Meta's chief AI officer says health will define the next generation of AI models

Meta has identified consumer health as the central focus of its next generation of artificial intelligence models. Speaking at the Bloomberg Tech conference in San Francisco on Thursday, Meta's Chief AI Officer Alexandr Wang said health capabilities would serve as the primary differentiator for its AI products and the feature most likely to reach billions of users across the company's platforms.

Wang's role and background

Wang, 29, joined Meta's board after Mark Zuckerberg invested $14 billion into his data-labelling company, Scale AI. He now leads Meta Superintelligence Labs, Zuckerberg's broader AI revamp initiative. The division released its first product, Muse Spark, last April.

Honest assessment of where Muse Spark stands

When addressing Muse Spark's current position in the market, Wang was candid. He acknowledged that the model had not yet reached the level of Anthropic's Claude or OpenAI's ChatGPT. That said, he noted it had surpassed internal expectations and expressed confidence that forthcoming versions would close the gap considerably with the market leaders.

Health singled out as Meta's strongest suit

Of all the capabilities Muse Spark offers, Wang highlighted health as its standout area and the domain through which Meta intends to build its competitive edge. "Health is an area that we view as really critical as we scale these models out to billions," he said.

The broader trend of users turning to AI tools such as ChatGPT and Claude for fitness advice, dietary guidance, health diagnoses, and mental health support has drawn growing attention from both businesses and regulators, who have raised concerns around patient safety and data privacy.

Bio-risk concerns shaped a key decision

Wang disclosed that Muse Spark had encountered significant bio-risk concerns during its development. He declined to elaborate on the specifics, but confirmed that Meta addressed those issues before the model was made available to the public. Those same risks, however, led directly to a consequential decision: Muse Spark would not be released as an open-source model.