Mark Zuckerberg Showed New Tech That Uses AI to Fully Recreate People

A person with curly brown hair, light skin, and blue eyes is seen in a close-up portrait against a black background. They are wearing a dark-colored shirt and a subtle smile is visible on their face.

Today at Meta Connect, the parent company of Facebook, Instagram, and Threads showed off a technology that is capable of fully recreating influencers as AI figures that can be spoken to as if they were the real person.

The technology at play here, as explained by The Verge and Engadget, was showcased for two purposes. Firstly, Meta CEO Mark Zuckerberg showed that its AI is able to dub Reels into other languages including changing the movement of a subject’s lips so that it appears to be spoken natively. Not only will the words be translated, but Zuckerberg says the technology will “simulate the speaker’s voice in another language,” too.

No exact timeline for this technology’s release was provided at Meta Connect, but Zuckerberg says that it will first roll out to “some creators’ videos” in English and Spanish in the United States and Latin America before slowly increasing language support and access over time — the company did say that it plans to add more languages “soon.”

Meta’s AI translation feature does sound like a good improvement from strictly an accessibility and reach perspective even if there is likely to be some discomfort around the changing of how a person’s lips move during speech. That said, if that small change is considered problematic, Meta’s next application of the technology is likely to be viewed with significantly more scrutiny.

As shown in Instagram Reels shared by The Verge, Zuckerberg presented a live demonstration of speaking with a fully AI-powered version of a real human influencer. The AI version looked and talked like the real person and was able to — or at least tried to — respond to questions as it believed the creator would. It was apparently very believable in the demonstration.

Meta did not apparently say when it plans to give access to this feature to content creators and influencers, but the company is likely to be very careful with the rollout. It is very possible for this technology to be abused and for many to not be able to tell if the words being spoken were actually something the real person said or if it was just an AI-generated answer.


Image credits: Meta

Discussion