The people who make artificial intelligence platforms have a tendency to make bold predictions about what this technology will be able to do. If you believe some of the biggest names in the industry, AI will cure cancer, transform education, make us live longer, help establish life on other planets, and will match humans' cognitive ability within just a few years.
Everyday Americans aren't buying the hype. A new poll by Pew Research Center found that regular people are much more pessimistic about the future of AI than the experts who work with it every day.
The survey presented the same set of questions to two groups: about 5,000 members of the public and a selection of nearly 100 experts who work with, research or produce some of the popular AI models that have taken over the tech world in the past few years.
What they found is that regular people are far more likely to believe that AI will do more harm than good. More than half of experts (56%) said that AI will have a positive impact on the U.S. over the next 20 years. Just 17% of non-experts agreed, half as many who said it will have a negative impact.
There was also a huge gap between the groups when it came to how AI will affect them personally. More than three-quarters of experts believe that AI will benefit them in the future. Only 24% of regular Americans said the same, and 43% said they think AI will harm them.
The survey then dug deeper into specific topics, once again finding wide variance between how experts and non-experts viewed AI. Experts were far more optimistic about the role that AI will play across a wide range of critical areas — including the economy, job performance, health care, education, the arts, the environment and personal relationships.
Within both groups, women were much more skeptical about AI than men. The difference was especially stark among experts. Just 36% of female AI experts believe the technology will have a positive impact on the country, compared to 63% of their male counterparts.
A significant share of both groups believes that AI will replace a lot of human workers, though that view is much more prevalent among non-experts. When asked which types of jobs are most at risk, there were some interesting points where the two sides were in agreement and others where there were large gaps.
Experts and regular Americans both see cashiers as the most “replaceable” positions. They also generally agree on how AI will affect employment for journalists, software engineers and therapists. However, experts were almost twice as likely to say that AI will lead to fewer jobs for truck drivers, while the public envisions AI as a bigger threat to teachers, musicians and doctors.
There are some important areas where experts’ views match those of the general public. Strong majorities of both groups are highly concerned about people getting inaccurate information from AI. Both also worry about racial and gender bias that may be built into AI systems.
“It's mainly straight white men or men of color who are really investing and excited about these technologies, but … [when] people start to get replaced by technology, it’s always going to affect those underrepresented groups first,” an unnamed AI expert told Pew.
The public and experts also agree that current laws around AI don’t go far enough to prevent its potential harms and lack confidence that our government will find a way to regulate AI effectively. There is also skepticism in both camps that the companies making AI can be trusted to develop and use the technology responsibly.
Despite these few areas of agreement, it’s clear that experts have strikingly different views on the potential and risks of AI. Pew’s report doesn’t attempt to explain why, but there are some of the data offers clues.
Experts, of course, are much more familiar with AI and use it more frequently. In question after question, larger shares of non-experts said they were not sure what impact the technology will have; that suggests that many people simply haven’t given the topic nearly as much thought as those who work with it every day.
Experts also appear to grossly overestimate how important AI is to the typical American. Nearly 80% said they believe people in the U.S. use AI “almost constantly” but just 27% of non-experts said that was true in their own lives.