Rethink Priorities is an independent, non-partisan, non-profit 501(c)3 think tank centeredon policy analysis. As an extension of its work on policy analysis, Rethink Prioritiesregularly conducts polling and analyses of public attitudes. Rethink Priorities is not fundedby any candidate or political party committee and does not poll on behalf of any politicalcandidate or party
The Monmouth poll also described AI as 530 Divisadero St. PMB #796, San Francisco, California 94117 www.rethinkpriorities.org 27 the creation of computers that can think for themselves, which may be more conducive to imagining hostile or frightening agentic AI. If accurate, our ndings indicate the US public is not as pessimistic about AI as some other polls might suggest. However, it is also plausible that this is dependent on exactly how AI is construed by the respondent. Male respondents were more likely to have a positive expectation for AI than female respondents, and both Democrats and Republicans had more positive expectations than
id: 20237e657dead486152f1139adc91bb8 - page: 27
530 Divisadero St. PMB #796, San Francisco, California 94117 www.rethinkpriorities.org 28 Independently aliated respondents. Female respondents were clearly more negative and also more likely to endorse neutrality. For political aliation, the dierence between groups seemed largely due to Independents being more likely to pick the Neutral option than for them to have reliably negative expectations. 530 Divisadero St. PMB #796, San Francisco, California 94117 www.rethinkpriorities.org 29
id: a52e02ad9abbaa09601767ea28023dc9 - page: 28
Associations among AI attitudes In addition to estimating public opinion, we also assessed some relationships of potential interest between the AI-related outcomes (see the Methodology section for a description of these models). For support of a pause on certain kinds of AI research, we found that higher expectations of harm, more worry, and greater expectations of extinction in the next 50 years from AI were positively associated with support for a pause. Those who reported a clear expectation that AI would do more good or much more good than harm were especially likely to oppose a pause.
id: 9f4991423ef7ebd405b91899ae97e152 - page: 30
530 Divisadero St. PMB #796, San Francisco, California 94117 www.rethinkpriorities.org 30 With respect to believing that AI should be regulated in a manner similar to how the FDA oversees food and drugs, the expectation of AI doing more good than harm was again associated with heightened disagreement with regulation. Worry and the belief in extinction from AI tended again towards being positively associated with support for regulation, but these associations were not robust. With respect to worry itself, we found that the anticipation of greater harm than good was strongly associated with worry more so than both the expectation of extinction and the
id: f9572ca9cc9906319a85c45c71696eba - page: 30