Most active commenters
  • bsenftner(3)
  • aydyn(3)
  • jdiff(3)

←back to thread

LLM Inevitabilism

(tomrenner.com)
1612 points SwoopsFromAbove | 18 comments | 15 Jul 25 04:35 UTC | HN request time: 4.289s | source | bottom
Show context
JimmaDaRustla ◴[15 Jul 25 13:50 UTC] No.44571157[source]▶
The author seems to imply that the "framing" of an argument is done so in bad faith in order to win an argument but only provides one-line quotes where there is no contextual argument.

This tactic by the author is a straw-man argument - he's framing the position of tech leaders and our acceptance of it as the reason AI exists, instead of being honest, which is that they were simply right in their predictions: AI was inevitable.

The IT industry is full of pride and arrogance. We deny the power of AI and LLMs. I think that's fair, I welcome the pushback. But the real word the IT crowd needs to learn is "denialism" - if you still don't see how LLMs is changing our entire industry, you haven't been paying attention.

Edit: Lots of denialists using false dichotomy arguments that my opinion is invalid because I'm not producing examples and proof. I guess I'll just leave this: https://tools.simonwillison.net/

replies(13): >>44571266 #>>44571325 #>>44571342 #>>44571439 #>>44571448 #>>44571473 #>>44571498 #>>44571731 #>>44571794 #>>44571923 #>>44572035 #>>44572307 #>>44572665 #
jdiff ◴[15 Jul 25 14:01 UTC] No.44571266[source]▶
The IT industry is also full of salesmen and con men, both enjoy unrealistic exaggeration. Your statements would not be out of place 20 years ago when the iPhone dropped. Your statements would not be out of place 3 years ago before every NFT went to 0. LLMs could hit an unsolvably hard wall next year and settle into a niche of utility. AI could solve a lengthy list of outstanding architectural and technical problems and go full AGI next year.

If we're talking about changing the industry, we should see some clear evidence of that. But despite extensive searching myself and after asking many proponents (feel free to jump in here), I can't find a single open source codebase, actively used in production, and primarily maintained and developed with AI. If this is so foundationally groundbreaking, that should be a clear signal. Personally, I would expect to see an explosion of this even if the hype is taken extremely conservatively. But I can't even track down a few solid examples. So far my searching only reveals one-off pull requests that had to be laboriously massaged into acceptability.

replies(10): >>44571346 #>>44571362 #>>44571401 #>>44571454 #>>44571514 #>>44571545 #>>44571653 #>>44571865 #>>44571882 #>>44572188 #
1. bsenftner ◴[15 Jul 25 14:09 UTC] No.44571346[source]▶
> I can't find a single open source codebase, actively used in production, and primarily maintained and developed with AI.

That's because using AI to write code is a poor application of LLM AIs. LLMs are better suited to summary, advice, and reflection than forced into a Rube Goldberg Machine. Use your favorite LLM as a Socratic advisor, but not as a coder, and certainly not as an unreliable worker.

replies(6): >>44571398 #>>44571404 #>>44571526 #>>44571609 #>>44571701 #>>44572177 #
2. komali2 ◴[15 Jul 25 14:15 UTC] No.44571398[source]▶
> Use your favorite LLM as a Socratic advisor

Can you give an example of what you mean by this?

replies(1): >>44571462 #
3. eddythompson80 ◴[15 Jul 25 14:16 UTC] No.44571404[source]▶
The entire hype for LLMs is that they can *do* anything. Even if only writing code, that could justify their hype. If LLMs mean Grammarly is now a lot better (and offered by big tech) then it’ll be very disappointing (economically speaking)
replies(1): >>44573957 #
4. hyperadvanced ◴[15 Jul 25 14:23 UTC] No.44571462[source]▶
You read the book and have the llm ask you questions to help deepen your understanding, e.g.
replies(1): >>44571871 #
5. Arch-TK ◴[15 Jul 25 14:29 UTC] No.44571526[source]▶
It's absolutely hilarious reading all these "you're holding it wrong" arguments because every time I find one it contradicts the previous ones.
6. robwwilliams ◴[15 Jul 25 14:38 UTC] No.44571609[source]▶
I support this comment. AI for coding does still involve much prodding and redirecting in my limited experience. Try getting Claude to produce even a simple SVG for a paper is a struggle in my experience.

But for helping me as a partner in neurophilosophy conversations Claude is unrivaled even compared to my neurophilosophy colleagues—speed and the responsivness is impossible to beat. LLMs are at pushing me to think harder. They provides the wall against which to bounce ideas, and those bounces often come from surprising and helpful angles.

7. JimmaDaRustla ◴[15 Jul 25 14:47 UTC] No.44571701[source]▶
Agreed. That argument they made was a straw-man which doesn't really pertain to where LLMs are being leveraged today.
replies(1): >>44572613 #
8. aydyn ◴[15 Jul 25 15:02 UTC] No.44571871{3}[source]▶
Or you dont read the book at all and ask the llm to give you the salient points?
replies(2): >>44572009 #>>44572013 #
9. noah_buddy ◴[15 Jul 25 15:13 UTC] No.44572009{4}[source]▶
Socratic method usually refers to a questioning process, which is what the poster above is getting at in their terminology. Imo
10. teg4n_ ◴[15 Jul 25 15:14 UTC] No.44572013{4}[source]▶
And cross your fingers it didn’t make them up?
replies(1): >>44572272 #
11. usrbinbash ◴[15 Jul 25 15:27 UTC] No.44572177[source]▶
> That's because using AI to write code is a poor application of LLM AIs

Then why is that exact usecase being talked about ad nauseam by many many many "influencers", including "big names" in the industry? Why is that exact usecase then advertised by leading companies in the industry?

replies(1): >>44573934 #
12. aydyn ◴[15 Jul 25 15:37 UTC] No.44572272{5}[source]▶
Yes. If it can get me through 10 books in the same time it takes you to get through 1 I am fine with an extra 1% error rate or whatever.
replies(1): >>44572634 #
13. jdiff ◴[15 Jul 25 16:07 UTC] No.44572613[source]▶
The claim that you made was "if you still don't see how LLMs is changing [the IT] industry, you haven't been paying attention." Pointing out that there is no visible evidence of that change in the industry you mentioned and inviting you to provide some where others have repeatedly failed is not attacking a strawman.
14. jdiff ◴[15 Jul 25 16:08 UTC] No.44572634{6}[source]▶
If I spend an afternoon on CliffNotes, I haven't read a hundred books in a day. This isn't one weird trick to accelerate your reading, it's entirely missing the point. If any book could be summarized in a few points, there would be no point to writing anything more than a BuzzFeed listicle.
replies(1): >>44572771 #
15. aydyn ◴[15 Jul 25 16:18 UTC] No.44572771{7}[source]▶
Lots of books CAN be summarized in a few points.
replies(1): >>44577411 #
16. bsenftner ◴[15 Jul 25 17:51 UTC] No.44573934[source]▶
It probably has a lot to do with those writing the code are not those marketing the finished software, and between the two groups they do not communicate very well. Once marketing gets any idea momentum going, they will go for it, because they can sell it, and then make engineering pay that bill. Not their problem.
17. bsenftner ◴[15 Jul 25 17:53 UTC] No.44573957[source]▶
I believe it was Flavor Flave that said: "Don't believe the hype", and the pretty much applies to everything humans do.
18. komali2 ◴[16 Jul 25 00:27 UTC] No.44577411{8}[source]▶
Yes, that's true, but bullet point ideology isn't very useful imo.

For example you can easily sum up "How to Win Friends and Influence People" into a few short recommendations. "Empathize with people, listen closely to people, let them do most of the talking, be quick to be upfront and honest with your faults." But it gets tenuous and some of the exact meaning is lost, and the bullet points aren't very convincing on their own of the effectiveness of the recommended methods.

The book fleshes out the advice with clarifications as well as stories from Carnegie's life of times he used the techniques and their results.

I think humans are good at remembering stories. For example I always remember the story about him letting his dog off the leash in a dog park and how he leveraged his technique for getting off the hook when a cop confronted him about it. I think I remember that better than the single bullet point advice of "readily and quickly admit when you're wrong." In fact I think the end of each chapter succinctly states the exact tip and I'm sure I'm misstating it here.

I could use anki to memorize a summary but I don't think I'd be able to as effectively incorporate the techniques into my behavior without the working examples, stories, and evidence he provides.

And that's just non fiction. I can't fathom the point of summarizing a fiction book, whose entire enjoyment comes from the reading of it.