New evidence on popular perception of “AI” risk
The Mozilla Foundation recently conducted a survey about AI perception. I participated after receiving an email invitation; I don’t know if they recruited elsewhere. They just released a writeup of the results today, which can be found here.
Mozilla seems to take a broad view of “AI”, with a lot of weight given to current and near-term issues and less to potential fate-of-humanity issues. This is, of course, a different perspective from many in the LW crowd.
The popular perception is nonetheless potentially relevant to AI safety because it may represent the perception of some AI capabilities researchers. It may also be relevant to predicting political intervention in AI research.
51k sample size, 67k including partial submissions (see pdf linked from writeup)
overall, respondents were optimistic (i.e. not especially worried about risks)
most optimistic demographics were 19-24 years old, male, South Americans
only 10% self-reported being “well-educated” about AI, only 4% unfamiliar with term “AI”
respondents generally interested in learning more about AI
“24% of respondents said AI will make our lives better. 41% of respondents think AI will make our lives both better and worse. Only 10% of respondents think AI will only make our lives worse”
″ Men (27%) are almost twice as optimistic as women (14%) that AI will make our lives better. Nearly half of South Americans (46%) are optimistic that AI is going to make their world better, making them the most positive region in the world. And young people 19 − 24 years old (35%) were the most likely age group to say AI will make the world better. Only 5% of this age group said they thought AI would make the world worse. ”
“Respondents aged 19 − 44 were most likely to be “very interested” in learning more about AI (52%).
Respondents 65 and over were most likely to say they are somewhat interested in learning more about AI (64%)
South Americans (61%) and Africans (63%) said they were most interested in learning more about AI.”