A plurality of experts believe the United States should create a new federal agency dedicated to artificial intelligence governance, according to a survey conducted by Axios, Generation Lab, and Syracuse University.
The survey polled 215 computer science professors across 65 of the most prestigious universities in the U.S. on topics related to AI.
we polled 213 computer science professors on A.I.
— The Generation Lab (@Generation_Lab) September 5, 2023
According to the data, when asked “What is the best entity to regulate AI,” the majority of respondents answered either “new “Department of AI” government agency” (37%) or “global organization or treaty” (22%).
Only 16% answered “congress,” just two percent higher than the number of respondents who chose “irrelevant: AI cannot be regulated.” The final 10% of respondents split their answers between “the White House” (4%), “the private sector” (3%), and “none: AI should not be regulated” (3%).
The survey also contained questions about how the AI sector will affect the future of employment. The majority of respondents indicated that they would advise a young person to pursue a career in AI, engineering, and data science.
At the other end of the spectrum, 31% of the professors polled said they’d advise against seeking a career in media and 19% said the same about the arts when asked which fields young people should avoid. “None of the above” was the most common response with 42%.
When asked if the respondents think there’s “a threshold in the evolution of AI after which humans cannot take back control,” the answers were split between “no, probably not” (41%), “yes, probably” (35%), “no, definitely not” (19%), and “yes, definitely (6%).
The overall sentiment of the experts seemed in juxtaposition with the general public and business leaders. Where the latter tends to poll bombastically about the potential for AI technology to rapidly change the economic and employment landscape in the near future, 73% of the professors said they believed that AI will be capable of performing less than 20% of tasks that humans do today at or above human-level.