China News Service, March 26. The 2024 Annual Meeting of the China Development Forum will be held on March 24-25, 2024. The "Artificial Intelligence Development and Governance Seminar" was held on the afternoon of March 24. Joseph Schiffakis, founder of Verimag Labs and winner of the 2007 Turing Award, said in his speech in the "Panel Discussion One" session: China can Leverage its vast industrial systems to lead the journey toward autonomous artificial intelligence.
Joseph Schifakith pointed out that in today's world, there are two different views on the ultimate goal of artificial intelligence. Some believe it should be general artificial intelligence, aiming to build superintelligent agents that transcend human thinking capabilities and make effective decisions in fields as diverse as healthcare, finance, transportation, and manufacturing. Others believe that the ultimate goal is to achieve autonomous artificial intelligence and build human-level intelligent machines that can replace human work in existing organizations, such as self-driving cars, smart factories, autonomous networks, etc. But what we have now is still a weak artificial intelligence, an intelligent assistant, which provides services through interaction with users. Joseph Schiffakis believes that there are three different ways to use artificial intelligence, including as an assistant, as a monitor, and as a controller. China can leverage its vast industrial system to take a leading position on the road to autonomous artificial intelligence.
Artificial intelligence faces two main risks, one is technical risk; the other is human risk, which can be controlled through regulatory or legal frameworks. The unemployment problem caused by automation brought by artificial intelligence can also be dealt with through appropriate social policies. In addition, there are two social risks. One is how to strike a balance between choice and performance. If we cannot ensure that the system uses reliable information in a fair and neutral way, we should not give decision-making power to the system. ;The other is performance improvement, and we have to consider whether it is balanced by the lack of human control.
Regarding how to establish a global regulatory framework, Joseph Schifakith said that people all realize that artificial intelligence must be regulated, but there is no consensus on how to regulate it in practice. There are also certain differences in artificial intelligence regulation between the EU and the United States. The EU has stricter and more complete artificial intelligence regulatory laws. In particular, it passed the Artificial Intelligence Act and the Digital Services Act, which adopts a risk management approach and requires artificial intelligence systems to be highly reliable in critical applications. sex. The U.S. regulations are relatively less mandatory. For example, the "Artificial Intelligence Executive Order" issued by it includes some suggestions and guidance. U.S. regulators and big tech companies are aligned in advocating for ethical AI principles. But from a technical perspective, it does not evaluate artificial intelligence systems based on very strict technical standards.
Joseph Schifakith believes that the current global regulatory framework for artificial intelligence advocated by the United Nations has little chance of reaching agreement. The United States is leveraging its strengths in generative artificial intelligence to pursue a self-regulating, market-centered approach to regulation. But in fact, only by achieving a leading position in the field of autonomous artificial intelligence and jointly establishing a safe artificial intelligence supervision system with relevant countries can we promote the great progress of artificial intelligence supervision.