Experts: Develop AI as safe, useful 'knife'

Greater efforts are needed to make artificial intelligence safe for human society — and regulators over the world must set ethical boundaries before AI technologies and algorithms make further advances, industry experts said on Thursday.

"A knife can be used for chopping food ingredients, but it can also be used to kill people. Ideally, we should develop AI into a sharp knife that can be a practical tool for human beings but will never harm us," said Zhu Shiqiang, director of Zhejiang Lab and deputy president of the Artificial Intelligence Industry Alliance, at a forum on AI and digital ethics at the 2022 World Internet Conference Wuzhen Summit.

Zhang Bo, president of the Institute of Artificial Intelligence at Tsinghua University and an academician at the Chinese Academy of Sciences, said measures to prevent the risk of people losing control over AI-enabled machines and potentially sentient robots must be put in place, or else, science fiction scenarios may come to pass.

"We used to believe that machines can become threats to human beings only when AI surpasses, or gets close to, that of humans. But the fact is, although AI is nascent, it appears as though we've started to lose control sooner than expected," Zhang said.

Currently, AI-enabled technologies can generate quality information, including text and images, to meet the requirements of users. But AI can also be freed of user control and manipulated to deliver biased and inaccurate information. Decisions based on such information can lead to disasters. Yet, people should give the AI industry liberty and autonomy to develop in the right direction, Zhang said.

"AI systems are also vulnerable to attack as algorithms aren't safe, trustworthy or reliable. When systems are under attack, they may even act destructively," he said. "Technologies like AI deepfakes — the manipulation of videos and images to create seemingly realistic but fabricated content — have had a huge impact on society."

Developers must assess AI systems scientifically during the development stage and have them closely monitored. Also, measures to prevent systems from being attacked or going rogue must be put in place, Zhang said.

Zhu of Zhejiang Lab said that, although currently, AI is still nascent, people must attach great importance to preventing it from being misused while pushing forward its development.

The lab, he said, has started developing methods to prevent AI from harming society and keep it under control.