The Man Behind AlphaGo Thinks AI Is Taking the Wrong Path (www.wired.com)

David Silver, a key figure behind the development of AlphaGo, has launched a new company called Ineffable Intelligence, focusing on building more general forms of AI superintelligence through reinforcement learning. Unlike the majority of AI companies that rely on large language models (LLMs), Silver argues that this approach is fundamentally flawed since LLMs learn from human-generated data rather than developing independent understanding. His vision is to create "superlearners" that can continually evolve and solve complex real-world problems, drawing parallels to scientific breakthroughs like Darwin’s theory of evolution. Ineffable Intelligence has already attracted significant attention, raising $1.1 billion in seed funding with a valuation of $5.1 billion, and recruiting top AI talent from leading research labs. Silver’s methodology involves placing AI agents in simulations, allowing them to learn and collaborate free from human biases, potentially addressing concerns about safety and alignment with human values. As the AI field increasingly races toward superintelligence, Silver seeks to establish a pathway that fosters creativity and discovery, arguing this could lead to unprecedented advancements in science and technology. His commitment to using profits for charity underscores a responsible approach to developing such powerful technologies.