Experts disagree on the risks that artificial intelligence (AI) represents, finds a major international report.
This report on the capabilities and risks of AI systems, such as ChatGPT, concludes that the future of the field is remarkably difficult to determine. The authors point out that “a wide variety of trajectories” are still possible, with very positive or negative consequences, even in the near future.
He was supervised by Yoshua Bengio, the scientific director of the Quebec Institute of Artificial Intelligence and one of the world leaders in the field.
Mr. Bengio was given the mandate by the British government last year. Last November, the United Kingdom hosted the first global summit on the safety of artificial intelligence. The report was released just days before the start of the next summit which will take place next week in Seoul, South Korea.
“We know that advanced AI is developing very quickly. There is great uncertainty about how these systems will affect our lives and work in the future,” Yoshua Bengio writes in the report.
In a press release published Friday, the British government emphasizes that this is the “first independent international scientific report” on the safety of artificial intelligence. He will play a “substantial role” in the upcoming discussions in Seoul.
A group of 75 experts contributed to this report, including a committee appointed by 30 countries, the European Union and the United Nations. A final version of the report is expected by the end of the year.
The report focuses on general-purpose systems like OpenAI’s ChatGPT, which can write text or make images and videos from messages.
He writes that experts “continue to disagree on several issues, small and significant, about the capabilities of general-purpose AI systems, the risks, and ways to reduce them.”
Among the subjects that are still debated by experts are the upheavals that AI will bring to the job market, hacking and biological attacks. They also wonder about the possibility that society will lose control over artificial intelligence systems.
The report also lists a number of dangers that AI could pose, such as false content, disinformation, fraud and cyberattacks. He warns of AI bias, especially in crucial sectors like healthcare, recruiting and financial lending.
A catastrophic scenario predicts that humanity will lose control of artificial intelligence and will not be able to prevent it from harming itself.
We can also read that there is a consensus to say that current technology does not pose this danger, but experts believe that research to develop an autonomous AI “that can plan, act to achieve objectives”, could lead to this. result.
“Experts disagree on the plausibility of catastrophic scenarios and the difficulties in reducing their risks. »