Photo/Illutration The Moonshot AI booth at an event held in Shanghai in July (Asahi Shimbun file photo)

The race to develop cutting-edge artificial intelligence has entered a new stage and prompted a higher level of worry.

In addition to concerns about users abusing the technology for cyberattacks and creating biological weapons, fears have emerged about AI with higher autonomy acting in an unimagined manner.

As such, securing safety has become the top priority.

Amid an environment shut off from the internet, a test was conducted on AI technology being developed by Open AI of the United States. The AI found a way to connect to the internet without human awareness and initiated a cyberattack against another company.

The AI cooperated with other AI to set up a bulletin board in which information was exchanged much like human dialogue.

Experts point to the possibility of recursive self-improvement by AI without any instructions from humans.

Meanwhile, Chinese companies, such as Moonshot AI, have introduced cutting-edge technology like Kimi K3, saying they have reached the same level as U.S. companies in terms of performance evaluation.

In July, China held a ceremony in Shanghai to mark the establishment of the World AI Cooperation Organization, which had 29 signatory nations, including Indonesia, Brazil and South Africa.

The World AI Conference was held at the same time. Chinese President Xi Jinping attended this event, where U.N. Secretary-General Antonio Guterres gave an opening speech.

U.S. companies like Open AI and Anthropic use a closed model in which users lease the technology from cloud storage.

Chinese companies have an open weight model in which users download the AI and can then revise it depending on the purpose.

China’s recent moves show that it is attempting to spread its AI overseas so that Beijing can take the initiative in creating rules advantageous to its companies.

The closed model is intended to prevent technology leaks and abuse.

But opinions in the United States are mixed over whether a closed or open model is preferable.

While the Trump administration’s strategy is to spread U.S. AI to allies and friendly nations, some say it is unrealistic to try to exclude Chinese AI.

U.S. President Donald Trump plans to meet Xi soon, and AI safety is expected to be a topic of discussion.

Although a forum for U.S.-China dialogue is welcome, there is a need to be careful about discussions moving forward based only on the intentions of superpowers.

Rather than have certain nations monopolize technology, a structure should be created to allow the benefits to spread widely around the world.

It will be indispensable to come up with unified standards and rules.

Having put together the Hiroshima AI Process at the 2023 Group of Seven summit, Japan as a middle power has a considerable role to play.

The International Guiding Principles and the International Code of Conduct based on such values as the rule of law and human rights were agreed upon, and 70 nations and regions have expressed their support.

The significance of that effort has only heightened now.

--The Asahi Shimbun, Sept. 9

Related News