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 -
TracingWoodgrains has been a fan of Opus, and seems a little frustrated by 4.7. That said, it may depend on your use case.
I'm generally not that surprised if there are occasional stinkers. I've given specific caveats around other vendors : it's just too easy to benchmax or find a bad local maxima such that there's some minor revisions that either don't have any benefit, or only have backend benefit. Repeated problems or broader-scale issues would say more, but there's been a number of surprisingly good models from other vendors recently, including small-parameter and open-model approaches.
I'm skeptical that LLMs are themselves enough to go to AGI, but I'm also skeptical that they're going to stop at exactly last month's level of capability, and last month's capabilities included solving some Erdos problems. There's a lot of low-hanging fruit just in terms of UI and process tooling, nevermind areas where we haven't applied existing tools.
That said, I recognize that a lot of the major AI vendors have ranged from scumbags to scammers. Altman's ridiculous behaviors, especially in relation to RAM, have made the most enemies (maybe even more than Musk's more conventional culture war), but the best PR the whole faction has got has come from anti-AI people, so that's a whole big mess.
Somewhat an aside, but I consider that first link to be a first-degree chart-crime. First of all radar plots are inherently iffy, since we pay close attention to the "area" and the area is highly dependent on how the categories are organized (a "spiky" radar plot has much less total area than if you sort the axes to create a "lopsided" plot, despite showing the same information). This is a little bit defensible if the adjacency of the categories is obvious and inherent, but they frequently are not. For example, "Occupational: Writing Literature and Language" is NOT next to "Text: Creative Writing" for no good reason at all. And furthermore, what is the scale of the chart? It's "Arena rank"... which is NOT equally spaced. The chart implies that the difference between #1 and #2 is the same as (or even slightly bigger than, considering how the radar chart "expands") that between #3 and #4, but this is plainly not the case. They should be using some kind of actual score instead, perhaps a scaled one. Sure, it allows consistency across axes, but if we are comparing a model to its successor, the rating scale definitely shouldn't be implicitly including other models like it does now (in one spot it drops from rank 2 to rank 5, does this mean in that category some other model class does abnormally well, or that did Claude truly degrade?). Even worse, the center of the plot, usually a natural "zero", is not a zero at all - it's rank 6. There are, as you know, dozens and dozens of models in the rankings, so rank 6 being a zero score is totally nonsensical.
More options
Context Copy link
More options
Context Copy link