This week, from the China Development Forum, to the Boao Forum for Asia, to the Zhongguancun Forum, these forums have focused on the economy and technology, among which at the level of science and technology, the development and discussion of artificial intelligence and humanoid robots have received special attention.

On 1 month 0, the annual meeting of 0 Zhongguancun Forum was officially opened, and this forum focusing on science and technology also set up more content related to artificial intelligence. Now, what is the stage of the development of artificial intelligence? Building on what is already there, how can AI continue to move forward? At the same time of development, how to solve the problem of security? "News 0+0" connected Huang Minlie, professor of the Institute of Artificial Intelligence of Tsinghua University, and Zhang Linghan, professor of the Institute of Data Rule of Law of China University of Political Science and Law, to bring analysis and interpretation.

How is the focus of AI changing?

Huang Minlie, Professor, Institute of Artificial Intelligence, Tsinghua University: I think there are two hot spots for this year's changes:

Embodied intelligence, this year's embodied intelligence is indeed in addition to the generative large model, the most important research hotspot and industrial layout direction, this direction of investment and financing is very active, because embodied intelligence is through a relatively embodied large model and robots combined, can be used to achieve such as humanoid robots we see today, autonomous driving, and so on such as some very visible application scenarios.

Another important change is AI for Sciences, why is AI for Sciences so important? It's because of the large models in the past, which may only be able to chat with people, help people write official documents, and do some documents, but can AI help do innovation? To solve some major scientific discoveries, such as to explore how to discover new materials, to solve some very important basic science problems, such as medical research, and even life science research. This direction is actually the future, how to really use AI to create new things, discover new knowledge, and then explore some important scientific discoveries that have not been explored by humans in the past, which is a very important focus for us.

At this year's Zhongguancun Forum, these two points were also very well reflected.

What is the key to developing AI applications in industrial scenarios?

Huang Minlie, Professor of the Institute of Artificial Intelligence of Tsinghua University: I think that the application of artificial intelligence in industrial scenarios in the future is a very important direction, and it is also the most important breakthrough direction to achieve new quality productivity. Because AI applications in industrial scenarios require high reliability, we need to solve three important problems:

Whether our intelligence can be improved to a higher level, although the current large model has been very intelligent in many aspects, but our data in industrial scenarios should be deeply combined with the actual workflow, and this combination of our intelligence can achieve a higher level.

The problem of its reliability is relatively small in industrial scenarios, and once an error occurs, it may cause destructive effects and is also relatively significant, so it needs to have a relatively high reliability. In terms of reliability, it is also a very important direction of current artificial intelligence research.

The issue of security, security means that we have applied a lot of new generation of artificial intelligence technology, in fact, the relationship between our whole person and AI, and what profound impact AI will have on the development of the whole society, in fact, this is a governance issue. Security and governance are also important issues to focus on in the future.

How to deal with the risks and security issues in the development of artificial intelligence?

中國政法大學數據法治研究院教授 張淩寒：實際上我們在監管法律法規的制定當中，已經充分考慮到了AI的風險問題。在去年9月，我國頒布了《人工智慧安全治理框架》，其中就把人工智慧的風險劃分為兩類——內生風險和應用風險，實際上就是AI的技術風險和濫用風險。對於內生風險的治理，我們提出的原則是“以技治技”，用技術來治理技術。人工智慧的技術，在進行研發和部署的階段，就要去充分進行測評，進行價值的對齊，並且去進行持續的運行監控。對於應用的風險，其實我們不光要關注對於AI的法律治理和監管，與AI相關的一些產業的法律也要充分去制定和考量，比如說像自動駕駛、智慧醫療、AI陪伴等等，充分地去做一個全面的規劃。

Editor/Zhou Chao