In a surprise Instagram Reel today, Meta CEO Mark Zuckerberg made an unexpected announcement. He revealed that the company is actively working on the development of open source artificial general intelligence (AGI). This ambitious project involves bringing together two of Meta’s AI research teams – FAIR and GenAI – with the shared objective of building full general intelligence and making it accessible to the broader community through open sourcing.
“Our long-term vision is to build general intelligence, open source it responsibly, and make it widely available so everyone can benefit,” wrote Zuckerberg in a caption accompanying the Instagram video.
Zuckerberg emphasized the importance of advancing AI capabilities in various domains, stating that the next generation of services would require the development of full general intelligence. This entails enhancing AI assistants, creating AIs for creators and businesses, and making advancements in areas such as reasoning, planning, coding, memory, and other cognitive abilities.
Zuckerberg also highlighted Meta’s ongoing efforts in AI training, specifically mentioning their work on training Llama 3. He noted the significant computational infrastructure being built, which includes plans to have 350,000 Nvidia H100s by the end of the year.
Additionally, Zuckerberg discussed the potential integration of AI with Meta’s metaverse and their Ray-Ban smart glasses. He expressed his belief that as AI becomes more prevalent in people’s lives, there will be a need for new devices that seamlessly merge with the AI experience. According to Zuckerberg, smart glasses are an ideal form factor for enabling AI to see and hear what the wearer does, making assistance always available.
It is worth noting that Zuckerberg’s announcement coincides with OpenAI CEO Sam Altman’s recent remarks about AGI at the World Economic Forum. Altman has softened his stance on the existential risks associated with AGI since his reinstatement in November 2023, generating significant interest and discussions surrounding this topic.
Despite Meta’s commitment to AGI development, their chief scientist, Yann LeCun, has expressed skepticism about the immediate arrival of AGI, particularly within the next five years. This divergence in opinions among key figures within the AI community adds an intriguing dimension to the ongoing conversations surrounding AGI.
The revelation that Meta’s future AGI could be open sourced follows closely on the heels of VentureBeat’s declaration that Llama and open source AI have dominated the AI landscape in 2023. Meta’s decision to potentially open source their AGI will undoubtedly fuel further debates regarding the benefits and drawbacks of open source versus closed source AI solutions. These discussions gain even more prominence following Anthropic’s recent publication, which warned about the potential risks associated with open models and the presence of hidden malicious elements.
