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 -
Predicting the future is really hard. In 2021 weren't you in despair at the prospects of a seemingly inevitable US world hegemony and centralized AI? But you changed your mind. Meanwhile I guess I was more bullish on China than has actually been warranted, not to mention many other more portfolio-relevant errors in prediction and modelling the future.
I was mostly impressed by him predicting what, to my non-expert eyes, resembles chain-of-thought and inference-time compute. Even being mostly wrong is pretty decent as long as you get some of the important parts right.
It's hard to account for human factor. Xi could just suddenly go senile and enact the sort of policies they predict, for example. Americans elected a senile president and then changed him for a tried-and-true retard with a chip on his shoulder who surrounded himself with ineffectual yes-men. That's history.
Technical directions are more reliable and are telegraphed years in advance.
Chain-of-thought is 2020 4chan tech. In 2020 also, Leo Gao wrote:
The idea of inference time compute was more or less obvious since GPT-3 tech report aka “Language Models are Few-Shot Learners”, 2019. Transformers (2017) are inherently self-conditioning, and thus potentially self-correcting machines. LeCun's Cake, aka unsupervised (then after Transformers, self-supervised) learning - Supervised – RL "cherry" is NIPS 2016. AlphaGo is 2015. And so on. I'm not even touching older RL work from Sutton or Hutter.
So in retrospect, it was more or less clear that we will have to
pretrain strong models with innately high or increased via post-training and synthetic data chain of thought capability
get a source of verifiable rewards and pick some RL algorithm and method
sample a lot of trajectories and propagate updates such that the likelihood of correct answers increases
Figuring out details took years though. Process reward models, MCTS have wasted a lot of brain cycles. But perhaps they could have worked too, we just found an easier way with another branch of this tech tree.
In this context, I find details of his predictions disappointing. The search space was narrowed enough that for someone in the know and trying to actually do a technically informed forecast could have done about as well as he did by semi-random guessing of buzzwords.
It's quite arrogant to say so without having written a better prediction (I predicted the chip war around 2020 too, but my guess was that we'd go way higher with way sparser models, a la WuDao, earlier). But this is just a low bar for claiming prescience.
More options
Context Copy link
More options
Context Copy link