AGI (artificial general intelligence) is a form of AI capable of understanding, learning, and applying intelligence across a wide range of tasks, akin to (or exceeding) human capabilities. It is the Grail Quest for all AI researchers. AGI could be achieved this afternoon, in a few weeks, next year, or never, but AI researchers are undeterred by the fact that AGI may not be achievable – it is “the” goal. The North Star. The apex, the zenith, the pinnacle, acme, aspiration, ambition… the final frontier… you get it.
In an interview with Alex Heath for The Verge, Meta CEO Mark Zuckerberg outlined his company’s ambitious goal to develop AGI and its strategic reorganization to integrate AI research and product development. Zuckerberg, without a definitive timeline or clear definition for AGI, emphasized the need for general intelligence to build desired products, attracting top researchers to work on challenging problems.
Meta plans to significantly invest in computing power, essential for training large AI models. By the end of 2024, Meta aims to possess nearly 600,000 GPUs, including more than 340,000 Nvidia H100 GPUs, positioning it potentially as the largest individual company in terms of AI computing capacity.
Mark also acknowledged the lack of a precise definition or understanding of when AGI will be achieved. He views its development as a gradual process rather than a singular event. This perspective influenced Meta’s approach to developing its large language model, Llama 2, and the upcoming Llama 3, which will include code-generating capabilities and advanced reasoning. Importantly, Near-AGI Technology Is Already Transforming the World.
Regarding the control of AGI, Zuckerberg contrasted Meta’s open-source approach with other companies’ increasingly closed methods. He argued for open-source to prevent concentration of power and ensure equal access, though he admitted the potential for power concentration in different forms under Meta’s approach.
Interestingly, he refuted the notion that Meta’s focus on AGI is a pivot from its metaverse initiative. He reaffirmed Meta’s commitment to the metaverse, with more than $15 billion in annual spending, and envisioned AI’s critical role in Meta’s hardware and social media platforms. He anticipates AI-generated virtual worlds and AI characters in Meta’s social apps.
Zuck’s interview revealed Meta’s strategic direction toward AGI, highlighting the challenges of talent acquisition, the necessity of substantial computing resources, and the philosophical and practical considerations of open versus closed AI development models. But – as you know – Meta is one of a handful of companies that has the talent and the resources to travel the path toward AGI. If they get there, what will they find?
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.