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 -
Over the last few months, I've followed someone named Alexander Kruel on Substack. Every single day, he writes a post about 10 important things that happened that day - typically AI breakthroughs, but also other of his pet concerns including math, anti-wokeness, nuclear power, and the war in Ukraine. It's pretty amazing that he is able to unfailingly produce this content every day, and I'm in awe of his productivity.
Unfortunately, since I get this e-mail every morning, my information diet is becoming very dark.
The advances in AI in the last year have been staggering. Furthermore, it seems that there is almost no one pumping the breaks. We seemed doomed to an AI arms race, with corporations and states pursuing AI with no limits.
In today's email, Kruel quotes Elizier who says:
Elizier is ahead of the curve. Where Elizier was in 2015, I am now. AI will destroy the world we know. Nate Soares, director of MIRI, is similarly apocalyptic.
What comes after Artificial General Intelligence? There are many predictions. But I expect things to develop in ways that no one expects. It truly will be a singularity, with very few trends continuing unaltered. I feel like a piece of plankton, caught in the swells of a giant sea. The choices and decisions I make today will likely have very little impact on what my life looks like in 20 years. Everything will be different then.
So, party until the lights go out? How do I deal with my AI-driven existential crisis?
Chess AI took decades to go from "possible" to "superhuman" (where the best AIs outperform the best humans), and then a decade or two more to go from "superhuman" to "ultrahuman" (where AI-human "centaur" teams no longer outperform AIs without humans). We're barely reaching the "possible" stage with AGI. I'd still say "party", maybe take that big vacation now rather than in retirement, but don't blow the cash you were saving for the electric bill, just in case the lights don't go out for a few more decades.
On the bright side, this also means there's likely to be no unexpected hardware overhang, because we're already throwing hardware at AI software as fast as we can. Imagine if we'd let FLOPs/watt and FLOPs/$ grow exponentially for another few decades and then discovered how powerful huge models can become. The AI arms race world is a world where a relatively slow takeoff, one where alignment failures occur on systems powerful enough to be educational but weak enough to be survivable, is conceivable. I'd say we're seeing the bare beginnings of that: the newest LLMs mostly don't pass the Turing test and mostly don't talk like psychopaths, but the exceptions are now common and blatant enough to get past the "nothing ever happens" bias of normies and the "my work just won't go wrong" bias of creators.
On the other hand, there's still going to be a software overhang, and I have no idea how big it will be. Some fields of computational mathematics in the last decades saw 3OOM of software speedup at the same time as they saw 3OOM of hardware speedup. 1,000 << 1,000,000, but if the first superhuman AGI can quickly make itself a mere thousand times faster just by noticing algorithms we've missed then we're still probably completely screwed.
Yeah, I agree that the focus on hardware limits is less important.. Training improvements mean these models have been getting orders of magnitude less expensive to train in a short time frame. I'd expect this to continue.
There's also the near term worry about emergent, unpredictable behaviors. Above a certain model size, there might be a discontinuous leap from 0 to 1. So I'm not particular reassured by the idea that we could pull the plug when necessary. And we wouldn't pull the plug anyway, would we? A superhuman AI would be so incredibly helpful and useful that some asshole corporation like Microsoft would say "damn the risks" and just let it go anyway.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link