Hugging Face Co-Founder Takes on AI Optimists: A Bold Challenge

Thomas Wolf, co-founder of Hugging Face, expressed skepticism about the potential of current AI systems, particularly large language models (LLMs), to drive significant scientific breakthroughs. Speaking to Fortune, he emphasized that while LLMs are adept at providing answers, they struggle to formulate original questions, which he sees as critical in scientific inquiry. Responding to a blog post from Anthropic’s CEO, who claimed AI could rapidly advance scientific discovery, Wolf noted that the ability to challenge existing paradigms and ask innovative questions is vital for genuine progress. He criticized LLMs for being “yes-men” that predict likely outcomes without generating original ideas. Furthermore, he argued that many current AI benchmarks focus on answering established questions rather than fostering new inquiries. Wolf concluded that achieving true scientific advancements depends on developing AI capable of formulating disruptive questions, likening it to creating a model that can invent new rules, rather than just excel at existing ones.

Source link