Photo: Dima Solomin
CEO and founder Mark Zuckerberg recently announced that Facebook parent Meta is developing artificial intelligence research to create worlds through speech, upgrade the process of chatting to voice assistants, and translate between languages.
Zuckerberg is outlining fundamental moves to developing the metaverse. He believes that the metaverse will be the future of the mobile internet. The metaverse is a realm of virtual and digital reality where users can interact, work, and play together.
“The key to unlocking a lot of these advances in AI,” said Zuckerberg, giving a speech at the giant’s live-streamed “Inside the Lab” event.
According to Zuckerberg, Meta is developing a new variation of the generative AI model that will enable people to illustrate a world and set out facets of it.
Shown in a pre-recorded demo, Zuckerberg presented an AI concept named Builder Bot, where he showed up as a 3D avatar with no legs on an island and gave speech commands to make a beach and then create other clouds, trees, and even a picnic blanket.
“As we advance this technology further, you’ll be able to create nuanced worlds to explore and share experiences with others, with just your voice,” he said. Zuckerberg, however, did not specify a timeline for these improvements or disclose additional details on how Builder Bot works.
He continued that Meta was developing AI research to enable people to possess additional natural conversations with voice assistants, moving towards how individuals will communicate with AI in the metaverse.
He added that the giant’s Project CAIRaoke was “a fully end-to-end neural model for building on-device assistants.”
The Project CAIRaoke tech presentation exhibited a family utilizing it to aid in cooking a stew, with the voice assistant warning the family and saying that they had already put salt in the pot. The assistant also regarded that they were running low on salt and ordered more.
Meta stated that it utilized the model within its video-calling Portal device and targeted coalescing it into virtual reality (VR) and augmented reality (AR). It was firmly limiting the responses of its new CAIRaoke-based model until it could guarantee that the system did not produce offensive language, said Jerome Pesenti, vice president for AI at Meta, in an interview.
“These language models are very powerful … so we are making a lot of effort to be able to control them,” Pesenti said.
Opinions expressed by CEO Weekly contributors are their own.