Mark Zuckerberg, the chief executive of Meta, has announced an ambitious plan to develop artificial general intelligence (AGI), a type of artificial intelligence that can match human cognitive abilities. In a post on Meta’s Threads social network, Zuckerberg stated that the company’s AI research efforts will focus on building full general intelligence, which he believes is necessary for creating the best AI assistants, AIs for creators, AIs for businesses, and more.
Zuckerberg emphasized that achieving AGI will require significant advancements in various areas of AI, including reasoning, planning, coding, memory, and other cognitive abilities. To support this endeavor, Meta will invest heavily in high-performance computing hardware, including 350,000 top-end Nvidia GPUs, which are considered the most advanced AI processors available. These GPUs will provide processing power equivalent to 600,000 H100s, making Meta’s AI research efforts among the largest and most powerful in the industry.
While current AI technology has made tremendous strides in specific domains, such as natural language processing and image recognition, AGI remains an elusive goal. Today’s AI systems are narrow in scope and lack true understanding and intuition. They excel at recognizing patterns in vast amounts of data but struggle to reason and learn in novel situations. AGI, on the other hand, would possess a broad range of cognitive abilities, allowing it to adapt, learn, and reason similarly to humans.
Zuckerberg acknowledges that achieving AGI will take time and requires careful consideration of safety and ethical concerns. However, he believes that AGI is essential for shaping the future of computing and keeping pace with rapidly evolving technology. Moreover, he envisions wearable devices like Meta’s Ray-Ban smart glasses as an ideal interface for letting AI see what humans see and helping them navigate reality.
Meta’s AI research efforts will merge the company’s two existing AI research teams: the older-school FAIR effort and the newer generative AI team responsible for building Llama. Opinions regarding the viability of large language models (LLMs) as a means to achieve AGI vary widely. Some researchers view LLMs as an early form of AGI, while others criticize them as mere “stochastic parrots” that lack genuine understanding.
Regardless, Zuckerberg’s announcement marks a significant milestone in the development of AGI. If successful, Meta’s AGI technology could revolutionize countless industries and transform society in ways previously thought impossible. However, the path forward will undoubtedly present numerous challenges and require painstaking caution to ensure that AGI-empowered machines align with humanity’s interests.