AI Index: Public vs. Insiders
The Stanford 2026 AI Index shows public sentiment toward AI has fallen while experts remain noticeably more optimistic, with the share of people saying they're 'mostly excited' dropping from 30% to 22% and 'hopeful' from 27% to 18%. (thenextweb.com) The Index also flags rising pressure around compute, emissions and public trust in AI systems. (spectrum.ieee.org)
Stanford’s 2026 AI Index says people are cooling on artificial intelligence even as experts stay markedly upbeat. (hai.stanford.edu) The report, released April 13 by Stanford’s Institute for Human-Centered Artificial Intelligence, draws on surveys across 30-plus countries and separate polling of United States adults and artificial intelligence experts. In the United States sample, the share saying they felt “mostly excited” about artificial intelligence fell to 22% from 30%, and “hopeful” fell to 18% from 27%. (hai.stanford.edu) The split widens when people are asked where the technology is headed. On jobs, 73% of experts said artificial intelligence will affect work positively, versus 23% of the public; on the economy the gap was 69% to 21%, and on medical care it was 84% to 44%. (hai.stanford.edu) Americans also expect more disruption than experts do. Nearly 64% of the public said artificial intelligence will lead to fewer jobs over the next 20 years, while only 5% said more jobs; experts were less pessimistic, with 39% expecting fewer jobs and 19% expecting more. (hai.stanford.edu) Stanford frames the gap as part of a broader mismatch between what the technology can do and what institutions can manage. The report says governance, evaluation, education systems, and the data needed to track impacts are all struggling to keep up with deployment. (hai.stanford.edu) That strain shows up in adoption numbers too. Stanford says generative artificial intelligence reached nearly 53% population-level adoption within three years, faster than the personal computer or the internet. (hai.stanford.edu) The build-out behind that adoption is getting bigger and harder to ignore. Stanford’s research chapter says global artificial intelligence compute capacity has risen 30-fold since 2021, and IEEE Spectrum’s review says it has grown more than threefold a year since 2022. (hai.stanford.edu) (spectrum.ieee.org) The same chapter says Grok 4’s estimated training emissions reached 72,816 tons of carbon dioxide equivalent in 2025, while artificial intelligence data center power capacity rose to 29.6 gigawatts, roughly comparable to New York state at peak demand. (hai.stanford.edu) Stanford also says the most capable systems are becoming less open about how they are built. Its research chapter says training code, dataset sizes, and parameter counts are increasingly withheld, and a separate Stanford transparency study published in December 2025 found disclosure had declined from 2024 levels. (hai.stanford.edu 1) (hai.stanford.edu 2) The public mood is not uniformly negative. Stanford says 59% of global respondents in 2025 still believed artificial intelligence products and services offer more benefits than drawbacks, up from 55% in 2024, even as 52% said the technology makes them nervous. (hai.stanford.edu) The result is a two-track picture of artificial intelligence in April 2026: faster uptake, bigger systems, and heavier infrastructure demands, alongside a public that is less excited than the people building and studying the tools. (hai.stanford.edu)
