On the afternoon of November 10, the World Internet Conference Artificial Intelligence and Digital Ethics Sub-Forum was held. In the intensive sub-forum agenda of the World Internet Conference, up to 12 sub-forums are held at the same time, and only this "exclusive" half-day meeting is held.

The theme of this sub-forum is "The Industrial Direction and Ethical Orientation of Artificial Intelligence", which is divided into three topics: the technical logic of artificial intelligence ethics and the space for human choice; the legal principles and principles of digital ethics - the integration of the real world and the digital world Order and rules; the impact of artificial intelligence and its industrial model optimization on future production and life forms of human society.

"The purpose of this forum is to deepen dialogue between scholars and the industry at home and abroad, promote the exploration and construction of the ethical research of artificial intelligence and the exploration and construction of the governance system, promote the research and development and application scenarios of safe, reliable and controllable artificial intelligence technology, and ensure that artificial intelligence always moves towards Developing in a direction of people-oriented and technology-oriented towards goodness. "Li Yuxiao, Secretary-General of the China Cyberspace Security Association, said.

In 2017, the "Asiloma Conference" was held in Asiloma, California, USA. Experts and scholars in the field of artificial intelligence focused on the potential problems of the rapid development of artificial intelligence, reached a number of principles, and proposed the goal to create beneficial artificial intelligence. Last year, my country issued the "Ethical Norms for the New Generation of Artificial Intelligence", which put forward six basic ethical requirements: to enhance human welfare, promote fairness and justice, protect privacy and security, ensure controllability and credibility, strengthen responsibility, and improve ethical literacy.

At the forum on the 10th, scholars and experts explained their views on the ethics of artificial intelligence from different angles, and The Paper () sorted out some of the content.

Zhang Bian, honorary dean of the Institute of Artificial Intelligence at Tsinghua University and academician of the Chinese Academy of Sciences, delivered a speech online. All pictures in this article are from the official website of the World Internet Conference

Zhang Yan, honorary dean of the Institute of Artificial Intelligence at Tsinghua University and academician of the Chinese Academy of Sciences: Clarifying ethical standards requires global cooperation

"When machines continue to improve and develop, one day their intelligence surpasses humans, especially after they have subjective consciousness, humans will lose control of the machines and bring disastrous consequences" - The risks of artificial intelligence have long been raised, and Hawking also issued a warning. , but we once thought it was a long-term view.

By the beginning of this century, when deep learning based on big data rose in artificial intelligence, people's understanding had changed, and they felt that the ethical risks of artificial intelligence were right in front of them and governance was imminent.

According to the current artificial intelligence technology, we can use deep neural networks to generate text and images that meet the requirements and have good quality according to the user's requirements. Similarly, we can also violate the requirements and generate biases such as race, gender, and injustice and Text and images with errors are not controlled by the user. If decisions and predictions are made based on these wrong texts, serious consequences may be brought.

If you want the machine to develop towards intelligence, you cannot let the machine be completely at the mercy of humans. You need to give it a certain degree of freedom and autonomy - Deep neural networks are based on this principle: using the data tool of probability, so that machines can generate rich Text and images. But for this reason, there must be a probability of generating unqualified and harmful texts and images.

The development of artificial intelligence will inevitably bring an impact on ethics and traditional norms. Fundamentally speaking, the research and development of artificial intelligence must be people-oriented and start from the ethical principles of fairness and justice. To this end, we need to work hard to establish explainable robust (transliteration, meaning "strong", which refers to the ability of the system to survive in abnormal and dangerous situations) artificial intelligence theory, and only on this basis can we develop security , trustworthy, controllable, reliable and scalable artificial intelligence technology.

The research and governance of artificial intelligence requires the cooperation of people in different fields around the world. In addition to those engaged in the research and development and use of artificial intelligence, legal, moral, ethical and other people also need to participate. It is necessary to clarify ethical standards, which requires global cooperation and jointly formulate standards that are in line with the interests of all mankind.

Zhu Shiqiang, Director of Zhijiang Laboratory and Vice Chairman of China Artificial Intelligence Industry Alliance

Zhu Shiqiang, director of Zhijiang Laboratory and vice chairman of China Artificial Intelligence Industry Alliance: Build a technical map to prevent potential harm and use Tao to control magic

Artificial intelligence technology brings convenience to production and life, and also brings negative things, causing consideration and even excessive concern about artificial intelligence ethics.

The reality is that artificial intelligence is still weak artificial intelligence, which has higher individual abilities than humans in some individual fields, such as Go and protein sequencing, but the comprehensive intelligence is still very weak. From a technical point of view, intelligence is relatively easy to obtain, but it is difficult to reproduce wisdom with machines.

My point of view is that artificial intelligence is an extension of human intelligence in the field of machines, and the real initiative is still in control of people. The most important thing at this stage is to develop technology, but at the same time, we must prevent the ethics of human beings because of the technology itself and its abuse. question.

Zhijiang Laboratory sorted out the AI ​​ethics map and classified it from three aspects (research and development subject, user, and management subject). We are also building a technological map to prevent artificial intelligence from harming society, trying to use technology to prevent the abuse and malicious application of artificial intelligence. Our logic is that the kitchen knife is used to cut vegetables. Although the kitchen knife has been used to kill people many times, it has no original sin itself. Our logic is to "us the Tao to control the demons", and use technical means to make the kitchen knife sharp, but let it last forever Can't cut people.

Yang Debin, member of the Digital Economic Development Committee of the Hong Kong Special Administrative Region Government, delivered a video speech

Yang Debin, member of the Digital Economic Development Committee of the Hong Kong Special Administrative Region Government: Human beings should improve themselves and avoid gradual "degeneration"

Computers have increasingly stronger computing power and can do more and more things for humans. Many of the work done by manpower have gradually been replaced by machines due to their repetition, danger and cost issues, such as the "intelligent manufacturing" Black light factory”, drones replace humans to detect sewers, robots replace security patrols, etc.

Machine learning triggered by big data has flourished in various industries in recent years. Robots defeated Go Jiudan Lee Sedol and world-leading Ke Jie in 2016. In recent years, machine learning has been in driverless driving, medical diagnosis, and even writing, drawing, and writing. Consequent progress has been made in poetry. Despite this, artificial intelligence is still far from human intelligence.

Artificial intelligence and human intelligence can be compared to the tortoise and hare racing. Humans are rabbits and machines are turtles. As humans become more and more dependent on machines and more machines can do, the situation will gradually change. When humans hand things over to machines, will we become lazy or even stupid? Before, we would remember the phone calls of several family and friends we often contacted. Now, as long as we call the phone automatically, we don’t need to remember it. Over time, our memory will become worse.

What I am more worried about is not that machine learning will surpass humans one day, but that humans become unwise due to over-reliance on machines. The tortoise and hare racing is a fable, which tells us that we cannot lose to our opponents because of complacency. We can teach machines to be smarter and to properly restrain machines so that they do not harm humans, but humans should improve themselves and avoid gradual degradation.

Joseph Sfakis, winner of the 2007 Turing Award and member of the World's Top Scientists Association, delivered a video speech

Joseph Sfakis, winner of the 2007 Turing Award and member of the World's Top Scientists Association: The Choice and Balance of Artificial Intelligence and Ethics

I want to say first that the advancement of artificial intelligence has enabled automated systems to be transformed into autonomous systems, which reflects the trend of transformation from weak artificial intelligence to universal artificial intelligence.

Normal ethics is to make the right decision. In order to make the right decision, first you should have a sense of choice. Nowadays, many people are confused by different opinions on how to use artificial intelligence. I think society must create conditions to engage in calm, informed and reasonable debates. We should achieve between strictly implementing the principles of risk prevention and accepting innovation to improve efficiency or quality. balance. When we download a system now, we will face the dilemma between choice risks and performance.

Another important dilemma is how to divide labor between humans and machines. Innovation and technological advancement may mean the loss of skills, and it is hard to imagine that machines can surpass human intelligence, and even if we are greedy for laziness and comfort, human intelligence can also be tamed and conquered. If we are defeated by the complexity of machines, it will be very bad for humans.

Jiang Bixin, Vice Chairman of the Constitution and Law Committee of the National People's Congress

Jiang Bixin, Vice Chairman of the Constitution and Law Committee of the National People's Congress: The "five origins" and "six complete" construction of artificial intelligence ethics

Why build artificial intelligence ethics and digital ethics? It can be simply summarized into "five sources": it originates from the duality of digital technology and artificial intelligence, which brings both benefits and risks; it originates from the boundary, transnational, professional, and non-transparent nature of digital networks; it originates from The limitations of legal norms stem from the incompetence of moral ethics to human beings; the dependence on human morality by the development of cybersecurity.

I believe that the basic rules of this ethical system are seeking goodness, truth and beauty. This is also a consensus of human beings, and specific rules are formed based on different periods, themes, environments, and fields - the current preliminary stage of artificial intelligence and the possible future supernatural powers In a strong stage, the specific rules are changing.

How to change the rules from ideas to reality and what mechanism can establish to make them reality? I summarize it as "six complete": all concepts, all subjects, all chains, all elements, all cycles, and all tools.

Gao Shiming, President of China Academy of Fine Arts

Gao Shiming, Dean of the China Academy of Fine Arts: Artificial intelligence can write poetry and draw, but does not have the impulse and passion to do these, while humans have

The significance of artificial intelligence to us is not only to simulate and extend human thinking, but more importantly, the development of AI technology has made great progress in our understanding of human language, thinking and intelligence itself, which helps explore human thinking, learning, analysis, What kind of intellectual process is reasoning, planning and imagination.

AI is not only a tool for research and manufacturing, but also a tool for perception and thinking. It is a mirror of human self-cognition and a mirror of human social consciousness. It is not even just a tool, it is also a partner. AI is part of us and a part of our future.

Desire and will may be the biggest difference between us and artificial intelligence. Artificial intelligence can write poetry, draw, and compose music, and can defeat humans on the chessboard, but it must accept instructions to write poetry, play chess, draw, and compose music. It does not have the impulse, desire and willingness to do these behaviors, and Humans not only have desires, but also passions.

How to use algorithms to deal with the complex, irrational and contingent interactions between countless social individuals is a difficult point in AI technology. There is no equation in the mind, and the social mind cannot be solved by algorithms.

There are many concerns about artificial intelligence. One of them is that artificial intelligence is gradually replacing and displacing our sensibility through computing and simulation. AI's infinite computing power also puts pressure and driving force on human imagination. I think the development of artificial intelligence is not to replace people. It in turn forces us to re-understand what creation is. Creation is the most meaningful production. Creation also involves curiosity, impulse, sensibility, care, desire, value, purpose, and the most important thing is What is the love for the unknown and unknown - as long as we still have these, we can boldly announce that we can let it go if we can be replaced by artificial intelligence.