People's Daily: Creating a smart age of benign interaction

The potential security risks of artificial intelligence algorithms, cognitive bias, legal ethics conflicts, etc. are often the reflection of real social defects. Only by focusing on machines and people using machines, can we create a future where artificial intelligence interacts with human beings.

Voice ordering, paying for the face, unmanned car to go home... More than 60 years ago, when a group of young scientists first proposed the concept of artificial intelligence, perhaps did not expect the huge changes it brought; nowadays, thanks to calculations The improvement of capabilities, the drive of big data and the breakthrough of deep learning, artificial intelligence is accelerating 'evolution' and is increasingly changing our production and life.

Technology is a double-edged sword. It should be noted that artificial intelligence has also raised new challenges to safety, legal ethics, etc. while triumphing. For example, some experts worry that artificial intelligence may blur the boundaries between the virtual world and the physical world, and even reshape Human living environment and cognitive form, and thus derived a series of difficult moral and ethical problems. Well-known physicist Hawking said that the short-term impact of artificial intelligence is determined by the person controlling it, while the long-term impact depends on labor. Whether intelligence is completely controlled by people.

From the perspective of technological development process and logic, these concerns may come early. Fortunately, we are not just stuck in the debate, the potential risks of artificial intelligence, scientists and engineers to prevent problems, from product design standards, normative constraints, etc. In this paper, a series of feasible schemes for trying to control intelligent machine systems are proposed. In the face of incomprehensible intelligent algorithms and lack of machine bias caused by transparency, people are strengthening the evaluation of data, algorithms and models used by artificial intelligence to eliminate potential factors that can lead to bias. For the future strong artificial intelligence environment, how the machine with quasi-personality can get along with people. Scientists also try to explore the embedding of human values ​​and norms by intelligent algorithms, so that they have the same compassion and responsibility as human beings. Shame and so on. We hope that while the artificial intelligence program is supervised, let the artificial intelligence algorithm follow the 'good law' and absorb the 'good' side of human morality, thus controlling the risk of the machine.

Of course, it is not easy to build a machine legal ethics system. In the real language, rich language expression, subtle emotional appeal can be understood and accepted by the machine? Is it possible to use quantitative indicators to measure ethical concepts such as good and evil? Determining algorithms, setting standards? These are the challenges we face in trying to regulate the development of artificial intelligence. But the difficulty of embedding ethics into machines does not mean that we can only do nothing. The judgment and behavior of machine learning output often depends on it. Accepted information. From this point of view, artificial intelligence itself is not right or wrong. Under ideal conditions, the current artificial intelligence algorithm based on deep learning should be neutral, objective, and efficient, its potential security risks, cognitive bias, Legal ethical conflicts, etc. are often reflections of real social deficiencies, but are magnified in the field of technology. Therefore, to make smart machines more perfect, in addition to reviewing or guiding algorithms through technical means, more importantly It is to face up, understand and improve the real social environment. Only when the machine and the person using the machine focus on each other, can we create an artificial intelligence Positive interaction with humans in the future.

Ethicists say that the moral level of artificial intelligence will be a mirror, which reflects the moral height of human civilization. Human society cannot be perfect, and the future of machine intelligence is also the same. We must continue to work tirelessly to create better The era of intelligence.