AI Hustle: News on Open AI, ChatGPT, Midjourney, NVIDIA, Anthropic, Open Source LLMs: Adobe Rewards Artists for Firefly AI Training Content

Welcome to the OpenAI podcast, the podcast that opens up the world of AI in a quick and

concise manner.

Tune in daily to hear the latest news and breakthroughs in the rapidly evolving world

of artificial intelligence.

If you've been following the podcast for a while, you'll know that over the last six

months I've been working on a stealth AI startup.

Of the hundreds of projects I've covered, this is the one that I believe has the greatest

potential.

So today I'm excited to announce AIBOX.

AIBOX is a no-code AI app building platform paired with the App Store for AI that lets

you monetize your AI tools.

The platform lets you build apps by linking together AI models like chatGPT, mid-journey

and 11 labs, eventually will integrate with software like Gmail, Trello and Salesforce

so you can use AI to automate every function in your organization.

To get notified when we launch and be one of the first to build on the platform, you

can join the wait list at AIBOX.AI, the link is in the show notes.

We are currently raising a seed round of funding.

If you're an investor that is focused on disruptive tech, I'd love to tell you more

about the platform.

You can reach out to me at jaden at AIBOX.AI, I'll leave that email in the show notes.

So Adobe's Firefly generative AI models have now transitioned from beta to a general release

available across Adobe's Creative Cloud, Adobe Express and Adobe Experience Cloud Platform.

So this groundbreaking move has really kind of made the technology more widely accessible

than ever.

And I think this is making a significant step in Adobe's quest for AI-powered creative

solutions.

So of course the road to Firefly's commercial launch wasn't without its challenges.

The company faced a whole bunch of questions around its compensation model for Adobe stock

contributors, especially those whose content contributed to Firefly's training data.

I think some of the tension kind of just revolves around the possibility of these photographers

and kind of illustrators losing business as users turn to Firefly for generating unique

images rather than just purchasing existing stock content.

In my opinion, I don't know what they expect because the alternative is people are just

going to go to like mid-journey and everything else that's out there right now and just get

it without them getting any compensation.

So in my opinion, if they're getting compensated, they should be happy.

Maybe in the future it can be kind of like Spotify where artists complain and got the

regulators or other people to kind of push up the price that they get paid per stream.

But as it stands, since the default would appear to be no one getting compensated, I

would take what you can get and then try to push the needle in the future.

That would be my strategy if I was a creator kind of complaining about this.

But of course, you know, people just want, maybe they just want their cake and they want

to be able to eat it as well.

So in any case, I think, you know, kind of addressing this concern, Adobe has announced

an annual bonus for eligible Adobe stock contributors, which is covering both standard and premium

stock collections.

So this bonus aims to recognize the value these contributors bring to Firefly's ecosystem.

Specifically, the initial bonus is going to go from, I believe, this like June of last

year to June of this year.

So essentially, that's kind of matching the data set Adobe used to train Firefly, aka

they used all of those things to train Firefly.

So they're going to say, like, hey, anyone that had content on the platform, during between

those two dates, we'll look at what the content was and we'll pay you for it.

I'm really curious like what they actually will, what the payout would be for that.

But in any case, determining these bonuses definitely is not straightforward, given the

difficulty in tracking how often, you know, a specific image contributes to a newly generated

Firefly image.

So Alexandru Kostin, who's Adobe's VP for generative AI and Sensei, provided a little

bit of insight into the company's approach here.

This is like actually so, so interesting.

But this is what he said, quote, the formula we have is a mix of the total number of assets

they have in the training data set, and the number of assets licensed.

So basically the popularity of the commercial success of their assets, which we take as

a proxy of the value of their assets through the lens of the marketplace.

Okay, for those of that weren't following, I think essentially what he's saying is they're

looking at like all of your assets that you have, all the pictures that you've taken,

how many times those pictures have been paid for just by users that are like, hey, this

is cool, pay for it.

And essentially, they're not like assigning a value to everything and every time it goes

through an AI model and trains something and gets spit out on the other end, the track

all the tokens and payments and then pay you.

It's not like that.

It's, I guess, too complicated.

It's not like streaming royalties on Spotify.

What they're actually doing is just saying you were like, you know, you sold a hundred

a month of this picture before.

So we're assuming you have like a popularity score of a hundred on this picture.

Now this, we're assuming that it's going to be when people are generating images with

AI, like that is your popularity score on the image and it's going to be the same for

images generated.

I know it's not scientific and so it's probably not exact.

It's probably fairly close.

So like big kudos to Adobe for kind of figuring out what I would call a quick and dirty like

approach to like they want to pay people.

They know this is controversial, but they also don't want to be slow on the technology.

They're not going to rewrite like the way these things, these whole models are rolling.

So in my opinion, I think that that's an acceptable response.

That is kind of cool, very interesting.

So Adobe kind of remains coy about disclosing any figures tied to these bonuses, right?

So they're going to pay bonuses, but no one knows how much it is.

And they're mentioning that these are largely dependent on individual contributors and how

their content is being utilized.

So the company did promise to divulge more details about the overall size of the bonus

pool in the future, whenever that is.

So on the user side, Adobe has implemented a quote generative credits system to meet

firefly usage.

So each time a user opts to generate an image using firefly, it's going to consume one credit

and the company has also redesigned the firefly web app to prevent automatic generation of

images until a user finalizes their actual settings.

So they have a bunch of different plans with different amounts of these generative credits

ranging from 25 for free users with an Adobe ID to 3,000 credits for Creative Cloud Pro

plus all apps subscribers, which is equivalent to I believe, if this is correct, what we're

seeing here around like 3,000 images that you'd get a month that you could generate.

So really impressive.

In any case, I think it's noteworthy to say, right, like this isn't, they can't just like

say, okay, pay 25 bucks a month and get as many images as you want.

It costs a lot of money, people would exploit that and like abuse it.

So they really have to like just assign credits and you can buy credits.

This is how a lot of things in AI are happening.

Makes a lot of sense.

You saw the same thing with like Jasper, the AI writing tool.

Yeah.

So I think this is the way it's going to go.

I think platforms are going to have to do credits for generating stuff because everyone

uses this stuff differently, unless, you know, sometimes maybe they'll just like, you

say it's a flat fee and then they kind of are averaging.

Some people are spending more than it's worth.

Some people are spending less than it kind of averages out as a profit for the company.

You may see that as well.

But I think credits is going to be especially for big companies like Adobe, the way that

this all ends up rolling out.

So I think, you know, running out of these credits doesn't lock users out of Firefly.

Instead the service is simply going to operate at a slower speed.

So that's another thing, right?

Adobe has also confirmed that additional subscription packs are going to be available,

although specific pricing details are yet to actually be revealed.

But I mean, that's kind of cool, right?

Like it's kind of like a chat GPT, you only have so many uses of GPT-4 in an hour.

But if you use them all up, it's not like you get locked out of your account.

It's just like you get, you know, put on the slower or worse GPT-3.5.

So Kostin elaborated on the need for the system stating, quote, all of these models

are very large.

They run in the cloud and are expensive to run and we are optimizing them for quality

of output, not for inference speed.

Because of that, we need to protect the user base to make sure they have access to the

service and to ensure fair use of the services.

So Firefly so far has generated, I think, over two billion images.

That's really impressive.

And I think, you know, Kostin kind of underscores that because AI was primarily trained using

Adobe stock images, the generated content is commercially safe for business use.

Also that's massive, right?

Like they have the licenses.

So when you use this, like no one's going to sue you and say, hey, like, I don't know,

whatever.

Businesses were concerned that there's going to be lawsuits, like, because this was trained

on an artist's data and the artist is going to go sue the business for generating something

with AI based off of their stuff without getting permission.

This is all this permission.

All the permissions done, Adobe had the licenses.

So this is good to go for businesses, which I think is also a big move and another reason

why Adobe is going to see a lot of success here.

So Adobe has also taken the extra step of indemnifying its enterprise users against potential lawsuits

arising from the use of Firefly generated images.

They said, like, look, if we did something wrong in the copyright and you're getting sued

because of anything that you generate on our platform, we're going to pay for the lawsuit.

So like big kudos to Adobe again.

So Adobe's dual focus, rewarding content contributors and, you know, regulating user access, I think

signals a balanced approach to ethical and business considerations as it scales its generative

AI capabilities.

The broader launch of Firefly showcases the company's kind of ongoing commitment to innovation.

They don't want to be left behind pretty much, right?

Like we're seeing massive innovation from all the top tech companies.

Adobe doesn't want to be left behind.

And I honestly think they've done an absolute standup job of getting this thing put together,

doing it right, you know, paying creators, you know, like telling businesses they're

not going to be liable to any lawsuits.

Like they really, like no one else is able to do what they've just done here.

And maybe it's because they have the deepest pockets, so it's left to be said if they're

going to completely sweep the floor when it comes to the enterprise use case, right?

And take some market share from Mid Journey, who seems to be the market leader right now.

But very, very well done.

Big kudos to Adobe.

That's all I can say.

This is going to be an interesting product to follow as we see this, get more features

and get more usage and more rollout broadly.

If you are looking for an innovative and creative community of people using chat GPT, you need

to join our chat GPT creators community.

I'll drop a link in the description to this podcast.

We'd love to see you there where we share tips and tricks of what is working in chat

GPT.

It's a lot easier than a podcast as you can see screenshots, you can share and comment

on things that are currently working.

So if this sounds interesting to you, check out the link in the comment.

We'd love to have you in the community.

Thanks for joining me on the OpenAI podcast.

It would mean the world to me if you would rate this podcast wherever you listen to your

podcasts and I'll see you tomorrow.

Machine-generated transcript that may contain inaccuracies.