Turing Award winners and academicians clash at Shanghai Dishui Lake: the security, computing power, and governance of AI's next decade

Add crowdsourcing testing promotional copy Add crowdsourcing testing promotional copy end Add App-exclusive traffic diversion

As technology extends into the physical world at a pace beyond imagination, a sense of crisis is becoming prominent in the fields of security, computing power and international governance.

On October 30, the 2026 World Laureates Forum opened in Lingang, Shanghai. At the AI parallel forum, an in-depth dialogue unfolded around "The Next Decade: The Development and Governance of Artificial Intelligence," bringing together multiple Turing Award laureates, top academicians and industry leaders. Discussions on computing power, chips and software repeatedly clashed with views on security governance.

Martin Hellman, professor emeritus at Stanford University and Turing Award laureate, pointed out that AI risks, the threat of nuclear war and climate change all ultimately share a common problem: technology has endowed humanity with almost "god-like physical power," yet humanity still behaves like an irresponsible teenager.

Hellman likened this imbalance to a 16-year-old who has just obtained a driver's license and somehow gotten hold of a high-performance sports car; he must mature quickly, and if he does not, he will bring about his own destruction.

In Hellman's view, both military and economic motives are driving the accelerated development of AI, and a confrontational international environment makes it even harder to jointly address risks. He called for building a cooperative world order, arguing that unilateral finger-pointing by any country will only backfire. Without cross-border regulatory mechanisms and safety guardrails, humanity faces the crisis of mutual destruction, "but as long as we build the world order we need, all these challenges will eventually be resolved."

China's top-level AI design philosophy also responds to this concern. He Jifeng, an academician of the Chinese Academy of Sciences and a lifetime professor at East China Normal University, proposed the concept of "trustworthy artificial intelligence" as early as 2017. He stressed that AI should never be an unpredictable black box. "Algorithmic bias, data hazards, hallucination problems and the global intelligence divide all remind us that AI must not only pursue the ultimate in performance, but also hold the bottom line of being trustworthy, safe and for the good," He Jifeng said.

Beyond security, computing power is also a key focus determining the next stage of AI development. At present, large models are driving continued growth in computing power demand and prompting the industry to increase equipment and investment. But Jack Dongarra, Turing Award laureate and professor emeritus at the University of Tennessee, warned that the peak performance of hardware does not equal actual application efficiency.

Data movement has already become a cost that cannot be ignored. For example, communication among the various parts of a computing system and memory bandwidth issues are equally important; floating-point operation speed alone is not enough to measure the performance of complex computing tasks.

Chen Yunji, deputy director of the Institute of Computing Technology of the Chinese Academy of Sciences and director of the State Key Laboratory of Processor Chips, likewise showed that the two-way relationship between chips and AI is being redefined: on the one hand, designing more efficient processors for AI; on the other hand, letting AI participate in processor design.

In Chen Yunji's view, if AI can design better AI chips and chips in turn support stronger AI, an RSI (recursive self-improvement) loop may form, and the pace of technological advancement will far exceed human imagination. This is his judgment on the important breakthrough direction for the next decade.

Similar exploration has also appeared in the field of industrial software. Kong Linghe, a professor at Shanghai Jiao Tong University, introduced that the team carried out software re-engineering based on the MiniMax large model: first understanding system semantics, then generatively reconstructing code, and finally verifying the results through multi-agent cross-comparison, differential testing and other methods. At the forum, the team announced that it would open-source about 5,000 core functions in dynamic system simulation software and release the related code, documentation and test cases.

When AI enters the physical world, responsibility requirements become more concrete. Gao Quanzhong, co-president of SWANCOR Advanced Materials and Qiyuan Robotics, said that consumer-grade personal robots are rolling off the line at a rate of thousands per month, but the generalization challenges and commercial thresholds for entering home scenarios are far higher than those for industrial scenarios.

Gao Quanzhong called on the robotics industry to accelerate the standardization of hardware bodies, operating systems and data interfaces, as the smartphone industry did in its early days, to avoid excessively high costs caused by fragmentation. At the same time, in the face of the red line of trust raised by robots entering private home spaces, the event also jointly released the "Personal Robot Security and Privacy Protection (SSP) White Paper," which clearly proposed three core principles for ethical governance of personal robots: a sense of boundaries, greater transparency, and controllability and accountability.

In the roundtable discussion, the topic further expanded to education and future technologies under AI. Yurii Nesterov argued that AI must be governed and proposed the idea of making the human-machine communication process inspectable; Noga Alon warned that the internal mechanisms of large language models are still not transparent enough, and competition may also hinder prudent development; Chen Yunji believed that the accuracy of current statistics-based large models is very difficult to reach 100%, and that physical rules and logical rules will need to be introduced in the future; He Jifeng stressed that humanity must still control the direction of AI development.

This heated discussion around artificial intelligence took place in Shanghai, and not by chance. As the permanent venue of the World Laureates Forum, the Lingang New Area of Shanghai is also one of the world's most concentrated testing grounds for artificial intelligence. From the deployment of high-performance chips at the "Dishui Lake AI Innovation Port" to the batch validation of embodied intelligent robots in real scenarios, Shanghai's current AI layout today advances in parallel across theory, chips, embodied intelligence and security governance, also reflecting Shanghai's open posture in accelerating the building of an AI highland with global influence. Statement: Exclusive article from Sina.com, unauthorized reproduction is prohibited. tech_web_article_v2015_block_bottom Non-targeted 300*250 button 17/09 wenjing begin

Non-targeted 300*250 button end