Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPT-4.5

openai.com

Post reply on HN
Print view View original

431–440 of 1001 posts

Re: GPT-4.5

#431

Post by npinsker »

porridgeraisin wrote: post #251
jefffoster wrote: post #187

Does anyone have any intuition about the how reasoning improves based on the strength of the underlying model? I’m wondering whether this seemingly underwhelming bump on 4o magnifies when/if reasoning is added.

It is possible to understand the mechanism once you drop the anthropomorphisms. Each token output by an LLM involves one pass through the next-word predictor neural network. Each pass is a fixed amount of computation. Complexity theory hints to us that the problems which are "hard" for an LLM will need more compute than the ones which are "easy". Thus, the only mechanism through which an LLM can compute more and solv…

Thanks for a well-written and clear explanation!

Re: GPT-4.5

#432

Post by throw234234234 »

eightysixfour wrote: post #115

Seeing OpenAI and Anthropic go different routes here is interesting. It is worth moving past the initial knee jerk reaction of this model being unimpressive and some of the comments about "they spent a massive amount of money and had to ship something for it..." * Anthropic appears to be making a bet that a single paradigm (reasoning) can create a model which is excellent for all use cases. * OpenAI seems to be betti…

> Anthropic appears to be making a bet that a single paradigm (reasoning) can create a model which is excellent for all use cases.

I don't think that is their primary motivation. The announcement post for Claude 3.7 was all about code which doesn't seem to imply "all use cases". Code this, new code tool that, telling customers that they look forward to what they build, etc. Very little mention of other use cases on the new model announcement at all. Their usage stats they published are telling - 80%+ or more of queries to Claude are all about code. i.e. I actually think while they are thinking of other use cases; they see the use case of code specifically as the major thing to optimize for.

OpenAI, given its different customer base and reach, is probably aiming for something more general.

IMO they all think that you need an "ensemble" of models with different capabilities to optimise for different use cases. Its more about how much compute resources each company has and what they target with those resources. Anthrophic I'm assuming has less compute resources and a narrower customer base so it economically may make sense to optimise just for that.

Re: GPT-4.5

#433

Post by swores »

DebtDeflation wrote: post #213
serjester wrote: post #58

Earlier quoted context omitted.

I suppose this was their final hurrah after two failed attempts at training GPT-5 with the traditional pre-training paradigm. Just confirms reasoning models are the only way forward.

GPT 5 is likely just going to be a router model that decides whether to send the prompt to 4o, 4o mini, 4.5, o3, or o3 mini.

My guess is that you're right about that being what's next (or maybe almost next) from them, but I think they'll save the name GPT-5 for the next actually-trained model (like 4.5 but a bigger jump), and use a different kind of name for the routing model.

Even by their poor standards at naming it would be weird to introduce a completely new type/concept, that can loop in models including the 4 / 4.5 series, while naming it part of that same series.

My bet: probably something weird like "oo1", or I suspect they might try to give it a name that sticks for people to think of as "the" model - either just calling it "ChatGPT", or coming up with something new that sounds more like a product name than a version number (OpenCore, or Central, or... whatever they think of)

Re: GPT-4.5

#434

Post by dgfitz »

svnt wrote: post #415
dgfitz wrote: post #390

Earlier quoted context omitted.

I think the trick is observing what is “better” in this model. EQ is supposed to be “better” than 4o, according to the prose. However, how can an LLM have emotional-anything? LLMs are a regurgitation machine, emotion has nothing to do with anything.

Words have valence, and valence reflects the state of emotional being of the user. This model appears to understand that better and responds like it’s in a therapeutic conversation and not composing an essay or article. Perhaps they are/were going for stealth therapy-bot with this.

But there is no actual empathy, it isn’t possible.

Re: GPT-4.5

#435

Post by dwaltrip »

mandevil wrote: post #361
crazygringo wrote: post #258

Earlier quoted context omitted.

> We don't really know what this is good for Oh come on. Think how long of a gap there was between the first microcomputer and VisiCalc. Or between the start of the internet and social networking. First of all, it's going to take us 10 years to figure out how to use LLM's to their full productive potential. And second of all, it's going to take us collectively a long time to also figure out how much accuracy is neces…

ChatGPT had its initial public release November 30th, 2022. That's 820 days to today. The Apple II was first sold June 10, 1977, and Visicalc was first sold October 17, 1979, which is 859 days. So we're right about the same distance in time- the exact equal duration will be April 7th of this year. Going back to the very first commercially available microcomputer, the Altair 8800 (which is not a great match, since tha…

So it’s barely been 2 years. And we’ve already seen pretty crazy progress in that time. Let’s see what a few more years brings.

Re: GPT-4.5

#436

Post by eightysixfour »

throw234234234 wrote: post #432
eightysixfour wrote: post #115

Seeing OpenAI and Anthropic go different routes here is interesting. It is worth moving past the initial knee jerk reaction of this model being unimpressive and some of the comments about "they spent a massive amount of money and had to ship something for it..." * Anthropic appears to be making a bet that a single paradigm (reasoning) can create a model which is excellent for all use cases. * OpenAI seems to be betti…

> Anthropic appears to be making a bet that a single paradigm (reasoning) can create a model which is excellent for all use cases. I don't think that is their primary motivation. The announcement post for Claude 3.7 was all about code which doesn't seem to imply "all use cases". Code this, new code tool that, telling customers that they look forward to what they build, etc. Very little mention of other use cases on t…

That's possible, my counter point would be that if that was the case Anthropic would have built a smaller reasoning model instead of doing a "full" Claude. Instead, they built something which seems to be flexible across different types of responses.

Only time will tell.

Re: GPT-4.5

#437

Post by dwaltrip »

bigstrat2003 wrote: post #334
crazygringo wrote: post #330

Earlier quoted context omitted.

And LLM's already have tons of productive uses. The biggest ones are probably still waiting, though. But this is about one particular price/performance ratio. You need to build things before you can see how the market responds. You say it's "not good business" but that's entirely wrong. It's excellent business. It's the only way to go about it, in fact. Finding product-market fit is a process. Companies aren't omnisc…

> And LLM's already have tons of productive uses. I disagree strongly with that. Right now they are fun toys to play with, but not useful tools, because they are not reliable. If and when that gets fixed, maybe they will have productive uses. But for right now, not so much.

Who do you speak for? Other people have gotten value from them. Maybe you meant to say “in my experience” or something like that. To me, your comment reads as you making a definitive judgment on their usefulness for everyone.

I use it most days when coding. Not all the time, but I’ve gotten a lot of value out of them.

And yes I'm quite aware of their pitfalls.

Re: GPT-4.5

#438

Post by xrisk »

fionic wrote: post #424
bigstrat2003 wrote: post #334

Earlier quoted context omitted.

> And LLM's already have tons of productive uses. I disagree strongly with that. Right now they are fun toys to play with, but not useful tools, because they are not reliable. If and when that gets fixed, maybe they will have productive uses. But for right now, not so much.

Hello? Do you have a pulse? LLMs accomplish like 90% of everything I do now so I don’t have to do it… Explain what this code syntax means… Explain what this function does… Write a function to do X… Respond to my teammates in a Jira ticket explaining why it’s a bad idea to create a repo for every dockerfile… My teammate responded with X write a rebuttal… … and the list goes on … like forever

It’s not that the LLM is doing something productive, it’s that you were doing things that were unproductive in the first place, and it’s sad that we live in a society where such things are considered productive (because of course they create monetary value).

As an aside, I sincerely hope our “human” conversations don’t devolve into agents talking to each other. It’s just an insult to humanity.

Re: GPT-4.5

#439

Post by throwaway314155 »

antirez wrote: post #253

In many ways I'm not an OpenAI fan (but I need to recognize their many merits). At the same time, I believe people are missing what they tried to do with GPT 4.5: it was needed and important to explore the pre-training scaling law in that direction. A gift to science, however selfist it could be.

> A gift to science

This is hardly recognizable as science.

edit: Sorry, didn't feel this was a controversial opinion. What I meant to say was that for so-called science, this is not reproducible in any way whatsoever. Further, this page in particular has all the hallmarks of _marketing_ copy, not science.

Sometimes a failure is just a failure, not necessarily a gift. People could tell scaling wasn't working well before the release of GPT 4.5. I really don't see how this provides as much insight as is suggested.

Deepseek's models apparently still compare favorably with this one. What's more they did that work with the constraint of having _less_ money, not so much money they could run incredibly costly experiments that are likely to fail. We need more of the former, less of the latter.

Re: GPT-4.5

#440

Post by adamtaylor_13 »

It’s crazy how quickly OpenAI releases went from, “Honey, check out the latest release!” to a total snooze fest.

Coming in the heels of Sonnet 3.7 which is a marked improvement over 3.5 which is already the best in the industry for coding, this just feels like a sad whimper.

Post reply on HN