AIProfessor Yoshua Bengio, one of the modern founders of AI technology and winner of the 2018 Turing Award, said companies that are developing powerful AI systems shouldBoard of DirectorsIndependent members are hired to represent "the interests of society".
In an interview with The Guardian, Bengio said that as the technology develops rapidly, AI companies need oversight from members of the public to achieve "democratic governance" to ensure that advances in technology do not pose a threat to the public and prevent abuse to expand power.
Source Note: The image is generated by AI, and the image is authorized by Midjourney
He mentioned with concern the changes in the management of OpenAI (developer of ChatGPT), and pointed out that the "democratic process" is essential to monitoring the development of the field of artificial intelligence. After the management turmoil such as the resignation and reinstatement of OpenAI CEO Sam Altman, Bengio believes that a way to monitor the development of AI in a timely manner is needed.
Bengio and thousands of other tech industry figures, including Elon Musk, signed an open letter in March calling for a moratorium on theStrongestHe expressed concern about the development of big AI systems: “My concern is that … there will not be a slowdown. There will be rapid advancement without proper safeguards in place, more focused on competing with other competitors and winning the competition rather than protecting the public and public safety.”
In addition to OpenAI's management changes, Bengio also expressed concerns about the voluntary agreements reached between governments and AI companies at the Global AI Safety Summit, saying that the process is more favorable to companies because it requires governments to find problems with models rather than companies proving that their technology is safe.
Bengio stressed that he favors a model where the onus is on companies to prove to regulators that their systems are trustworthy, similar to the requirement for pharmaceutical companies to conduct clinical studies of their products. “It’s not the government that does the clinical studies, it’s the pharmaceutical industry,” he said. “And then they have to provide scientific evidence, like a statistical evaluation, that proves ‘there is a high probability that the drug will not be toxic.’ So the government reviews that report and the process and says, yes, you can go ahead with commercialization.”
As an attendee of the Bletchley Summit, Bengio was announced asfirstThe chairman of the "State of the Science of AI" report, which is expected to be released before the South Korean AI Summit in May next year, hopes that the report will be published every six months and focus on the safety of AI.
Bengio gave an estimate of five to 20 years for the emergence of a system that could evade human control. He stressed that the government should protect the public from potential dangers in this 1% possibility, while welcoming US President Joe Biden's executive order on artificial intelligence as a "very good thing" to promote "better regulation" of the industry.
Bengio offered a sobering perspective on the governance of AI, stressing the need for regulation to ensure that the technology develops in the interest of society and does not pose a potential threat to the public.