Things we learned about LLMs in 2024 • Hacker News Board

Re: Things we learned about LLMs in 2024

#51

Post by parpfish » Tue, Dec 31, 2024, 7:31 PM UTC

datadrivenangel wrote: post #33

shihab wrote: post #20

Earlier quoted context omitted.

The last OpenAI valuation I read about was 157 billion. I am struggling to understand what justifies this. To me, it feels like OpenAI is at best few months ahead of competitors in some areas. But even if I am underestimating the advantage, it's few years instead of few months, why does it matter? It's not like AI companies are going to enjoy the first-mover advantage internet giants had over the competition.

It's justified if AGI is possible. If AGI is possible, then the entire human economy stops making sense as far as money goes, and 'owning' part of OpenAI gives you power. That is of course, assuming AGI is possible and exponential, and that marketshare goes to a single entity instead of a set of entities. Lots of big assumptions. Seems like we're heading towards a slow-lackluster singularity though.

Well, AGI would make the brainy information worker part of the economy obsolete. Well still need the jobs that interact with the physical world for quite a while. So… all us HN types should get ready to work the mines or pick vegetables

Re: Things we learned about LLMs in 2024

#52

Post by mjburgess » Tue, Dec 31, 2024, 7:31 PM UTC

Animats wrote: post #10

> Some of those GPT-4 models run on my laptop That's an indication that most business-sized models won't need some giant data center. This is going to be a cheap technology most of the time. OpenAI is thus way overvalued.

I don't think openai's valuation comes from a data center bet -- rather, I'd suppose, investors think it has a first-mover advantage on model quality that it can (maybe?) attract some buy-out interest or otherwise use in yet-to-be-specified product lines.

However, it has been clear for a long time that meta are just demolishing any competitor's moats, driving the whole megacorp AI competition to razor thin margins.

It's a very welcome strategy from a consumer pov, but -- it has to be said -- genius from a business pov. By deciding that no one will win, it can prevent anyone leapfrogging them at a relatively cheap price.

Re: Things we learned about LLMs in 2024

#53

Post by mjburgess » Tue, Dec 31, 2024, 7:33 PM UTC

neom wrote: post #13

"learned out about" - is that an Australian phraseology by chance? Sounds Australian or British of some manner.

Good ear -- the use of 'out' as an abbreviation of anything is a britishism.

Nowt, owt, -- nothing, anything

Re: Things we learned about LLMs in 2024

#54

Post by antirez » Tue, Dec 31, 2024, 7:33 PM UTC

About "people still thinking LLMs are quite useless", I still believe that the problem is that most people are exposed to ChatGPT 4o that at this point for my use case (programming / design partner) is basically a useless toy. And I guess that in tech many folks try LLMs for the same use cases. Try Claude Sonnet 3.5 (not Haiku!) and tell me if, while still flawed, is not helpful.

But there is more: a key thing with LLMs is that their ability to help, as a tool, changes vastly based on your communication ability. The prompt is the king to make those models 10x better than they are with the lazy one-liner question. Drop your files in the context window; ask very precise questions explaining the background. They work great to explore what is at the borders of your knowledge. They are also great at doing boring tasks for which you can provide perfect guidance (but that still would take you hours). The best LLMs (in my case just Claude Sonnet 3.5, I must admit) out there are able to accelerate you.

Re: Things we learned about LLMs in 2024

#55

Post by lowercased » Tue, Dec 31, 2024, 7:34 PM UTC

refulgentis wrote: post #34

Animats wrote: post #10

> Some of those GPT-4 models run on my laptop That's an indication that most business-sized models won't need some giant data center. This is going to be a cheap technology most of the time. OpenAI is thus way overvalued.

Been in the Mac ecosystem since 2008, love it, but there is, and always has been, a tendency to talk about inevitabilities from scaling bespoke, extremely expensive configurations, and with LLMs, there's heavy eliding of what the user experience is, beyond noting response generation speed in tokens/s. They run on a laptop, yes - you might squeeze up to 10 token/sec out of a kinda sorta GPT-4 if you paid $5K plus for…

I have a 2023 mbp, and I get about 100-150 tok/sec locally with lmstudio.

Re: Things we learned about LLMs in 2024

#56

Post by simonw » Tue, Dec 31, 2024, 7:34 PM UTC

Der_Einzige wrote: post #49

simonw wrote: post #43

Earlier quoted context omitted.

It doesn't matter how good the generated text is: it is still slop if the recipient didn't request it and no human has reviewed it.

By that definition machine to machine communication that happens "organically" (like how humans do it, where they sometimes strike up conversations unprompted with each other) is "slop". You're not seeing how the future of the world will develop.

If you ask me to read an unguided conversation between two LLMs then yes, I'd consider that slop.

Some people might like slop.

Re: Things we learned about LLMs in 2024

#57

Post by wslh » Tue, Dec 31, 2024, 7:36 PM UTC

antirez wrote: post #54

About "people still thinking LLMs are quite useless", I still believe that the problem is that most people are exposed to ChatGPT 4o that at this point for my use case (programming / design partner) is basically a useless toy. And I guess that in tech many folks try LLMs for the same use cases. Try Claude Sonnet 3.5 (not Haiku!) and tell me if, while still flawed, is not helpful. But there is more: a key thing with L…

Right, in simpler terms: The measure of LLMs success is how effectively they help you achieve your goal faster.

Re: Things we learned about LLMs in 2024

#58

Post by dmd » Tue, Dec 31, 2024, 7:36 PM UTC

dtquad wrote: post #28

What is the current status on pushing "reasoning" down to latent/neural space? Seems like a vaste of tokens to let a model converse with itself especially when this internal monologue often has very little to do with the final output so it's not useful as a log of how the final output was derived.

See https://news.ycombinator.com/item?id=42555320

Re: Things we learned about LLMs in 2024

#59

Post by simonw » Tue, Dec 31, 2024, 7:36 PM UTC

k2xl wrote: post #38

Something not mentioned is AI generated music. Suno's development this year is impressive. Unclear what this will mean for music artists over next few years.

Yeah, this year I decided to just focus on LLMs - I didn't touch on any of the image or music generation advances either. I haven't been following those closely enough to have particularly useful things to say about them.

Re: Things we learned about LLMs in 2024

#60

Post by Alex-Programs » Tue, Dec 31, 2024, 7:37 PM UTC

adsharma wrote: post #31

In spite of all this progress, I can't find LLMs that solve simple tasks like: Here is my resume. Make it look nice (some design hints). They can spit html and css, but not Google doc. On the other hand, Google results are dominated by SEO spam. You can probably find one usable result on page 10. The problem is not technology. It's a business model that can support the humans feeding data into the LLM.

Why would they be able to output a Google doc? It's a proprietary format. The closest thing would be rich text format to copy paste.