With the development of artificial intelligence, deep forgery technology is becoming more and more mature, especially using deep learning technology to make personal voice, body movements, facial expressions and other false content, such as AI face changing, voice simulation, gesture generation and so on. With the improvement of machine learning ability, the content of deep forgery becomes more and more realistic, and in many cases it can achieve the effect of confusing the fake with the real. For example, in a case of telecom fraud using AI technology issued by Baotou Public Security Bureau, criminals used AI face-changing technology to impersonate friends to defraud the victim of transferring 4.3 million yuan. In the era of artificial intelligence, how to prevent the risks of artificial intelligence technologies such as deep forgery is in urgent need of a clearer response from laws and regulations.
Preventing artificial intelligence technology from being used maliciously and guiding digital technology to be good is the meaning of governing the country according to law.
First of all, we should be alert to the new risks brought by technologies such as deep forgery.
The birth of new technology will always be accompanied by new risks, which is an inevitable law in the history of technology development. Cars cause traffic accidents, the Internet brings cyber crimes, and artificial intelligence causes intelligent crimes. In the United States, generative artificial intelligence has caused academic fraud and other problems. It is conceivable that more artificial intelligence technologies will be applied to crimes in the future. "Baotou AI fraud case" is only the application of artificial intelligence technology in the field of crime, and there has been a black and gray industrial chain of "AI face-changing", that is, providing customized video face-changing services on the network platform. Laws and regulations need to be followed up in time to improve the risk loopholes that artificial intelligence technology may bring.
Second, we should set a bottom line for the application of artificial intelligence technology.
Faced with the risks brought by artificial intelligence technology, countries are taking measures to actively respond. The Biden administration has begun to study whether it is necessary to check artificial intelligence tools. China is also actively legislating to prevent the risks of artificial intelligence. The National Internet Information Office issued the Administrative Measures for Generative Artificial Intelligence Services (Draft for Comment), which takes generative artificial intelligence as the core and restricts the algorithm design, training data selection, model generation and optimization, service provision, generated content, personal information and privacy protection.
Laws and regulations should reduce the risk of artificial intelligence to an acceptable level. The main problem of artificial intelligence products is to confuse the fake with the real, and then produce false information or be used for fraud. In this regard, the method of "internal and external double identification" can be used to prevent the misleading of artificial intelligence products. In terms of user visibility, publishers should use the same tips as advertisements for videos and voices with deep forgery technology, and set up eye-catching "artificial intelligence products" signs, which can effectively prevent the public from misunderstanding. In terms of machine visibility, we should set up a "digital watermark" for artificial intelligence products, that is, embed the digital identification information that cannot be recognized by the naked eye into the images, audio and video generated by artificial intelligence. Whether it is copied, reprocessed or re-read by artificial intelligence, the software can identify it as an artificial intelligence product, thus preventing the wrongly generated content from interfering with the deep learning of other artificial intelligence products and leading to misinformation.
Third, the risk level of artificial intelligence technology cannot be exaggerated.
From the perspective of human historical development, new technologies have generally made society safer, not more dangerous. Guns are much more dangerous than wooden sticks, but the murder rate in the industrial age is much lower than that in the Stone Age. In his book "Good Angels in Human Nature-Why Violence Decreases", an American scholar introduced how human beings gradually evolved from a predatory community to a safe and orderly society. Even in the worst years of war, the mortality rate in modern western countries is less than 1/4 of the average mortality rate in non-state society. From a big historical perspective, although the course of human development is troubled by terrorism, human society is still an increasingly safe society. There are always risks in technology, but advanced technology makes human beings more civilized.
On the one hand, artificial intelligence technology should be carefully regulated.
The purpose of preventing technical risks is to better develop technology, and we cannot blindly pursue zero risk at the expense of development efficiency. Digital technologies such as the Internet, big data and artificial intelligence are neutral in themselves. While improving the efficiency of social development, digital technology has also become a tool to spread pornography and implement telecom fraud. You can't give up eating just because you choke. Laws and regulations deny the malicious use of technology, not the technology itself. Laws and regulations never prohibit new technologies, and they will not preset barriers before industrial development. In the early stage of industrial development, sensitive legislation should not be implemented.
On the other hand, even risky technologies are relatively neutral.
Good people can use guns to get rid of violence, and bad people can also use guns to kill people and rob goods. In order to prevent the risk of accidents caused by cars, the Red Flag Act of that year stipulated that when cars passed by, someone should wave a red flag to warn passers-by. As a result, Britain, the first country to invent the steam engine, surrendered its status as a big automobile country to Germany and the United States. The same is true of digital technology. Laws and regulations should firmly encourage technological innovation while preventing major risks caused by technology, and should not stop because of possible risks.
In short, the application of artificial intelligence will indeed bring new risks, but "evil of human nature" cannot be simply understood as "evil of technology". Laws and regulations should promote artificial intelligence technology as a good tool to benefit the country and the people while doing a good job of bottom-line supervision.