In a recent update, Mark Zuckerberg shed light on Facebook’s AI endeavors, emphasizing a long-term vision to develop general intelligence, responsibly open-source it, and ensure widespread accessibility for everyone to benefit. This commitment involves the integration of two major AI research initiatives, FAIR and GenAI, to propel their shared goals.
A focal point of Facebook’s AI efforts is the ongoing training of their next-gen model, Llama 3. Zuckerberg highlighted the establishment of a robust compute infrastructure to support their future roadmap, with a staggering commitment to deploy 350,000 H100s by the end of the year. The cumulative compute power, including other GPUs, is anticipated to reach nearly 600,000 H100s equivalents.
Excitement surrounds Facebook’s foray into creating AI-centric computing devices, exemplified by the Ray-Ban Meta smart glasses. Zuckerberg expressed anticipation for further advancements in this domain, promising a stream of forthcoming updates. As Facebook pushes boundaries in AI research, the tech community eagerly awaits the realization of their ambitious vision.