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

751–760 of 1001 posts

Re: GPT-4.5

#751

Post by youssefabdelm »

torginus wrote: post #183

My 2 cents (disclaimer: I am talking out of my ass) here is why GPTs actually suck at fluid knowledge retrievel (which is kinda their main usecase, with them being used as knowledge engines) - they've mentioned that if you train it on 'Tom Cruise was born July 3, 1962', it won't be able to answer the question "Who was born on July 3, 1962", if you don't feed it this piece of information. It can't really internally co…

Yep. I've often said RLHF'd LLMs seem to be better at recognition memory than recall memory.

GPT-4o will never offhand, unprompted and 'unprimed', suggest a rare but relevant book like Shinichi Nakazawa's "A Holistic Lemma of Science" but a base model Mixtral 8x22B or Llama 405B will. (That's how I found it).

It seems most of the RLHF'd models seem biased towards popularity over relevance when it comes to recall. They know about rare people like Tyler Volk... but they will never suggest them unless you prime them really heavily for them.

Your point on recommendations from humans I couldn't agree more with. Humans are the OG and undefeated recommendation system in my opinion.

Re: GPT-4.5

#752

Post by j_maffe »

freehorse wrote: post #687
chefandy wrote: post #273

Earlier quoted context omitted.

That $7 trillion dollar ask pushed me from skeptical to full-on eye-roll emoji land— the dude is clearly a narcissist with delusions of grandeur— but it’s getting worse. Considering the $200 pro subscription was significantly unprofitable before this model came out, imagine how astonishingly expensive this model must be to run at many times that price.

Or, the model is nowhere as expensive as in the api pricing and they want to pump the user value of their pro subscription artificially?

Most people can evaluate whether the model improvements (or lack thereof) are worth the price tag

Re: GPT-4.5

#754

Post by PeterStuer »

jsemrau wrote: post #748
PeterStuer wrote: post #747

Currently my daily API costs for 4o are low enough and performance/quality for my usecases good enough that switching models has not made to to the top of application improvements. My cases' costs are more heavily slanted towards input tokens, so trying 4.5 would raise my costs over 25x, which is a non-starter.

Interesting observation.It seems capability has reached a plateau. Like a local maximum.

I'm not sure that is the right conclusion.

It is more like the AI part of the system for this specific use case has reached a position where focusing on that part of the complete application as opposed to other parts that need attention would not yield the highest return in terms of user satisfaction or revenue.

Certainly there is enormous potential for AI improvement, and I have other projects that do gain substantially from improvements in e.g. reasoning, but then GPT 4.5 will have to compete with Deepseek, Gemini, Grok and Claude on a price/performance level, but to be honest the current preview pricing would make it (in production, not for dev) a non starter for me.

Re: GPT-4.5

#755

Post by DebtDeflation »

JohnnyMarcone wrote: post #555
swores wrote: post #433

Earlier quoted context omitted.

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 nami…

They already confirmed GPT-5 will be a unified model "months" away. Elsewhere they claimed that it will not just be a router but a "unified" model. https://www.theverge.com/news/611365/openai-gpt-4-5-roadmap-...

If you read what sama is quoted as saying in your link, it's obvious that "unified model" = router.

> “We hate the model picker as much as you do and want to return to magic unified intelligence,”

> “a top goal for us is to unify o-series models and GPT-series models by creating systems that can use all our tools, know when to think for a long time or not, and generally be useful for a very wide range of tasks,”

> the company plans to “release GPT-5 as a system that integrates a lot of our technology, including o3,”

He even slips up and says "integrates" in the last quote.

When he talks about "unifying", he's talking about the user experience not the underlying model itself.

Re: GPT-4.5

#756

Post by netdevphoenix »

Is it official then?

Most of us have been waiting for this moment for a while. The transformer architecture as it is currently understood can't be milked any further. Many of us knew this since last year. GPT-5 delays eventually led to non-tech voices to suggest likewise. But we all held our final decision until the next big release from OpenAI as Sam Altman has been making claims about AGI entering the workforce this year, OpenAI knowing how to build AGI and similar outlandish claims. We all knew that their next big release in 2025 would be the final deciding factor on whether they had some tech breakthrough that would upend the world (justifying their astronomical valuation) or if it would just be (slightly) more of the same (marking the beginning of their downfall).

The GPT-4.5 release points towards the latter. Thus, we should not expect OpenAI to exist as it does now (AI industry leader) in 2030, assuming it does exist at all by then.

However, just like the 19th century rail industry revolution, the fall of OpenAI will leave behind a very useful technology that while not catapulting humanity towards a singularity, will nonetheless make people's lives better. Not much consolation to the world's super rich who will lose tons of money once the LLM industry (let us remember that AI is not LLM) falls.

EDIT: "will nonetheless make people's lives better" to "might nonetheless make some people's lives better"

Re: GPT-4.5

#758

Post by fergonco »

netdevphoenix wrote: post #756

Is it official then? Most of us have been waiting for this moment for a while. The transformer architecture as it is currently understood can't be milked any further. Many of us knew this since last year. GPT-5 delays eventually led to non-tech voices to suggest likewise. But we all held our final decision until the next big release from OpenAI as Sam Altman has been making claims about AGI entering the workforce thi…

> will nonetheless make people's lives better

Probably not the lives of translators or graphic designers or music compositors. They will have to find new jobs. As llm prompt engineers, I guess.

Re: GPT-4.5

#759

Post by energy123 »

~40% hallucinations on SimpleQA by a frontier reasoner (o1) and a frontier non-reasoner (GPT-4.5). More orders of magnitude in scale isn't going to fix this deficit. There's something fundamentally wrong with the approach. A human is much more capable of saying "I don't know" in the correct spots, even if a human is also susceptible to false memories.

Probably OpenAI thinks that tool use (search) will be sufficient to solve this problem. Maybe that will be the case.

Are there any creative approaches to fixing this problem?

Re: GPT-4.5

#760

Post by entropi »

netdevphoenix wrote: post #756

Is it official then? Most of us have been waiting for this moment for a while. The transformer architecture as it is currently understood can't be milked any further. Many of us knew this since last year. GPT-5 delays eventually led to non-tech voices to suggest likewise. But we all held our final decision until the next big release from OpenAI as Sam Altman has been making claims about AGI entering the workforce thi…

> will nonetheless make people's lives better

While I mostly agree with your assessment, I am still not convinced of this part. Right now, it may be making our lives marginally better. But once the enshittification starts to set in, I think it has the potential to make things a lot worse.

E.g. I think the advertisement industry will just love the idea of product placements and whatnots into the AI assistant conversations.

Post reply on HN