Zoom CEO imagines AI deepfakes attending meetings instead

Zoom CEO Eric Yuan has a vision for the future of work: sending your AI-powered digital twin to attend meetings on your behalf. In an interview with The Verge’s Nilay Patel published on Monday, Yuan shared his plans for Zoom to become an “AI-first company,” using AI to automate tasks and reduce the need for human involvement at work. diary.

“Let’s say the team is waiting for the CEO to make a decision or maybe some meaningful conversation, my digital twin can actually represent me and can also be part of the decision-making process,” Yuan said in the interview. “We’re not there yet, but that’s one of the reasons there are limitations in current LLMs.”

LLMs are large language models: AI models that predict text and power AI assistants like ChatGPT and Microsoft Copilot. They can generate very convincing, human-like texts based on probabilities, but they are far from being able to replicate human reasoning. Still, Yuan suggests that instead of relying on a generic LLM to impersonate you, in the future, people will train customized LLMs to simulate each person.

“Everyone shares the same LLM [right now]. It does not make any sense. I should have my own LLM: Eric’s LLM, Nilay’s LLM. “All of us will have our own LLM,” he told The Verge. “Essentially, that’s the basis of the digital twin. Then I can count on my digital twin. Sometimes I want to join, so I join. If I don’t want to join, I can send a digital twin to join. That’s the future.”

Yuan believes we are five or six years away from this kind of future, but even the suggestion of using LLM to make decisions on someone’s behalf is enough to frustrate and confuse some AI experts.

“I’m not a fan of this idea of ​​people building LLM systems that try to simulate individuals,” AI researcher Simon Willison recently wrote in X, regardless of Yuan’s news. “The idea that an LLM can usefully predict an individual’s response seems obviously flawed to me. It is equivalent to receiving business advice from a talented impersonator/improv artist: just because they can ‘sound like’ someone doesn’t mean they can provide “genuinely useful information.”

In the interview, Patel rejected Yuan’s claims, saying that LLMs hallucinate, drawing inaccurate conclusions, so they are not a stable basis for the vision Yuan describes. Yuan said that he is confident that the hallucination problem will be solved in the future, and when Patel also rejected that point, Yuan said that his vision would advance later.

“In that context, that is the reason why, today, I cannot send myself a digital version during this call,” Yuan told Patel. “I think that’s more of the future. The technology is ready. Maybe it needs some change in the architecture, maybe a transformer 2.0, maybe the new algorithm to have that. Again, it’s very similar to 1995, 1996 , when the Internet was born “There are many limitations. I can use my phone. It is very slow. Basically it doesn’t work. But look at it today. “I think those problems will be solved.”

Patel also mentioned the privacy and security implications of creating a convincing replica of yourself that others could hack. Yuan said the solution was to make sure the conversation was “very secure,” pointing to a recent initiative by Zoom to improve end-to-end encryption (an issue that, we should note, the company has lied about in the past ). And he says Zoom is working on ways to detect deepfakes and create them, in the form of digital twins.

Leave a Comment