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

Adobe Firefly: AI Art Generator

adobe.com

Post reply on HN
Print view View original

221–230 of 496 posts

Re: Adobe Firefly: AI Art Generator

#221

Post by unreal37 »

ftufek wrote: post #218
adam_arthur wrote: post #211

Earlier quoted context omitted.

Creating AI models has proven to simply be easier than other past innovations. Much lower barrier to entry, the knowledge seems to be spread pervasively within months of breakthroughs. People seem to take offense at this idea, but the proof is in the pudding. Every week there's a new company with a new model coming out. What good did Google's "AI Talent" do for them when OpenAI leapfrogged them with only a few hundre…

It's easier for large rich companies with infrastructure and datasets. It's very hard for small startups to build useful real world models from scratch, so you see most people building on top of SD and APIs, but that limits what you can build, for example it's very hard to build realistic photo editing on top of stable diffusion.

Someone was able to replicate GPT 3.5 with $500. The training of models is getting very cheap.

[1] https://newatlas.com/technology/stanford-alpaca-cheap-gpt/

Re: Adobe Firefly: AI Art Generator

#222

Post by adam_arthur »

ftufek wrote: post #218
adam_arthur wrote: post #211

Earlier quoted context omitted.

Creating AI models has proven to simply be easier than other past innovations. Much lower barrier to entry, the knowledge seems to be spread pervasively within months of breakthroughs. People seem to take offense at this idea, but the proof is in the pudding. Every week there's a new company with a new model coming out. What good did Google's "AI Talent" do for them when OpenAI leapfrogged them with only a few hundre…

It's easier for large rich companies with infrastructure and datasets. It's very hard for small startups to build useful real world models from scratch, so you see most people building on top of SD and APIs, but that limits what you can build, for example it's very hard to build realistic photo editing on top of stable diffusion.

Most of the cutting edge models are coming from companies with a few dozen to a few hundred people. Stability AI is one example.

Training an AI model, while expensive, is vastly cheaper than most large scale products.

This wave will be nothing like the SaaS wave. Hyper competitive rather than weakly-competitive/margin preserving

Re: Adobe Firefly: AI Art Generator

#223

Post by LegitShady »

stravant wrote: post #83
LegitShady wrote: post #27

I wonder about how companies are approaching letting their customers understand what is and isn't copyrightable with the results of these ai models. The us copyright office has made it clear that elements of designs that lack human authorship cannot be copyrighted. A prompt is not enough - the same way if you provided a prompt to an artist while commissioning a work does not grant you copyright to that work. The resu…

I don't see how this will possibly be relevant. AI generated content is about to sprint so far ahead of the existing legal framework that something will just have to give.

on the opposite - existing legal frameworks protect human works, and altering them to allow corporations to own any idea they can provide a prompt for is directly harmful to society and the future of intellectual property in every country.

using ml is already powerful and fast. protecting ideas from mass corporate ownership so that your grandchildren will be allowed to think freely is more important than chat gpt or stable diffusion or whatever algorithm replaces them next week.

Re: Adobe Firefly: AI Art Generator

#224

Post by ftufek »

unreal37 wrote: post #221
ftufek wrote: post #218

Earlier quoted context omitted.

It's easier for large rich companies with infrastructure and datasets. It's very hard for small startups to build useful real world models from scratch, so you see most people building on top of SD and APIs, but that limits what you can build, for example it's very hard to build realistic photo editing on top of stable diffusion.

Someone was able to replicate GPT 3.5 with $500. The training of models is getting very cheap. [1] https://newatlas.com/technology/stanford-alpaca-cheap-gpt/

I've tried it, sure it's good, but not even close to the real thing. But yes it's getting cheaper through better hardware, better data and better architectures. Also it builds on Facebook's models that were trained for months on thousands of A100 GPUs.

Re: Adobe Firefly: AI Art Generator

#225

Post by lmarcos »

petilon wrote: post #65
m3kw9 wrote: post #21

This will supercharge artists, it won’t replace them because details matter and when you need to get it just right, you cannot just keep rolling the dice by “prompting better”

Yes, but by "supercharging" artists, each artist will be able to do more, which means fewer artists will be needed.

That artists will be able to do more not necessarily mean that fewer of them will ne needed. I bet the opposite actually. Companies will want to produce more (with more resources) not the same (with less resources).

Re: Adobe Firefly: AI Art Generator

#226

Post by mesh »

egypturnash wrote: post #149
mesh wrote: post #124

Earlier quoted context omitted.

>That is, these companies are largely not doing the hard part, which is creating and training these models in the first place. fyi, for Adobe Firefly, we are training or models. From the FAQ: "What was the training data for Firefly? Firefly was trained on Adobe Stock images, openly licensed content and public domain content, where copyright has expired." https://firefly.adobe.com/faq (I work for Adobe)

I am definitely glad to see attention being paid to ethical sourcing of training images but I am curious: did the people who made all those stock images get paid for their work being used for training? Did they check a box that explicitly said "Adobe can train an AI on my work"? Or is there a little clause lurking in the Adobe Stock agreement that says this can be done without even a single purchase happening?

We are working on a compensation model for Stock contributors for the training content, and will have more details by the time we release.

The training is based on the licensing agreement for Adobe contributors for Adobe Stock.

(I work for Adobe)

Re: Adobe Firefly: AI Art Generator

#227

Post by ftufek »

adam_arthur wrote: post #222
ftufek wrote: post #218

Earlier quoted context omitted.

It's easier for large rich companies with infrastructure and datasets. It's very hard for small startups to build useful real world models from scratch, so you see most people building on top of SD and APIs, but that limits what you can build, for example it's very hard to build realistic photo editing on top of stable diffusion.

Most of the cutting edge models are coming from companies with a few dozen to a few hundred people. Stability AI is one example. Training an AI model, while expensive, is vastly cheaper than most large scale products. This wave will be nothing like the SaaS wave. Hyper competitive rather than weakly-competitive/margin preserving

I wrote it from the perspective of a small startup (But yes I agree, it will be very competitive with much smaller margins.

Re: Adobe Firefly: AI Art Generator

#228

Post by krisoft »

unreal37 wrote: post #219
egypturnash wrote: post #149

Earlier quoted context omitted.

I am definitely glad to see attention being paid to ethical sourcing of training images but I am curious: did the people who made all those stock images get paid for their work being used for training? Did they check a box that explicitly said "Adobe can train an AI on my work"? Or is there a little clause lurking in the Adobe Stock agreement that says this can be done without even a single purchase happening?

Nobody owes creators who have been paid fully for their work "extra" compensation just because AI is involved. Assuming they have been paid, the work belongs to Adobe.

> Assuming they have been paid

That is the question we are asking, yes. Based on the reading of the contributor agreement it sounds like Adobe doesn’t have to pay a cent to the creators to train models on their work.

Does that sound fair to you?

Re: Adobe Firefly: AI Art Generator

#229

Post by freedomben »

airstrike wrote: post #172
open-paren wrote: post #81

Adobe employee, not in Creative Cloud. We got access to this in beta a week ago and it was an instant hit across the whole company. This is just the tip of the iceberg and there is a lot of really cool, in-house products around generative AI. The team is going to great lengths to make this ethical and fair (try and generate a photo of a copyrighted character like Hello Kitty or Darth Vader). I'm excited to see the fi…

> The team is going to great lengths to make this ethical and fair (try and generate a photo of a copyrighted character like Hello Kitty or Darth Vader) are you saying it won't work? if that's the case, that seems really silly. actually, it goes against everything I believe in (as well as my understanding of even the kindest meaning of the word "hacker"). it drives me up the wall, it makes my blood boil who is going…

GP works for Adobe, and Adobe's bread and butter are the professional creators who would love a world where there is hardware DRM on your eyes and you can't even see their creations or a likeness of them without paying a royalty (and one to "rent" the memories of the visualization, not to "own" the memories like we do now). While I largely agree with you, the GP post is exactly what I would expect from an Adobe person.

Re: Adobe Firefly: AI Art Generator

#230

Post by UberFly »

Adobe will need to allow the importing of custom models, or else their product will be too limited. That then allows the "unofficial" use of copyrighted material for image generation. They're definitely starting from a place of advantage leveraging their already vast open-license stock portfolios.
Post reply on HN