January 20, 2023

Will Adobe de-throne Dall-e and become the go-to choice of creative AI?  

Adobe has been secretly collecting data and training it's AI.

Many fortunes were built in the 19th century on oil, railway, and grabbing up Manhattan real estate. Vanderbilts, Rockefellers, whateverellerilts.

The family patrons did such a great job that their descendants still live off of that haul.  

In the 21st century, we don’t care how cliche and overused this sounds, the equivalent of that “oil” is data. But with a slight difference.

Every man could’ve tried to find oil, snatch up some Manhattan real estate or figure out how to reap the immense fortunes railway can choo choo in.

Mere access to capital was the hard part.

But with data, it isn’t the case. You have a chicken and an egg problem.

To use the data for insights for machine learning, you need to get the data somewhere.

And if you’re a starting business, you can’t easily get thousands of terabytes of data.

And if you don’t have data, you can’t build your own AI, etc, which might be a table-stakes feature in many industries in a couple of years.

You might try to scrape data from somewhere, but this could mean legal issues. Already, this is becoming more and more of a problem, and companies are turning off their data wells.

So small up-and-comers aren’t competitive.

Therefore the big companies get even bigger and harder to de-throne.

Proprietary data rules.

And the bigger and more established you are, the more data you have.

And the more Ferraris you can buy with all OF THAT FCKING MONEY!

Adobe knows that.

And as some people just discovered, they are using their user’s data to train their creative AI.

Whenever you’re uploading photos to lightroom or editing something in Photoshop or premiere. By default, Adobe is using ALL OF THAT to train their AI.

That’s user-generated content right there.

Everyone is, by default, opted into that. So if you’d like to opt out, here’s a guide.

Oh and all the artists, who of course feel threatened about it, are furious

Adobe is much better equipped for visual AI success than OpenAI.

Or any other player in the space because they have so much data. And a huge distribution engine.

OpenAI has to navigate the restrictions on scraping the web, or existing art, literature, etc. Maybe purchase data, and it might get locked out of different sources.

Meanwhile, Adobe has millions of new files edited-created in its ecosystem daily.

Would love to see what comes out of that.

And most of all, does their AI succeed at creating hands, faces and text?

But here’s a wild idea for the end.

Could companies create data APIs for others to use their data?

No personable information gets shared with 3rd party, but solely the information they could use for training their machine learning systems do that.

Why should they share the data? Well yeah, it is a moat.

But certain companies, especially VC-backed ones, must always be hitting their outrageous revenue & growth targets. And this might be an extra set of legs for them.

Vixus is a mediatech content hub covering stories from generative AI to VR and beyond.

Liked this story?

Subscribe for our weekly 5-minute brief and get more stories like these straight into your inbox.