Meta Is Banking On Its AI Reaching “Human Levels”


According to Mark Zuckerberg, who hosted an event on Wednesday in which he announced Project Caraoke, the codename for developing the company’s AI technology so that it is 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 the context and learn as humans do,” Zuckerberg said in the meeting.

In order to do so, Meta is investing heavily in machine-learning technology and has recently started to implement self-supervised learning tools which allow AI models to recognize patterns on their own and learn from them.

A typical AI learning model algorithms work by learning from labels and tags that have been previously inputted by humans, before making decisions based on those labels.

The metaverse will require technology that is easy to implement in order to function properly, which means that Meta will have to focus heavily on machine learning.

Meta’s managing director for the company’s AI research group Joelle Pineau divulged the company’s work in its robotics division in the meeting.

Pineau said that the social media giant was working on robotics equipped with humanlike skin and said that it was using AI to rest robots in real-world settings, including homes.

“Learning is much faster, deeper, and transferable once it’s done in the real world through trial and error, rinse and repeat,” Pineau told Business Insider.

During the event, Zuckerberg openly called for more people to apply to work for Meta in its AI and metaverse divisions after the company has endured struggles with retaining its talent due to the company’s bad press surrounding harmful content, political misinformation, and other social ills.