The survey found that the majority of AI researchers believe there is a five percent chance that AI itself will lead to the extinction of humanity or other “extremely bad consequences.”

The study also found that opinions are divided on when machines will surpass humans in all tasks, with 10 percent predicting it will happen in 2027 and half predicting it will happen in 2047.

It’s not that simple

Researchers highlight significant disagreements and uncertainties within the AI ​​community. Differing views on whether AI progress should be accelerated or slowed underscore the difficulty of predicting the future of advanced AI.

A five percent probability of risk, although not definitive, is considered an important indicator of danger perception. Katya Grace of the Machine Intelligence Research Institute in California noted:

This is an important signal that most AI researchers do not think it is implausible for advanced AI to destroy humanity.

Aware of the difficulty of predicting the future, the authors of the study emphasize the importance of taking into account the opinions of experts based on knowledge and experience in this field. Researchers say such predictions need to be supported by data from a variety of sources, including computer hardware trends, artificial intelligence capabilities, economic analysis and the opinions of forecasters.

In the short term, AI researchers are focusing on issues such as deepfakes, manipulating public opinion, creating dangerous viruses, and artificial intelligence systems that allow individuals to achieve prosperity at the expense of others. As the US presidential election approaches, the potential for artificial intelligence to distort reality and influence public opinion continues to be a major focus.