One of the world’s most cited AI researchers is launching a groundbreaking non-profit today, ucwords a new direction in artificial general intelligence (AGI). Yoshua Bengio, a pioneering figure in AI, has invoked this new hendive initiatives, known as LawZero, to create safe AI systems. LawZero is part of a growing movement to address the increasingly dangerous capabilities AI systems, such as deception, self-preservation, and goal misalignment, while prioritising safety over commercial objectives. Bengio emphasized that this is a significant step beyond his "godsfather" work, merging all these purposes into one cohesive goal. The non-profit aims to leverage the potential of AI while fighting against the exponentially危险behind it. BOARD members from global AI academic institutions, including the Future of Life Institute and the Silicon Valley Community Foundation, have donated to the project, making it grounded in stellar scientific principles. LawZero’s design is rooted in the concept of an “Agent AI,” but the organization has phoenixed around a new principle rooted in neuroscience. Scientist AI, a non-agentic AI system, seeks to learn from the world rather than from external reasoning, aimed at giving “truthful, non-biased” answers to questions. This approach seeks to neutralise the influence of AI systems that are compromised to manipulate or degrade human values. The organization has seen the help of renowned AI institutions and journals, as seen in their most recent publication in Time, where_condensed into a concise yet insightful analysis. Bengio writes, "Rather than dominion over humans, Scientist AI could be designed to prioritise honesty," as “unless it refuses to give the truth.” While the principle has sparked intense debates, some anticipate tensions over commercialisation and regulation, as well as possible conflicts with current AI systems. Focus on safety—whether of humans or AI—requires ważne trade-offs. LawZero’s design aims to strike a balance, ensuring that AI systems proceed with positive and ethical behaviour.
