Adobe Inc. rolled out new features of its Firefly AI Tuesday and is so confident its artificial intelligence model protects copyrighted material that it is offering to pay legal bills of users sued for infringement.
On Tuesday, Adobe
ADBE,
kicked off its MAX conference in Los Angeles, introducing its next generation of AI image generator, Firefly Image Model 2, and the company is using the controversy surrounding AI — and one of the main sticking points in recent actors and writers strikes — as a selling point.
“We’re trying to use a model that doesn’t contain IP [intellectual property], so it’s safe for the commercial user, so it cannot generate anything that could infringe,” Alexandru Costin, who heads generative AI at Adobe, told MarketWatch recently in an interview.
“This can be seen as a limitation,” Costin said. “We see it as a strength because it actually enables you to be confident that you can’t be sued because you’re infringing.”
If you were to prompt Firefly to generate a certain cartoon family property owned by Fox drawn in the style of a certain Warner Brothers superhero team, the AI would be somewhat hampered in the request as it is not trained on unlicensed copyrighted material, Adobe said. To back that up, Costin said Adobe is offering an indemnity to customers.
“If they get sued because we generate somebody else’s IP, we cover their legal fees because we trust that our model doesn’t, cannot infringe someone else,” Costin said.
Costin added that Adobe advocates that artists should be in control of their style, as Adobe Chief Legal Officer Dana Rao testified before Congress.
If, however, someone else’s copyrighted style is used, Costin said Adobe stores thumbnails on its servers to guarantee that if an infringement claim comes forward, the company can provide support in the future.
Read: Adobe hopes to acquire new users, firm up margins as it lets Firefly AI out of its jar
Adobe, like Microsoft Corp.
MSFT,
and Alphabet Inc.’s
GOOG,
GOOGL,
Google and every other software company competing in the AI field, is incorporating features of the technology into their existing suite of subscription products.
Adobe also introduced its Firefly Vector Model, which generates scalable vector images that the company bills as the “world’s first generative AI model focused on producing vector graphics.”
In a demo provided to MarketWatch, the model was able to create scalable vector images on demand that could be easily placed into an existing picture. The model also made it easier for the user to manipulate objects by separating them from their background with a click, rather than having to manually outline the object.
The company also introduced its Firefly Design Model, which includes hundreds of templates for things like a new design model that includes templates for posters, flyers and invitations.
Back in September, Adobe released the commercial version of Firefly, which has generated more than 3 billion images since being released as a beta version in March. In its earnings report right after the launch, Adobe said it expected to keep operating margins in the mid-40s even with the high cost of maintaining AI computing costs.
Adobe is scheduled to host a meeting with Wall Street analysts later in the day, beginning at 5 p.m. Eastern.
Read the full article here