May 21, 2022

The Indie Toaster

Complete News World

Meta Zuckerberg unveils AI projects aimed at building the future of Metaverse

Meta Zuckerberg unveils AI projects aimed at building the future of Metaverse

Feb 23 (Reuters) – Facebook owner dead (FB.O) He’s working on artificial intelligence research to generate worlds through speech, improving the way we chat with voice assistants and translate between languages, CEO Mark Zuckerberg said Wednesday, while he maps out key steps to building the metaverse.

Zuckerberg is betting that the metaverse, a futuristic idea of ​​virtual environments where users can work, socialize and play, will be the successor to the mobile internet.

“The key to unlocking a lot of these advances is artificial intelligence,” he said, speaking at a livestreamed “Inside the Lab” event.

Register now to get free unlimited access to Reuters.com

Zuckerberg said Meta was working on a new class of generative AI models that would allow people to describe the world and create aspects of it. In a pre-recorded demo, Zuckerberg demonstrated an AI concept called Builder Bot, appearing as a legless 3D avatar on an island and giving speech commands to create a beach and then adding clouds, trees, and even a picnic blanket.

“As we advance in this technology even further, you will be able to create micro-worlds to explore and share experiences with others, just with your voice,” Zuckerberg said. He did not specify a timeline for these developments or provide more details about how the Builder Bot will work.

He said Meta has been working on AI research to allow people to have more natural conversations with voice assistants, a step toward how people communicate with AI in the metaverse. He said the company’s CAIRaoke project was a “complete neural model for building assistants on a device.”

See also  Dow rises, but Netflix crashes 35% on profit; Tesla Earnings Due

A demo of Project CAIRaoke technology showed one family using it to help cook soup, with the voice assistant ringing to warn that salt was already added to the pot. The assistant also noticed that the salt was running out and asked for more.

Meta said it was using the model inside a video call gateway device and intended to integrate it into devices with augmented reality (AR) and virtual reality (VR). In an interview with Reuters, Meta Vice President of Artificial Intelligence Jerome Pesenti said he is severely restricting responses from his new assistant at CAIRaoke so he can make sure the system doesn’t generate abusive language.

“These language models are very powerful … so we put a lot of effort into being able to master them,” said Bisenti.

Zuckerberg also announced that Meta was working on a universal speech translator, with the goal of providing instant speech translation across all languages. The company previously set a goal for its AI system to translate all written languages.

Miniature toy figures in front of the Meta show the new Facebook logo displayed in this illustration taken on October 28, 2021. REUTERS/Dado Ruvic/Illustration/File Photo

The social media company, which recently lost a third of its market value after a poor earnings report, has invested heavily in its new focus on building the metaverse and changed its name to reflect that ambition. Meta this month announced a 2021 net loss of $10.2 billion from Reality Labs, the company’s augmented and virtual reality company. Read more

different monster

Meta’s director of AI told Reuters that Meta is exploring how AI can be used to modify content and activity in the metaverse.

Pesenti, who said this was an “evolving work” and that Meta was still learning about metaverse activity policies.

See also  Dow futures point to 'rapid' market losses as Russian banks targeted amid invasion of Ukraine

At the AI ​​event, Zuckerberg said Meta was preparing for how the AI ​​would interpret and predict the kinds of interactions that might occur in the metaverse, by working on “self-supervised learning” — where the AI ​​is given raw data rather than trained on a lot of data. classified.

Zuckerberg said Meta has also been working on self-centered data, which includes seeing worlds from a first-person perspective. He said it has brought together a global coalition of 13 universities and labs to work on the largest ever self-centered dataset, called Ego4D.

Meta also said it would expand free education initiatives aimed at involving more ethnic minorities in technology, which researchers say is needed to create bias-free artificial intelligence systems. About 80% of data analytics and AI executives are men and 65% are white, according to a recent survey across the US and Europe by recruiter Heidrick & Struggles.

In a sign of transparency, Meta plans to create an open source TorchRec recommendations library that is used to customize products like Facebook’s News Feed, Pesenti said at another event session. He said the company will also publish a prototype feed ranking to show how its algorithms prioritize the content you show users on Instagram.

Some of the projects Meta announced on Wednesday, such as Project CAIRaoke and its algorithmic transparency efforts, follow similar innovations announced in recent years by competitors like Alphabet Inc’s (GOOGL.O) Google.

Meta also recently announced that its research team has built a new artificial intelligence supercomputer that it believes will be the world’s fastest when completed in mid-2022. Read more

(Elizabeth Culliford reports in New York). Additional reporting by Dave Parrish in Oakland, California and Katie Ball in Palo Alto. Editing by Lisa Schumaker, Bernard Orr

Our criteria: Thomson Reuters Trust Principles.