- Mark Zuckerberg spoke Wednesday about "challenges" Facebook was working on to build the metaverse.
- Hiring and the further development AI were areas Zuckerberg touched on.
- "AI needs to understand context and learn like humans do," he said.
Facebook is banking on far-away advances in artificial intelligence and machine learning, and a lot of new hiring, to build its version of the metaverse.
During an event for the company, now known as Meta, on Wednesday CEO Mark Zuckerberg announced Project Caraoke, an effort to get AI to a "human level" of intelligence where it's capable of learning, predicting, and acting on its own.
"In the metaverse, we'll need AI to help people navigate between the virtual world and the physical world, and because these virtual worlds will be changing constantly, AI needs to understand context and learn like humans do," Zuckerberg said.
Facebook is building AI that can learn on its own
The project, pronounced like the sing-along activity, is almost exactly like Google's DeepMind machine-learning subsidiary. To advance its AI capabilities, Facebook is investing heavily in machine-learning technology. The company said it had started to implement self-supervised learning, a tool that allows AI models to recognize patterns on its own.
In a typical AI learning model — including those used by Facebook — algorithms learn from labels and tags inputted by humans, like Facebook content moderators. The AI proceeds to make decisions based on those labels. So far, Facebook's tests of SSL are outperforming previous AI models for text and video, Zuckerberg said.
"SSL is still developing, but we think it will be an important tool for the metaverse," he added. "Information in the metaverse will be too great to be captured by labeled data sets."
For Facebook's metaverse to be what Zuckerberg has said it can become, an immersive virtual world full of fun and useful applications that will be the next version of the mobile internet, it will need technology that's less cumbersome to deploy and operate. Artificial intelligence that can learn on its own is a step in that direction.
In a separate segment, Joelle Pineau, a managing director for the company's AI-research group, spoke of the company's work in robotics, including machines equipped with humanlike skin. Pineau also said the company was using AI to test robots in real-world settings like homes, ostensibly to allow them to learn as humans do.
"Learning is much faster, deeper, and transferable once it's done in the real world through trial and error, rinse and repeat," Pineau said.
Project Caraoke may improve Facebook's language capabilities
Caraoke is also a part of the company's plans to expand its language capabilities — what Zuckerberg called "a foundational problem" for the metaverse.
Two projects were announced during the event directed at helping Facebook improve its language capabilities: No Language Left Behind, an AI-based translating system for the internet that is expected to learn every language without using English as a translation base, and Then Universal Speech Translator, a tool to translate information and conversation into any language when in an augmented-reality or virtual-reality experience. Neither have been deployed.
All the executives who spoke during the event acknowledged that the level at which this AI technology needed to be at for any of Facebook's metaverse ambitions to become reality — be it in the virtual or physical world — were at least several years away.
Meta plans on hiring more AI researchers to build the metaverse
As the event ended, Zuckerberg and other executives openly called for people to apply to work for Meta in AI and other departments working on the metaverse. The company has lately struggled to retain and recruit talent given its many issues with harmful content, political misinformation, and other social ills.
But even as Meta opens up positions for AI scientists, self-learning algorithms still have a long way to go before they become useful to the company.
As Yoshua Bengio, an AI expert, put it succinctly during the conference: "We are still far away from human-level AI."