Turing Award winner Joseph Sifakis calls for establishing global AI regulatory standards
Joseph Sifakis, winner of the 2007 Turing Award, called for establishing global standards and regulatory standards in the field of AI at the "2026 Sustainable Global Leaders Conference," and said this is a shared responsibility of humanity. Citing ChatGPT and DeepSeek scoring higher than humans on medical exams as an example, he pointed out that large models do not understand social norms and value systems and therefore cannot replace doctors. He also warned that large companies may use regulatory gaps to shirk responsibility and seek commercial gain.
Body · AI translation
2026-10-20 16:31 Sina Tech - -
The "2026 Sustainable Global Leaders Conference" was held in Shanghai from October 20 to 22. Joseph Sifakis, winner of the 2007 Turing Award and founder of the Verimag laboratory, attended and delivered a speech.
Joseph Sifakis said that as superintelligence has formed a kind of deepening narrative, people have begun to worry that these systems may exceed the scope of authorization and instead pursue their own survival and dominance. Therefore, humanity needs to understand the meaning and role of AI's self-evolution of capabilities, as well as the risks and benefits it brings.
Joseph Sifakis pointed out that "autonomy" refers to taking purposeful action, which is a characteristic of living beings. A single-agent intelligence must understand the world in order to better realize intelligence, which means it needs to understand a world model and obtain perception from that world model.
Joseph Sifakis believes that at present, agents and humans do not think about problems in the same way. He gave an example: even if ChatGPT or DeepSeek perform better than humans in medical examinations, this does not mean we can let large models serve as doctors.
His explanation was that human doctors understand the norms of human society and the relevant value systems, and know well that once they violate certain norms, they will damage their reputation and may even have their freedom restricted. But obviously, large models cannot do this.
Joseph Sifakis emphasized that attention must be paid to the moral deficiencies of AI systems, and that AI's self-evolution does not always develop in a good direction. "If agents exceed our control and go beyond the initial preset, self-evolution may deviate far from reality, leading to uncontrollable risks or systems being intentionally manipulated. We have already had such cases, so the urgent task is not to prepare to deal with this kind of self-evolution, but to govern what we already have."
Joseph Sifakis also pointed out that the real danger lies in the possibility that superintelligence may conceal the real risks brought by insecurity and uncontrollability, while some large companies, for commercial interests, use this to shirk responsibility and at the same time exploit regulatory gaps for profit.
He called for the establishment of global standards and regulatory standards in related fields. "We must join hands and conduct regulation together. On the road ahead, we need to reshape the balance of power, work with like-minded countries to regulate AI, and coordinate the relationship between technological development and feasibility."
"AI is only a small problem; it can make the world better or destroy it. This is a human-led transformation, and we must maximize the benefits and minimize the risks. This is our shared responsibility," said Joseph Sifakis.
Sina statement: All meeting transcripts are compiled from on-site stenography and have not been reviewed by the speakers. Sina publishes this article for the purpose of conveying more information and does not mean that it agrees with the views or confirms the descriptions.