Potential Adverse Factors of Artificial Intelligence Systems

2019-11-12

Stanislav Petrov may not be well-known in the field of computer science, but the story that happened to him taught a valuable lesson to artificial intelligence developers. On September 26, 1983, the Soviet defense alert system announced that the United States had launched five intercontinental missiles. As a leader in the system, Stanislav Petrov was supposed to order the launch of missiles in retaliation. However, at the critical moment, he hesitated and began to think critically. Why? At that time, although the global political situation was tense, it had not reached the point of using missiles to resolve conflicts. Moreover, if the United States had chosen to initiate a devastating war, why would they only launch five missiles? After careful consideration, this former Soviet official made a wise decision not to retaliate by launching missiles. The final result proved that he was right - the United States did not launch any missiles, and it was just a false alarm. His decision saved the world. This story tells us that in social technological algorithm systems, neglecting the ability to take action based on human reasoning in specific situations is undoubtedly a disaster. If the missile launch button had been controlled by a machine, what would the world be like today? Personifying the language of artificial intelligence systems is the true meaning of artificial intelligence, rather than just code or algorithms. Real-world artificial intelligence often has flaws, is not perfect, and lacks the universality to be widely promoted. Technological chauvinism is a term coined by Meredith Broussard, and the concept of transforming humans is an extension of this term. It means adjusting humans to fit the limitations of artificial intelligence systems, rather than adjusting technology to serve humans. This is undoubtedly a reversal of priorities. Some studies have shown that due to the perception of the objectivity of data and artificial intelligence, humans are more willing to believe in artificial intelligence and algorithms rather than trust themselves, even in low-skilled tasks and despite the potential flaws in algorithms. Many people still prefer to trust cold machine algorithms rather than believe in the superiority of human judgment. Flawed designs can lead to biased results, and the imperfection of artificial intelligence fundamentally leads to the undesirability of blind trust. In today's highly advanced technology era, many people have blind faith in data, which is reasonable to some extent because data is precise and objective. Compared to subjective human judgment, data is undoubtedly more reliable. However, it is important to clarify that although data is mostly accurate, it does not mean it is always correct. Furthermore, it is humans, not data itself, who exercise subjective agency. Even if the data is incredibly accurate, if it is used in the wrong context, the result will be disastrous.

RECENT AI NEWS

RECENT AI TOOLS