‘Godfather of AI’ Warns It’s Unsafe—Proposes Scientist AI to Rein It In
Yoshua Bengio—renowned as one of AI’s “godfathers” and a 2018 Turing Award winner—now issues a stark warning: modern AI has become “unsafe”, urging urgent intervention. His answer? A new nonprofit, LawZero, is developing Scientist AI to serve as a transparency-enhancing and safety-improving model. 1. From Deep Learning Pioneer to Safety Advocate Bengio—alongside Yann LeCun and Geoffrey Hinton—initially shaped the foundations of deep learning. In recent months, he's shifted focus to the era’s perils, citing irresponsible AI use, such as instructions for bomb-making generated by AI, as a warning sign for broader misuse. 2. Introducing Scientist AI Scientist AI is being built on three core safety-by-design pillars: • Confidence Scores: The model self-assesses how sure it is, reducing “hallucinations.”• Explainable Reasoning: It can detail its reasoning process—restoring transparency lost in today's fast-moving AI models.• World Model Integration: Embedded understanding of real-world dynamics […]