Meta’s Superintelligence team, which is exploring the advanced use of artificial intelligence, has released Muse Spark, the first in its initial series of large language models. The company said these models are “rooted in the relationships and context already at the center of your life.”
Meta said Muse Spark represents a reimagining of its entire AI development stack. And it’s generating strong results. Meta noted that even in the early stages, Muse Spark has already achieved industry-leading results across most of the critical AI benchmark tests.
As explained by Meta: “Over the last nine months, Meta Superintelligence Labs rebuilt our AI stack from the ground up, moving faster than any development cycle we have run before. Muse Spark is the first model in our new Muse series — a deliberate and scientific approach to model scaling where each generation validates and builds on the last before we go bigger.”
It’s the company’s first big effort to lead the AI race, after committing hundreds of billions of investment dollars. And within its initial release notes, Meta has also provided insight into how the company envisions AI being integrated into people’s everyday lives.
Muse Spark includes advanced multimodal capabilities, which enable the system to better understand visual cues, and assess items as humans do.
The system will be better able to provide context-based advice and insight, and therefore, more in-depth reasoning, Meta said. That could help to improve the complexity of its responses and reduce the need for follow-up queries.
Though this note seems concerning: “Health is one of the top reasons people turn to AI, so we worked with a team of physicians to develop the model’s ability to provide helpful information on common health questions and concerns.”
It’s worth noting that AI tools should not be trusted for health advice. Not even advanced models that have had hundreds of billions of dollars invested into them.
Muse Spark also has a new shopping mode, which could help users decide what to wear, how to style a room or what gifts to buy.
Though why these AI developers keep looking to automate tasks that people actually enjoy doing remains a mystery.
Fashion sense and creativity are uniquely human traits, so why are AI teams so keen to automate them out of existence? Buying a gift is, at least ostensibly, intended to reflect affection and connection — so why would people want to pass this task off to a robot?
It’s possible that Meta’s AI team, and other AI developers, are highlighting these benefits without realizing they’re more reflective of a developers’ optimization mindset than market demand.
For example, Meta Chief Mark Zuckerberg wore the same clothes every day for years in order to streamline his decision-making process. Maybe that says something about the thinking here.
Maybe these aren’t the people who should be developing new structures for how we live.
The release of Muse Spark offers an intriguing view into Meta’s future vision, which, in Zuckerberg’s mind at least, will eventually lead to expanded AI use.
Muse Spark is now powering Meta’s AI assistant in the Meta AI app, with Meta planning to make it a foundational model for future AI projects.
