[ad_1]
At the annual GPU Technology Conference, it announced a suite of cloud services designed to help businesses build and run creative AI models trained on custom data and created for “domain-specific tasks” like writing ad copy.
Each individual cloud service under the Nvidia AI Foundation brand – Nvidia NeMo for language models and Nvidia Picasso for image, video and 3D content generation – includes pre-trained models, data processors, APIs and support from Nvidia engineering staff. Once models are ready for deployment, enterprises can run them in Nvidia’s cloud or elsewhere.
Starting today, the NeMo Creator AI cloud service is in early access. The Picasso service is in private preview. Enrollment for both requires submitting an application.
“Generative AI is accelerating the adoption of AI and revolutionizing countless industries,” said Jensen Huang, founder and CEO of NVIDIA, in a press release. “Nvidia AI Foundation enables enterprises to customize foundational models with their own data to harness the most valuable human resources – intelligence and creativity.”
With the NeMo cloud service, Nvidia said developers can add domain-specific knowledge and application skills to text generation models they use, such as chatbots, enterprise search and customer service. Models on the service are regularly updated with additional training data, the company said.
About Nvidia Picasso, a cloud service for building text-to-image, text-to-video and text-to-3D object models. (Although text-to-video and text-to-3D object models are now smaller than their text-to-image brethren, vendors are increasingly experimenting with them — anticipating increased demand.) Customers can use Picasso to train. and deploy models on their data, either using pre-trained models from Nvidia or optimizing and running their own models.
In a big customer win, Nvidia says it’s working with Adobe to jointly develop “the next generation of commercially viable generative AI models” on the Picasso platform. They will be brought to market through Photoshop, Premiere Pro and After Effects, according to Nvidia, complementing Adobe’s recently launched Firefly generative AI products.
No doubt eager to avoid getting into a copyright wrangle over generative AI, Nvidia says it wants to partner with another party, Getty Images, to train “responsible” generative text-to-image and text-to-video models. The models are created using content from the Getty Images library, says Nevidi, with Getty Images paying royalties to the artists on any revenue generated from the models.
That’s significant, as Getty Images currently uses another proprietary AI service provider called Stable Diffusion to train the images on Stable AI without obtaining permission from Getty or the rights holders. Nivea, it seems, didn’t want to take a chance — or maybe it wanted to hedge its bets, especially since similar lawsuits are working their way through the courts.
In a related announcement, Nvidia said it plans to partner with Getty Images competitor Shutterstock, specifically to train text-to-3D object models using Picasso. Shutterstock’s assets and metadata will be used for training, and Shutterstock will compensate artists through a recently launched contribution fund, according to Nvidia.
Shutterstock plans to complement its existing text-to-image tool through a partnership and licensing agreement with OpenAI, and to offer the model on its own platform once it’s ready to go.
[ad_2]
Source link
