obvious, but my point is that you responded to my criticism of my Scott with points that are in the 1-5 area.
Do I think that people are actually going to stop making LLMs? No, I don't. If I had to bet, I'm in the camp that sees LLMs as a 'normal technology'. I think they will exist, more use will be made of them, and that the uses found will be moderate and not that interesting. I think both the doomers and the singularitans will be disappointed.
The point that I made that you took issue with was that it is ludicrous to say that if we don't make LLM-based 'AI', we're '[consigned] to a permanent civilisational childhood or an early grave'. That is absurd. If we don't pursue that technology, we're not even locking ourselves out of artificial intelligence, much less locking ourselves out of civilisational maturity.
The Babel example was meant to be evocative, not an argument as such. You're an atheist, fine, whatever. Scott made a comparison from Judaism so I felt it was within-bounds to reply with another comparison from Judaism.
Fifteen years is short term.
This is one of the weird things about the debate to me - the absurdly short timelines everybody seems to be approaching it with. From my perspective, Scott and the 'AI pause' people are basically one subgroup of accelerationist AI-enthusiast arguing with another subgroup of accelerationists. Five, ten, fifteen years, that's all still the short term. That is the immediate future. Medium-term for me means 'within the next hundred years or so', i.e. within the lifetimes of people alive today, and long-term means after that.
I think there is a very good argument to the effect that 'civilisational maturity', if it is a meaningful measure, refers to more than sheer destructive power.
Can you give some examples where the "more" would diverge meaningfully from the "sheer destructive power"? What's an example of a "mature civilization" with minimal destructive potential? The Amish?
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link