In a recent presence ”Possible“Linkedin’s co-founder Reid Hoffman’s co-host-host, Google Dipmind CEO Demis Hassabis says Google has finally planned to combine his Gemny AI models with its VO video-generated models to improve the previous understanding of the physical world.
Hassabis said, “We have always created our Foundation Model, our Foundation Model, our Foundation Model,” Hassabis said, “and the reason we did [is because] Universal Digital Assistant, we have a vision for this idea about an assistant […] Actually helps you in the real world. ”
The AI industry is slowly moving towards the “Omni” models, if you want – models that can understand and synthesize different forms of media. Google’s new Gemi models can create audio as well as images and texts, while OpenAI’s default models in ChatGPT can make locally images-surely, including the Gibli-style industry. Amazon also announced plans to launch a “any-to-to-” model “later this year.
These Omni models require a lot of training data – images, videos, audio, text, etc. Hasabis indicated that video data for VEO is mostly coming from YouTube, it is a platform that owns Google.
“Basically, watching YouTube videos – lots of YouTube videos – [Veo 2] You know, physics in the world, “Hassabis said.
Google had previously called TechCrunch its models to “be trained in some” YouTube content “according to the deal with YouTube manufacturers. Report, Google has widen the terms of its service Last year to allow the company to tap the more data to training its AI models in the part of the last year.
