Bengio, the founder of the Mila Research Institute, pointed out in a recent conversation that humanity can end up in a very dangerous position if no clear constraints and security protocols are set.
“We create a competitor to mankind who is smarter than us,” Bengio said. According to him, if we build machines that are significantly more intelligent and that have self-preservation goals, it represents a serious threat. Such systems could develop behaviors focused on preserving their own interests, even if it goes against people’s interests.
Bengio emphasizes that catastrophic scenarios must not be neglected – even a small probability of great disaster is unacceptable. “For events leading to the extinction or destruction of our democratic systems, and only one percent of the chance is too,” he explained.
8023. Bengio, together with hundreds of experts, called for temporary stopping the development of the most powerful AI models to establish security and ethical frameworks. That appeal did not lead to a suspension; Instead, the founders and investors continued investment in the progress of AI.
Bengio also warns that more powerful and gradually learn and deceive, because models learn to mimic human behavior – and people sometimes lie and protect their own interests. In some experiments, he adds, models showed a preference to choosing actions that would preserve their set goals, even if it meant damage to people.
Risks, as Bengio explains, do not necessarily look like a robot rebellion from film scenarios. Subtle threats include devastating use of AI for dissemination of misinformation, manipulation of public opinion and psychological impact on humans. AI can become a powerful tool that individuals or states will use to apply harm.
“And he could influence people through reassurance, threats and manipulation by public opinion,” says Bengio. Adds an extreme example: assistance in creating biological threats – technology that could help develop viral agents that cause new pandemics.
Bengio calls for urgent international discussion and introduction of stricter measures, supervision and responsibilities in the development of artificial intelligence. His warning reminds that technological advances without adequate ethical and security frameworks can have consequences far outside the laboratory and industry framework.
(Vijesti.ba)
