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

ChatGPT Images 2.0

openai.com

Post reply on HN
Print view View original

891–900 of 1001 posts

Re: ChatGPT Images 2.0

#891

Post by fennecfoxy »

marricks wrote: post #316
simonw wrote: post #15

I've been trying out the new model like this: OPENAI_API_KEY="$(llm keys get openai)" \ uv run https://tools.simonwillison.net/python/openai_image.py \ -m gpt-image-2 \ "Do a where's Waldo style image but it's where is the raccoon holding a ham radio" Code here: https://github.com/simonw/tools/blob/main/python/openai_imag... Here's what I got from that prompt. I do not think it included a raccoon holding a ham radio…

Like... this has things that AI will seemingly always be terrible at? At some point the level of detail is utter garbo and always will be. An artist who was thoughtful could have some mistakes but someone who put that much time into a drawing wouldn't have: - Nightmarish screaming faces on most people - A sign that points seemingly both directions, or the incorrect one for a lake and a first AID tent that doesn't exi…

No, it won't be. I did indeed get the same problems when trying to generate my own image for it.

However as someone who's mucked about with local image generation as well - I'd say that this is a problem with their implementation, it doesn't resolve fine detail because majority of requests it won't matter/it drastically increases compute requirements.

With local image generation bad features/incorrect fingers/disfigurement etc has been solved for a long time.

I think their new process involves multiple steps including sketching/fleshing out the idea before adding detail. The step that would fix this would be outpainting or similar to tile based upscaling.

From what I understand of image generation models they also struggle with fine detail in general because they aren't really trained for that. However for each tiny chunk of a detailed image like that there's nothing to say they can't allocate a 500x500 chunk for it to work in as its "idea/reference space" and then transpose that into the main image being generated - i.e. generate image features separately rather than all together.

Re: ChatGPT Images 2.0

#892

Post by Kiro »

makerofthings wrote: post #752
bananaflag wrote: post #741

Earlier quoted context omitted.

Code can be beautiful or ugly but that doesn't make it art. Art is not just about beauty, it is about expressing the mind (feelings, experience etc) of the author. AI will never do that (except if it learns to express its own experiences, which would be art, but not something competing with human art; it would be like if we had contact with alien art).

Code is my art and is how I express myself. I agree that nothing that AI does is art.

Code in general is obviously not art, which is all that matters here.

Re: ChatGPT Images 2.0

#893

Post by fennecfoxy »

tsimionescu wrote: post #774
arghwhat wrote: post #719

Earlier quoted context omitted.

This is, in my opinion, attempting to say the right thing with entirely the wrong perspective: The people you say are getting "shafted" always got shafted. Their works are the inspiration for all artists and people who lay their eyes on it - maybe they got paid when they made the work, maybe they managed to sell it, but probably not. And still, other artists (and machines) will use remember and be inspired by it, som…

You're ignoring the biggest problem here: the concentration and extraction of wealth. The sum total of human artists were previously getting those billions of dollars, and now it's OpenAI (and Anthropic, and Google, and Microsoft, and maybe a handful of other players) getting it. Now, maybe it actually used to be hundreds of millions of dollars, and they've grown it to billions, and maybe they deserve some of that -…

Yes, look at how many historical inventors (like the Blue LED, the guys struggling to convince Gates and Balmer to make the Xbox) etc get/got nothing for their efforts compared to the huge sums raked in by the very people actively trying to prevent them from building the idea that made all the money.

AI is hugely beneficial to our species. Our tribalism and "yeah well they earned it!" response to capitalism's rampant production of billionaires is the real problem, not technology.

Why are footballers and movie celebrities paid 50$m a year? There's the answer.

Re: ChatGPT Images 2.0

#894

Post by fennecfoxy »

Oarch wrote: post #668

Every groundbreaking new AI release feels like a volley of cannonfire towards the soul. Oof.

Lmao, and pre-AI the world wasn't trending towards cost-reduced mass-produced executive-driven content?

And the average member of our species lapping it up like being a mindless consumer is going out of style?

Re: ChatGPT Images 2.0

#895

Post by fennecfoxy »

mercacona wrote: post #683

Every improvement in image generation seems to reduce the value of the images themselves. When anything can be faked or created in seconds, what is an image really worth? With text or code, you can dig into a meaningful dialogue because their reality is digital too. But images become like the plain people to show up photo frames. I guess it's just a completely personal feeling.

Nah I'm gonna generate the hecky out of "relevant for this one presentation" little cliparts to add to my powerpoints.

But then I'm still going to take photos on film and enjoy Sunday afternoons in the darkroom doing prints.

It's possible to compromise and/or use the right tool for the right job. Saving me time for something of little or fleeting importance? AI. Making me feel good/physical work with hands/emotion chemicals in my brain? Film/traditional media.

Re: ChatGPT Images 2.0

#897

Post by valarauko »

vunderba wrote: post #83

OpenAI’s gpt-image-1.5 and Google’s NB2 have been pretty much neck and neck on my comparison site which focuses heavily on prompt adherence, with both hovering around a 70% success rate on the prompts for generative and editing capabilities. With the caveat being that Gemini has always had the edge in terms of visual fidelity. That being said, gpt-image-1.5 was a big leap in visual quality for OpenAI and eliminated m…

That's lovely. My own personal benchmark has been to ask the various models to generate a functional pair of novelty New Year's Eve glasses on a person, that don't just plonk the year onto the top of regular frames.

Re: ChatGPT Images 2.0

#898

Post by arghwhat »

BeetleB wrote: post #864
arghwhat wrote: post #719

Earlier quoted context omitted.

This is, in my opinion, attempting to say the right thing with entirely the wrong perspective: The people you say are getting "shafted" always got shafted. Their works are the inspiration for all artists and people who lay their eyes on it - maybe they got paid when they made the work, maybe they managed to sell it, but probably not. And still, other artists (and machines) will use remember and be inspired by it, som…

> The person (singular) that is actually getting "shafted" at each use is the artist you didn't hire to do the job of making your new work, because it is their skill that got replaced. 1% Yes, and 99% No. Over 99% of uses would not have resulted in hiring someone to do the work had these models not existed as you yourself acknowledge.

Yes, but this is a bit of an oversimplification. The "99%" tends to be either: 1. Pointless throwaway content which we can just ignore as a new source of noise, 2. Something that could have ended up being a $5 commission[^1] to a kid somewhere out there but now never will be.

Those numbers are also a bit too aggressive - it's easy to miss what kind of gig work exist out there. PowerPoint as a service is a thing on Fiverr for example. A horrible, horrible thing, but a thing none the less.

^1: not at all what art costs, but someone trying to get started might do quick sketches at those prices

Re: ChatGPT Images 2.0

#900

Post by wahnfrieden »

fennecfoxy wrote: post #886
wahnfrieden wrote: post #860

Earlier quoted context omitted.

That's what makes it a fair evaluation of its limits

I mean asking these transformers to do maths has always been the wrong task. It's like we're now considering "it doesn't have x tools built with traditional code built in". Though I suppose we're testing their model + agent harness here as well. It really _should_ have all of those tools/reasoning available to accomplish a task like the above without issue.

It's only been the wrong task because they've been deficient at it and expensive to use, so we had workarounds. They are getting better at these tasks and cheaper (sometimes). It's fair to evaluate even if there are more economical and accurate alternatives available.
Post reply on HN