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

941–950 of 1001 posts

Re: GPT-4.5

#941

Post by chamomeal »

vjerancrnjak wrote: post #697
tmaly wrote: post #95

Earlier quoted context omitted.

rethinking your comment "was that all" I am listening to the stream now and had a thought. Most of the new models that have come out in the past few weeks have been great at coding and logical reasoning. But 4o has been better at creative writing. I am wondering if 4.5 is going to be even better at creative writing than 4o.

I still find all of them lacking on creative writing. The models are severely crippled by tokenization, complete lack of understanding of language rhythm. They can’t generate a simple haiku consistently, something larger is more out of reach. For example, give it a piece of poetry and ask for new verses and it just sucks at replicating the language structure and rhythm of original verses.

I might sound crazy but honestly fine-tuned GPT-3 absolutely blows all of these modern models out of the water when it comes to creative writing.

Maybe it was less lobotomized, or less covered in the prompt equivalent of red tape. Or maybe you just need to have a little bit of lunacy for fun creative writing. The new models are so much more useful, but IMO they don’t have even come close to GPT-3.

Re: GPT-4.5

#943

Post by herval »

MichaelZuo wrote: post #889
herval wrote: post #848

Earlier quoted context omitted.

I don't have an accurate benchmark, but in my personal experience, gpt4o hallucinates substantially less than gpt4. We solved a ton of hallucination issues just by upgrading to it...

How much did you use the original GPT-4-0314? (And even that was a downgrade compared to the more uncensored pre-release versions, which were comparable to GPT-4.5, at least judging by the unicorn test)

I don't recall the original version we used unfortunately :(

in our case, the bump was actually from gpt-4-vision to gpt-4o (the use case required image interpretation)

It got measurably better at both image cases and text-only cases

Re: GPT-4.5

#944

Post by iainctduncan »

So will we get people admitting they've been total jerks to Gary Marcus yet? Is he hyperbolic and over the top sometimes? Sure. Is he right about scaling not getting LLMs to AGI? Sure is looking like it.

I, for one, am so sick of listening to LLM fanboys wax on about "AGI" when they don't know the first goddamned thing about actual human cognition. For all his faults, Marcus studied human intelligence at a PhD level. I have only done a wee bit (music cognition as part of an interdisciplinary PhD I'm doing) and it's obvious to me, my supervisor (AI prof for 25 years) and anyone who knows anything about human cognition that LLMS are not going to get anywhere close to "thinking as well as a human" by scaling.

Bubble can't burst soon enough for me, sigh.

Re: GPT-4.5

#945

Post by ethbr1 »

tinco wrote: post #788
ethbr1 wrote: post #591

Earlier quoted context omitted.

It's nice to know the new Turing test is generating effective VC pitch decks.

Joke's on us, the VC's are using LLM's to evaluate the pitch decks.

We all thought the singularity was going to be exceeding human capacity for change.

It'd be funny if it's actually full-automated, closed-loop automation of capital allocation markets.

"Why are we doing this? How much money are we getting?" -> "I dunno. It's what the models said."

Re: GPT-4.5

#946

Post by cynicalpeace »

sponnath wrote: post #922
cynicalpeace wrote: post #897

Earlier quoted context omitted.

"There Is No AI Revolution" Good write-up. But it focuses too much on the big companies. Many indiehackers have figured out how to make profit with AI: 1. No free tier. Just provide a good landing page. 2. Ship fast. Ship iteratively. Employ no one besides yourself. 3. Profit. The old silicon valley idea that you need to raise a bunch of money, hire a bunch of devs, and scale a ton to satisfy investors is dying rapid…

If this were true wouldn't we be seeing a massive devaluation of software? (or alternatively increased demand for complex software)

It is true because levelsio, eric smith, maker of interviewcoder, etc exist.

Re: GPT-4.5

#947

Post by janoc »

cynicalpeace wrote: post #897
janoc wrote: post #867

Just going to put this here: https://www.wheresyoured.at/wheres-the-money/

"There Is No AI Revolution" Good write-up. But it focuses too much on the big companies. Many indiehackers have figured out how to make profit with AI: 1. No free tier. Just provide a good landing page. 2. Ship fast. Ship iteratively. Employ no one besides yourself. 3. Profit. The old silicon valley idea that you need to raise a bunch of money, hire a bunch of devs, and scale a ton to satisfy investors is dying rapid…

And pretty much most of them just resell OpenAI/Anthropic/Google/Meta's APIs and model access, with something repackaged on top.

And none is remotely profitable.

Re: GPT-4.5

#948

Post by hadlock »

matwood wrote: post #660
HDThoreaun wrote: post #572

Earlier quoted context omitted.

I use LLMs everyday to proofread and edit my emails. They’re incredible at it, as good as anyone I’ve ever met. Tasks that involve language and not facts tend to be done well by LLMs.

> I use LLMs everyday to proofread and edit my emails. This right here. I used to spend tons of time making sure my emails were perfect. Is it professional enough, am I being too terse, etc…

The first profitable AI product I ever heard about (2 years ago) was an exec using a product to draft emails for them, for exactly the reasons you mention.

Re: GPT-4.5

#949

Post by gom_jabbar »

ethbr1 wrote: post #945
tinco wrote: post #788

Earlier quoted context omitted.

Joke's on us, the VC's are using LLM's to evaluate the pitch decks.

We all thought the singularity was going to be exceeding human capacity for change. It'd be funny if it's actually full-automated, closed-loop automation of capital allocation markets. "Why are we doing this? How much money are we getting?" -> "I dunno. It's what the models said."

This is basically Nick Land's core thesis that capitalism and AI are identical.

> "I dunno. It's what the models said."

The obvious human idiocy in such things often obscures the actual process:

"What it [capitalism] is in itself is only tactically connected to what it does for us — that is (in part), what it trades us for its self-escalation. Our phenomenology is its camouflage. We contemptuously mock the trash that it offers the masses, and then think we have understood something about capitalism, rather than about what capitalism has learnt to think of the apes it arose among." [0]

[0] https://retrochronic.com/#romantic-delusion

Re: GPT-4.5

#950

Post by hadlock »

chamomeal wrote: post #941
vjerancrnjak wrote: post #697

Earlier quoted context omitted.

I still find all of them lacking on creative writing. The models are severely crippled by tokenization, complete lack of understanding of language rhythm. They can’t generate a simple haiku consistently, something larger is more out of reach. For example, give it a piece of poetry and ask for new verses and it just sucks at replicating the language structure and rhythm of original verses.

I might sound crazy but honestly fine-tuned GPT-3 absolutely blows all of these modern models out of the water when it comes to creative writing. Maybe it was less lobotomized, or less covered in the prompt equivalent of red tape. Or maybe you just need to have a little bit of lunacy for fun creative writing. The new models are so much more useful, but IMO they don’t have even come close to GPT-3.

Do you have an example prompt? I've been trying to get ChatGPT to tell a customized children's story similar to what you would see in a commercial story book but it just keeps giving me what's basically a summary of what you might read about in the book.
Post reply on HN