New AI could be used to create biological weapons

:Phone showing OpenAI's GPT-o1 learning interface - Photo: FINANCIAL TIMES

According to the Financial Times on September 13, the company OpenAI admitted that the GPT-o1 model, which was just launched on September 12, "significantly" increases the risk of artificial intelligence (AI) being exploited to create biological weapons.

A paper explaining how GPT-o1 works published by OpenAI states that the new model has a “medium risk level” for issues related to chemical, biological, radiological and nuclear weapons. This is the highest risk level OpenAI has ever assigned to its AI models.

According to OpenAI, this means GPT-o1 could “dramatically improve” the ability of experts who want to abuse them to create bioweapons.

Mira Murati, chief technology officer (CTO) of OpenAI, said that although GPT-o1 is still in the experimental phase, the model's advanced capabilities still force the company to be extremely "cautious" about its roadmap to the public.

Ms Murati also confirmed that the model had been thoroughly tested by a multidisciplinary team of experts hired by OpenAI to test the limits of the model developed by the company.

Testing showed that the GPT-o1 model meets general safety criteria much better than its predecessors, Murati said.

Not stopping there, OpenAI representatives also reassured that the trial version of GPT-o1 is "safe for deployment" and cannot create more complex risks than the developed technologies.

Despite the above assurances, a large part of the scientific community and those interested in AI are still not very optimistic about the development direction of this technology.

Concerns that AI is developing too quickly and outpacing the pace at which regulations are being developed and approved to govern it are growing.

If OpenAI now poses a “medium risk” of chemical and biological weapons, “it only reinforces the importance and urgency” of legislation, such as the one being hotly debated in California to regulate the field, said Yoshua Bengio, a professor of computer science at the University of Montreal and one of the world’s leading AI scientists.

This scientist emphasized that the above institutions are especially necessary in the context of AI increasingly approaching the level of artificial general intelligence (AGI), which means having the ability to reason on par with humans.

"The closer pioneering AI models get to AGI, the more risks will increase if proper precautions are not taken. AI's improved reasoning ability and its use for fraud are extremely dangerous," Mr. Bengio assessed.

Source: https://tuoitre.vn/ai-moi-co-the-bi-dung-che-tao-vu-khi-sinh-hoc-20240914154630363.htm