On August 29 in Shanghai, a debate between Jack Ma, founder of China's tech giant Alibaba, and Elon Musk, CEO of U.S. tech firms Tesla and SpaceX, took place at the World Artificial Intelligence Conference 2019 (WAIC), which was described as "Mars hits Earth."

While Musk has his heart set on sending 1 million people to Mars, Ma has his set on providing sustainable solutions for 7 billion people on Earth. Along with this contrast, the two also have differing views on artificial intelligence (AI).

Musk said he believes "people underestimate the capability of AI" and warned that human beings "will be surpassed by AI in every single way." Meanwhile, Ma said "a computer is clever but human beings are much smarter," adding in response to the threats of AI, "young people will find solutions."

Debates on the threat of AI have never ceased, but scientists and business people in China and the United States have also never ceased their efforts to tap into the market.

Hidden dangers

Recently, a Chinese AI-based face-changing app, which allows users to substitute their faces with celebrities' in various video scenarios and share them with friends through social media, went viral. But to use it, they have to agree to its registration provision, which stipulates that any photos uploaded can be permanently used for free by the app. One day after its debut, social media platform WeChat banned the app's links on the grounds that it was an invasion of users' portrait rights that could lead to illegal uses. The news once again awakened people's anxiety about AI's threat to personal privacy.

"The real security threat brought about by this face-changing software is that when this technology is popularized, traditional face recognition applications will be challenged," said Pei Zhiyong, a researcher with Qi An Xin, a Chinese security provider.

For example, there was a case in China a few years back where photos were being changed into dynamic photos with eyes and mouths that could move, which were then used to hack people's facial recognition logins.

Such problems arouse worldwide concern. In February, 26 expert groups, including Musk's research team, released a report examining the challenges related to AI.

The report said AI can be used to threaten the safety of personal information, be applied to make personal attacks and even imperil existing political systems based on digitalization, which could undermine the whole world security landscape in the future.

In addition, the utilization of AI can enhance the power of all kinds of people including terrorists, which could result in damage to the stability and security of human society, according to the report, which added that in the future, AI attacks will be beyond imagination and truth will no longer exist.

Liu Shanquan, an official with the Shanghai Municipal Commission of Economy and Informatization, said since the development of AI technologies is at its initial stage and moving at a fast pace, giant tech corporations and startups worldwide are investing a great deal of resources in related research programs, but paying little attention to the security of data.

Currently, AI security threats include the quality of data and the uncontrollability of operation. Since AI is a data-driven technology, the quality of data can directly influence the result of operation and malicious manipulation can skew the results. Moreover, what if machines' decisions are different from people's decisions? These are the problems that need to be solved but are currently being overlooked.

Visitors watch AquaJellies at the World Artificial Intelligence Conference 2019 in Shanghai on August 29 (XINHUA)

95 percent left

Nevertheless, China is accelerating its pace in AI development. According to a recent report released by the International Data Corp., a global market intelligence company, it is estimated that China's AI market will maintain a compound annual growth rate of 44.9 percent, hitting $17 billion in the next five years.

Some experts believe there is still a long way to go for AI to surpass and control human beings. Song-Chun Zhu, a Chinese-American AI and computer vision expert, raised the example of chair recognition at the WAIC. He said to teach a machine to recognize chairs, people need to collect a large number of samples including different shapes, materials, colors, etc. "But new chairs will continue to emerge, so if machines just remember the existing information of chairs instead of understanding why new chairs appear and what their function is, they can never make a satisfactory recognition," he said.

But human beings can understand chairs differently since they can feel which part of their body hurts or is comfortable in one, so they can train machine learning models from other angles such as the function of chairs and their purpose.

"Visible things account for only 5 percent of what we know, while invisible things, what we call cognition, occupy 95 percent," Zhu said. "What AI lacks is not the 5 percent that can be seen, but the capacity to understand the remaining 95 percent."

Li Deyi, President of the Chinese Association for Artificial Intelligence, divides cognition into three categories: The preliminary one is feeling, the intermediate one is self-consciousness and the highest one, group consciousness.

At the WAIC, Li said people's skin can feel and react when something stimulates it,

taking just a millisecond for its nerve endings to transfer the message to the brain. To create such a skin would take a hundred years, he said. Besides, higher intelligence comes from memory, which is never equal to storage. He also emphasized the social properties of intelligence as language, knowledge and cultural traditions, which rely on postnatal acquisition. All these things are how human intelligence is distinguished from machine intelligence.

In addition, Li said AI is an extension of human intelligence, which can exist independently without cognition. "I believe as AI's creator, human beings are smart enough to not allow machines to have cognition," Li said, "Instead, we can use AI as our tool for driving, counting and serving us."

He also said the evolution of human beings is in line with their capability to create and use tools. From sickles to atomic bombs, tools are the extension of people's physical ability, while AI is the extension of people's intelligence. Both power and intelligent tools can exist without consciousness.

"Whether AI will lead to a new creature on Earth that can override human beings, and whether it will be our friend or enemy is too early to tell," Li said, adding that even if in the future there is an artificial skin that is able to distinguish physical boundaries, two other conditions are still needed for the new creature to compete with human beings: their own language and characteristics.

Copyedited by Rebeca Toledo

Comments to zhangshsh@bjreview.com