At the World Economic Forum in Davos, AI pioneer Yoshua Bengio stressed the risks posed by autonomous AI agents, indicating they could trigger catastrophic scenarios if artificial general intelligence (AGI) is developed under such frameworks. He emphasized the importance of designing non-agentic AI systems, which could still contribute significantly to fields like science and medicine without the inherent dangers of independence. As AI technology matures and businesses see its potential for return on investment, the focus should shift towards safer implementations of AI that prioritize human oversight and safety.
"All of the catastrophic scenarios with AGI or superintelligence happen if we have agents," Bengio told BI in an interview.
"All of the AI for science and medicine, all the things people care about, is not agentic," Bengio said. "And we can continue building more powerful systems that are non-agentic."
Collection
[
|
...
]