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 -
(As an aside: this text is confusing, you jump from psychologizing to chess to Moravec, and it's hard to keep track of the core assertion).
Hlynka do you really want to go another round. Because I don't particularly care for it, especially not enough to respond with pure object-level to your Bulverism – even though this is what we do here all the time, to the point of obsessiveness. How boring is your life exactly? Maybe go on a trip to ol' Europe (or Asia), come here and we talk it over. Bring some good liquor if you can, Turks are lousy in this regard.
The real inferential distance here seems to come from inferences made on the basis of evidence, versus whatever you're trusting. Say, Caplan isn't known for being impressed by GPT's chess skills – he tests it on exam questions he believes are tricky. You tried that too: last time you've been claiming that autoregressive LLMs cannot not hallucinate your daughter's name, due to how they're trained. I've shown that ChatGPT replies as well as you said a human would, admitting it doesn't know. Since then, it's become possible to have coherent dialogues with files below 4 Gb and very sensible ones with stuff like Vicuna-13B. I assume you haven't tested that yet, despite insistence of multiple people here, because I haven't seen you concede the issue. Now you're simply dropping that and pivoting to saying they can't play chess, again due to fundamentals of their training. It's «just» regression, «just» predicting words, see. And words aren't chess moves or facts, so of coursh' flawlessly modeling grammar ought to be unsurprising, trivially feasible for an LLM – unlike modeling the logic of the game board. Or something. Although Chomsky, another guy who does not check whether his punches land, still seems to think that grammar cannot be solved with «just» statistics either. And while we're at it, Minsky's arguments were also obsolete at release. Now Dreyfus, this Heidegger scholar, was largely correct about Minsky's Talmudic-symbolic approach, but of him, Minsky only had to say that he doesn't understand and should be ignored.
On a meta-level, your rhetorical similarity to all those eggheaded paper-pushers is a rather bigger indictment of your position than whatever you say specifically about the tech. You scoff at pure math guys, at ivory tower symbol manipulators, but your arguments here are: brandishing your degree, discussing the history of academic debate, a bit of homegrown epistemology, dismissive blogposts and nerdy web comics, throwing around applause lights and rat lingo. You do not apply the pragmatic and entrepreneurial American lens, the «does it work tho» ethos. You treat LLM enthusiasts (and by proxy, developers who say the same) with the sort of casual disdain you believe pure math bros have for signal-processing researchers; where do you think notions like gradient, dropout and channel came from to LLMs? Read about Hyena Filters some time to stay ahead of the curve.
As a man involved with engineering, you ought to know that a skilled engineer can make bits and bytes perform bizarre magical circus tricks a limp-wristed intellectual would not see coming, and on the other hand that some problems are vastly harder than a sedentary utopian imagines, for not everything is deducible from first principles; that processes can be very deep and counterintuitive, that it can take a lifetime to figure out the nitty-gritty of how something actually works, so it is sensible to defer to reality over theory and assumption; worse, you preach this attitude. But you do not practice what you preach. Are you really a thing-manipulator or a symbol-manipulator? Or maybe more of a people-manipulator, at this stage of your career?
You are wrestling with your own shadow.
Congrats on your nine-year old never making illegal moves, by the way. You teach them well. Recently I've learned that my gainfully employed backend dev friend, 32, doesn't know how castling works, and is uncertain about pawn's inability to attack straight. I'd say he should be able to get to 1600 ELO, at least, with a little bit of finetuning. It's an issue of knowledge and experience, not just ineffable innate properties of the mind.
Do you have enough experience with LLMs to justify your conclusions?
I'll cite again Piantadosi again.
More options
Context Copy link