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

Search 5.8B images used to train popular AI art models

haveibeentrained.com

Post reply on HN
Print view View original

141–150 of 187 posts

Re: Search 5.8B images used to train popular AI art models

#141

Post by ghaff »

mrweasel wrote: post #129
ericjang wrote: post #38

It would be a stunning twist of irony if this website uploaded images to a proprietary image dataset used for training AI models, pitching "uncorrelated data"

That highlights one of my "concerns" about AI training sets. There seems to be a very real risk of accidentally adding an image that you have no rights to, so what happens if someone finds out and demand their image removed from all models trained on that image? You can't really back an image out of a model, you can only retrain without that image. Creative Commons should add a new license that prohibits the use of o…

Because Creative Commons really needs another non-open license variant like non-commercial that no one really understands what they allow and don't allow so you're better off just not using them if you're being conservative.

Re: Search 5.8B images used to train popular AI art models

#142

Post by bdowling »

JimDabell wrote: post #130
bdowling wrote: post #123

Earlier quoted context omitted.

If there was an actual case, the question would be whether the defendant had access to the original work. If the original work was used to train the AI model, then the answer is yes. It's not necessary for the model to contain a copy of the original work.

That wouldn’t be the question because nobody is disputing the access to the original work. You mentioned two factors. That’s the first, which nobody is questioning. The thing people are questioning is the second factor, the reproduction of the original work. A copy must be made for there to be copyright infringement. Who has shown a copy has been made? If a copy has been observed to occur, it’s trivial to demonstrate…

I may have misunderstood. I thought we were talking about an AI output that might be substantially similar to an original used to train the AI. If we're considering only the use of the original as a training image, then I don't know if that would infringe.

Re: Search 5.8B images used to train popular AI art models

#143

Post by poulpy123 »

gigel82 wrote: post #9

It's surprising how poorly labeled these images are; who is curating this collection? Can't they crowd-source a proper labeling project - I wonder how much better things like Stable Diffusion would be if its training would include correct, complete labels for the images. I'm sure lots of folks would willingly spend a few minutes here and there to aid with the labeling if it means they get to enjoy the model for free.

labeling 6 billions pictures, even at 10 labels per second would take 20 years, so I guess nobody

Re: Search 5.8B images used to train popular AI art models

#144

Post by tomrod »

andybak wrote: post #138
tomrod wrote: post #90

Earlier quoted context omitted.

> "I stole your money and several other people's money and then put it into a bank account. Tough luck! It's too late, you are unlikely to prove which of the dollar bills in account are yours!" Sort of falls flat.

I don't think this analogy holds up.

If someone's copyrighted material was used to train a model, what analogy would you use?

Re: Search 5.8B images used to train popular AI art models

#145

Post by Kye »

ma2rten wrote: post #59
gigel82 wrote: post #9

It's surprising how poorly labeled these images are; who is curating this collection? Can't they crowd-source a proper labeling project - I wonder how much better things like Stable Diffusion would be if its training would include correct, complete labels for the images. I'm sure lots of folks would willingly spend a few minutes here and there to aid with the labeling if it means they get to enjoy the model for free.

Those images are not really "labeled". They just scraped the alt text. A lot of the recent advances in AI have been by using lower quality large scale web data, instead of hand labeling. The noise will average out. Hand labeled data can be used for finetuning.

This is the same thinking that leads to incredible astrophotography using stacked long exposure images that appear to be pure noise individually.

Re: Search 5.8B images used to train popular AI art models

#146

Post by yreg »

poulpy123 wrote: post #143
gigel82 wrote: post #9

It's surprising how poorly labeled these images are; who is curating this collection? Can't they crowd-source a proper labeling project - I wonder how much better things like Stable Diffusion would be if its training would include correct, complete labels for the images. I'm sure lots of folks would willingly spend a few minutes here and there to aid with the labeling if it means they get to enjoy the model for free.

labeling 6 billions pictures, even at 10 labels per second would take 20 years, so I guess nobody

Could be done by reCAPTCHA, although might be more difficult to classify the results (whether a user passed the test) compared to their usual challenges.

Re: Search 5.8B images used to train popular AI art models

#147

Post by yreg »

simandl wrote: post #16
gigel82 wrote: post #9

It's surprising how poorly labeled these images are; who is curating this collection? Can't they crowd-source a proper labeling project - I wonder how much better things like Stable Diffusion would be if its training would include correct, complete labels for the images. I'm sure lots of folks would willingly spend a few minutes here and there to aid with the labeling if it means they get to enjoy the model for free.

This is Laion-5B, you can read more about it here: https://laion.ai/blog/laion-5b/ Imagen and Stable-Diffusion both used subsets of this full 5.8B image set.

Is Imagen actually trained on a subset of Laion-5B and nothing else? I've heard they used huge internal data sets.

Re: Search 5.8B images used to train popular AI art models

#148

Post by 0xrisk »

chongli wrote: post #127
UncleEntity wrote: post #44

Earlier quoted context omitted.

The same artists who have been pitching a fit about how the robots are coming for their jobs? Pretty sure they are trying to figure out a way their art can’t be used to train an AI and won’t be willingly participating in making it easier.

I’m skeptical that Al will replace artists. The existence of computer chess and go, though superior in every way, has not killed interest in human competition. I see no reason we couldn’t have the same thing happen with art: people see AI art as a novelty and something worth studying but otherwise preferring human art for its relevance to the human condition.

I agree. It was telling that in what I perceived to the non story of the midjourney created image winning the county art fair, many of the other prizes were for activities that machines have been better at than humans for decades. Fruit canning. Furniture construction. The replacement narrative is overblown, but these tools are going to augment human artists considerably.

Re: Search 5.8B images used to train popular AI art models

#150

Post by throw_m239339 »

quantified wrote: post #111
throw_m239339 wrote: post #98

Earlier quoted context omitted.

> but porn is by definition art No, that's why it's called pornography, very much to differentiate it from something "noble" like art, it's vulgar and trivial, by the very definition of the word it isn't art.

Etymology is "writing about prostitutes" [ https://www.merriam-webster.com/dictionary/pornography ]. So you got me there. In the realm of visual depictions, it's just as much art as cave art. Art is a broad classifier. Still lifes are art. Landscapes are art. Audubon's drawings of birds are art. Porn is film, photography and drawing, with a specific subject matter. If Iron Man comics are art, it is too. "Noble" is no…

> Art is a broad classifier.

"everything I say is art is art", you basically. Well anything I say isn't art isn't art. Especially pornography in general, which sole purpose is sexual gratification of the audience through sexual exploitation, rape via human trafficking.

There is nothing snob is saying that. There something egregious and dehumanizing in saying what you say on the other hand and trying to normalize the very definition of vulgarity.

Post reply on HN