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've been watching Claude Plays Pokemon a bit and while it does seem AGI-ish moment to moment, the lack of proper memory is rather crippling as it often gets stuck in loops for hours. Some kind of better memory/online learning may be the last step before we have entry-level AGI agents, but it seems non-trivial.
Claude 3.7 Sonnet has a context window of 200k tokens. That is massive compared to the first commercial LLMs, which ranged from 4k to 16k.
It is, however, utterly dwarfed by other models like Gemini 2.0 Pro by Google. That one has a whopping 2 million token large context window, and I've personally made good use of it by throwing absurd amounts of text, including massive textbooks, into it.
There's plenty of ongoing research into both online learning, as well as drastically extending context lengths. We've gone from 4k to 2 million in about 2 years, without the original quadratic scaling of memory use still holding. I presume @DaseindustriesLtd would be better placed to answer how they pulled it off. Even accounting for performance degradation with very large context windows (needle in a haystack tests don't capture this), you can probably get around strictly needing online learning.
We don't know how Gemini is made. At this point I assume it's something incredibly dumb like Noam Shazeer's reduced attention schemes and not, say, DeepSeek's NSA. In short though, attention inherently allows for sparsity.
More options
Context Copy link
In context learning is not equivalent to training. There are certain model behaviors that are easily attainable through finetuning which cannot be achieved through in context learning. For example a properly safety tuned model will not output toxic content no matter how any of those two million tokens you use to insist that it should. The fact is that the model's ability to reason with in-context information versus embedded information is fundamentally different, so you will not be able to reach agi with just an infinite context window.
Of course I'm not saying you need agi to beat pokemon and the current state of the art should be capable of doing it if some minor adjustments are made.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link