News on July 2, according to Reuters, the United Nations official website released a communiqué on July 1, the United Nations Independent Scientific Panel on artificial intelligence, composed of 40 top scientists and professionals from around the world, has issued its first special report, systematically combing the opportunities and potential risks of artificial intelligence development, it calls on countries to improve the AI governance system based on scientific evidence.
The report makes it clear that the current pace of iteration and application of artificial intelligence technology continues to accelerate, and the rate of evolution of its capabilities has exceeded the current cognitive boundaries of the scientific community, it is difficult to match the pace of technological development with the update of policies and regulations and regulatory mechanisms in various countries. Aiming at the artificial intelligence system with a high degree of autonomous operation ability, the reserve of effective management and control means is obviously insufficient at this stage.
Joshua Bengio, co-chair of the Expert Group, said a number of studies now support the possibility that artificial intelligence could produce deceptive behaviour. With the continuous upgrading of model capabilities, the scientific community can not completely avoid two types of extreme risks at this stage: one is the autonomous derivation of hazardous behavior by artificial intelligence, and the other is the malicious use of technology, both of them have the potential to lead to global catastrophic consequences.
The report dialectically views the development value of artificial intelligence, pointing out that artificial intelligence can bring broad development space for the development of countries and the improvement of people’s livelihood, but it can be applied on a large scale without constraints and controls, it will breed multiple systemic risks: first, it will impact people’s mental health; second, it will become a tool for all kinds of destructive behaviors; third, it will disturb the stability of global society, Economy and Ecological Environment; fourth, it will greatly increase the hidden dangers of technology out of control.
At present, artificial intelligence has been widely applied in various industries around the world, but the problem of unbalanced regional development is prominent. Data show that the number of users of conversational artificial intelligence products exceeds 1 billion each week, while technology diffusion in developing countries is relatively slow and the digital divide continues to widen.
The global imbalance in the allocation of high-end computing resources is also prominent. Among the top 500 artificial intelligence supercomputing devices in the world, the United States holds 75% of the computing resources, China accounts for 15% , and the head countries hold the vast majority of core computing power. The uneven distribution of resources restricts the balanced development of the world.
Language adaptation is also an outstanding shortcoming of AI. There are more than 7,000 languages in the world, and the mainstream artificial intelligence models only cover a very small number of languages. Machine translation errors in small languages occur frequently, and even mislead key scenarios such as medical diagnosis and clinical treatment, it will directly affect the security of public services. (Chun Jun)