Center for Strategic Assessment and forecasts

Autonomous non-profit organization

Home / Science and Society / Future of the Information Society / Articles
"People will have to think about whether they are ready to entrust their lives to a computer." What will AI achieve in 20 years
Material posted: Publication date: 15-12-2021
Kaifu Li is the former head of Google in China and the author of the book "Superpowers of Artificial Intelligence: China, Silicon Valley and the New World Order." Recently, Li and the author of the cyberpunk novel "Garbage Surf" Chen Qufan wrote a joint work on the near-term prospects for the development of artificial intelligence - "AI 2041". Noema magazine published an interview with Lee, where he told how AI will change healthcare, education and international politics.

But this will not happen soon, since we are talking about training algorithms on personal data of patients. In addition, the use of software in medical practice is associated with legal and moral difficulties, for example, with the question of responsibility for incorrectly chosen treatment. Therefore, the AI should be nothing more than an assistant. The final decision will be left to the doctor.

Materials scientist Vaclav Smil recently noted that although mobile phones have become lighter than before, there are a lot more of them. "The total volume of materials used in the production of mobile phones has increased, not decreased. People often confuse relative and absolute dematerialization. It is important to take into account the absolute consumption of energy and materials," he said.

Given all this, does the development of AI depend on the transition to renewable energy sources?

At the moment, we are a little stuck, because at the current stage of AI development, very complex calculations need to be carried out. As more and more companies follow the path of Microsoft, Google and OpenAI, the need for computing power will grow. That's why you need to invest in efficient software that will make advanced technologies more accessible, even at the cost of reduced performance. In addition, of course, we must switch to more environmentally friendly and cheaper energy sources. In the book, I wrote about distributed energy, the simultaneous use of solar panels and the latest, non-lithium batteries. According to the optimistic scenario, in a few decades we will be able to reduce the cost of energy by 10%. Although this may not be enough.

But there is a downside. We need to figure out how to improve AI in other ways. Otherwise, we will never be able to create a common artificial intelligence.

— What are the limitations of AI? Some critics argue that although smart machines perform many tasks more efficiently than people, they "do not understand" what they are doing. Understanding is born out of context and requires a unique human ability to fill gaps between data fragments.

Can AI do everything we do? At each stage of AI development, we formulate questions taking into account what we humans are capable of: can he play chess? If he can play chess, can he play go? If he can understand what is written, can he reach our level of self-awareness?

It is encouraging that, despite all the geopolitical problems, scientists are still working together. At major AI conferences, American, European and Chinese scientists share their discoveries and share their experience with others. I want to believe that one day we will come to the point where China and the United States will compete in some areas and cooperate in others.


RELATED MATERIALS: Science and Society