Mark Zuckerberg is building Meta’s artificial intelligence voice capabilities this year as the social media giant pushes forward with its plans to generate revenue from the rapidly developing technology.
Meta plans to introduce improved voice features in its latest open-source large language model, LLaMA 4, expected in the coming weeks, according to people familiar with the matter. This move signifies the company’s bet that future so-called AI-powered agents will be conversational, rather than text-led.
One individual mentioned that the company is specifically focusing on making the interaction between a user and their voice model more like a two-way natural conversation, allowing users to interrupt, rather than sticking to rigid question-and-answer formats.
The voice push comes at a time when CEO Zuckerberg has outlined bold plans to position the $1.7 trillion Silicon Valley company as an “AI leader,” designating 2025 as a critical year for several of its AI products. Meta is in a race with rivals like OpenAI, Microsoft, and Google to commercialize the technology.
This has also prompted the company to test premium subscriptions for its AI assistant, Meta AI, for agent-like tasks such as reservation booking and video creation, according to two people familiar with the matter. It is also considering introducing paid ads or sponsored posts in its AI assistant’s search results, one of the people said.
/Read More/
The real AI upgrade for Siri might still be years away
Discover the latest update on Siri’s AI upgrade, as Apple plans a conversational version by 2027. Learn about expected changes, iOS 18.5 updates, and Apple’s AI challenges amid competition. Stay informed on the future of Siri and the latest developments in AI technology at Apple.
Zuckerberg revealed plans earlier this year to create an AI engineering agent with coding and problem-solving capabilities similar to those of a mid-level engineer. This would involve converting voice to text, sending it to an LLM (large language model), receiving text, and converting it back to speech. Speaking at the Morgan Stanley Technology, Media, and Telecom conference, he said, “I think this is a big deal for the product interface, the idea that you can talk to the internet and just ask it anything. I think we’re still thinking about how powerful this is.”
Meta is also discussing what should and should not be included in the latest LLaMA model, according to two people familiar with the matter. This conversation is happening amidst a flurry of launches by competitors and warnings from Silicon Valley venture capitalist and newly appointed ‘AI czar’ David Sacks, who has expressed concerns about ensuring that U.S. AI models do not become politically biased or “woke.”
OpenAI launched its voice mode last year, focusing on giving it a distinct personality, while Elon Musk’s xAI introduced Grok 3 with voice features for select users on the X platform at the end of last month.
According to the company, Grok’s model was specifically designed with fewer safety measures, including an “unhinged mode” that intentionally responds in an “offensive, inappropriate, and aggressive” manner. Meta unveiled a less “sanitized” version of its AI models for its third LLaMA iteration last year after criticism that LLaMA 2 was refusing to answer innocuous questions.
Allowing users to interact with the AI assistant via voice commands is a key feature of Meta’s Ray-Ban smart glasses, which have recently become a major hit among consumers. The company has accelerated its plans to create lightweight headsets that could replace smartphones as
consumers’ primary computing device.
The social media giant has introduced enhanced features as it bets on technology that promotes the development of so-called AI agents. Mark Zuckerberg is building Meta’s artificial intelligence voice capabilities this year as the social media giant pushes forward with its plans to generate revenue from the rapidly developing technology.
Meta plans to introduce improved voice features in its latest open-source large language model, LLaMA 4, expected in the coming weeks, according to people familiar with the matter. This move signifies the company’s bet that future so-called AI-powered agents will be conversational, rather than text-led.
One individual mentioned that the company is specifically focusing on making the interaction between a user and their voice model more like a two-way natural conversation, allowing users to interrupt, rather than sticking to rigid question-and-answer formats.
The voice push comes at a time when CEO Zuckerberg has outlined bold plans to position the $1.7 trillion Silicon Valley company as an “AI leader,” designating 2025 as a critical year for several of its AI products. Meta is in a race with rivals like OpenAI, Microsoft, and Google to commercialize the technology.
This has also prompted the company to test premium subscriptions for its AI assistant, Meta AI, for agent-like tasks such as reservation booking and video creation, according to two people familiar with the matter. It is also considering introducing paid ads or sponsored posts in its AI assistant’s search results, one of the people said.
Zuckerberg revealed plans earlier this year to create an AI engineering agent with coding and problem-solving capabilities similar to those of a mid-level engineer. This would involve converting voice to text, sending it to an LLM (large language model), receiving text, and converting it back to speech. Speaking at the Morgan Stanley Technology, Media, and Telecom conference, he said, “I think this is a big deal for the product interface, the idea that you can talk to the internet and just ask it anything. I think we’re still thinking about how powerful this is.”
