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 -
Maybe old news to this forum, but I stumbled upon two articles. The first is a Harvard professor using Claude to write a paper that second year graduate student could, and the second is a Less Wrong post about how Go players have disempowered themselves to AI.
I bit the AI bullet finally a couple weeks ago: first using it to help me fix up my software package that had some pretty annoying bugs, and later to help me make a couple Anki add-ons I had sitting in my brain. Even the free version of Claude is very good, and although it made mistakes, at least for coding, it's as least as smart as me and works 100s of times faster and doesn't need to take breaks. With some back and forth I can get it to make something functional it a tenth of the time it would take me to do it by myself. This is almost exactly the experience the Harvard Physics Professor has with trying to get Claude to write the theoretical physics paper: it takes about two weeks of back and forth, but eventually Claude manages to produce what the author and other experts think is a pretty solid paper, in 20x the time as a graduate student.
This is all great for experienced faculty: at least in theoretical disciplines they can already greatly speed-up the research process without spending more or time on pesky grad students with the bonus that Claude doesn't mind if you're mean to it. For younger faculty, postdocs, and graduates students not so much. Not only do these students/researchers not have the experience or knowledge to critique Claude in the way to actually produce usable research, the fact that more experienced faculty can use AI in this way means that there's less demand for educated individuals in general, and slowly but probably surely, the pyramid scheme of Academia will collapse.
Which brings me to the second article. In contrast to chess, which has been a "solved" game for a long time and thus has had time to develop antibodies against heavy AI/computer use, the Go world has been taken completely by storm by AlphaGo, which has led to rampant cheating with AI and rapid deskilling of the player-base. The author of the article highlights that this was a choice made by the community, both not to punish cheaters, and to care more about abstract values like ranking and points rather than having fun or development of some kind of genuine skill.
How does this relate to the first article? I worry that this Harvard professor is missing the large scale implications of wide-spread adaption of his AI practices on the university system, and how on emphasis on one aspect of the system (solving problems efficiently) can destroy the whole thing if not left unchecked. Many people both inside and outside academia take the purpose of the academy to be the generation of new knowledge, mainly in the form of research papers. This is certainly its most important role, at least to society at large. But academia also needs to perform two other key functions: disseminate that knowledge to industry/the general public so it can actually be used by society, and to reproduce itself so knowledge can continue to be generated. AI puts both of these secondary functions in grave danger.
Consider two scenarios. In the first, AI agents get no better than they are now at solving scientific problems. In this case they are still useful to higher level faculty in producing novel research output. Due to enhanced productivity at the top, there is a lesser need for graduate students, postdocs and younger faculty, and the ones that do remain in the system receive inferior training because of heavy reliance on AI to pass coursework and generate their own novel research questions. Over time as traditionally trained professors retire the effectiveness of the system declines because the new professors aren't as competent and are thus unable to use AI as effectively. In the long-term output shrinks, and less knowledge is able to be translated to the general public.
In the second scenario AI continues to improve its capabilities. In this scenario research output continues to go up indefinitely, but we begin to lose the ability to comprehend what much of it means or how much of it can be applied. Academia basically stops needing to exist at all, and we are reliant on the proper alignment of AI to get anything of value out of the research it performs.
In both scenarios, academia has basically signed its own death wish. For the reward of extremely high-productivity for very AI-savvy professors over the next few years, the system that brought us so many world-altering discoveries will basically be dismantled. Like the Go players who have seen all enjoyment and skill taken out of the game by heavy AI use, I think we are going to see the same in academia, except at the very top. And if you're not effectively training students to take your place, the system can't last very much longer.
Luckily being in biology I still have quite a bit of time left before AI can automate away my hands, but this whole thing has made me very personally scared.
If AI becomes "Real" AI then...good?
Automated high quality research and scientific improvement is in all likelihood going to be a great thing for mankind as long as alignment is dealt with.
If AI remains like shitty LLMs.....don't worry?
The biggest use case in medicine for LLMs is for ambient note generation. It sucks for this. Far below the ability of a human scribe or an actual physician, however people like it because it's fast and many choose not to spend the time proofreading. But if you do then it doesn't really save much time (read: value).
The same thing occurs for quite a few of the uses case. It doesn't work or it requires so much verification that it isn't really that helpful.
The economy is often more or less run by tech and finance, and it does seem actually useful in those disciplines, that doesn't mean it really is going to have much use outside those when stress tested without improvements (and if it improves who cares).
People will realize how to reallocate Human Resources or get mad at the problems with the technology.
I wrote this story in 2022. I did not write it about LLMs. It's a problem that dates back to the invention of calculators, or arguably to the development of writing.
I think it's more relevant today than four years ago.
We have solutions, of a kind. Programmers pick up (and build) esolangs, hardware hackers hunt for antiques that may never have been documented, sysadmins can learn to install archlinux without archinstall, woodworkers try random crap with random scrap, mechanics do custom work, so on. Math Olympiads and spelling bees can bypass the development of calculators and always-on spell-check. But these are solutions to the old versions of the problem.
In the real FOOM case, we don't care. It's the Minds' universe, or CelestAI, and either we live with it or we don't and it doesn't really matter.
If the extant development is bounded, or even unbounded in many ability but constrained in others, it gets significant more bad.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link