Will AI improve your life? Here’s what 4,000 researchers think

Scientists working on artificial intelligence are more confident than the public that the technology will benefit people

Credit: Andrey Rudakov/Bloomberg via Getty

Researchers working at the forefront of artificial intelligence (AI) are much more optimistic than members of the public about the future of AI, reveals a survey of 4,260 scientists in the field: 54% think that the technology will bring more benefits than risks, compared with 13% of the UK public (see ‘Views on AI risks vary by country’).

But the researchers share the public’s concerns about the technology’s role in disinformation, data use and cybercrime. In the survey, 77% of researchers and 68% of the public said that AI makes misinformation a problem, and 65% of researchers and 71% of the public expressed issues with tech companies using people’s personal data without consent.

That means researchers’ opinions run counter to some governments’ plans to implement a mechanism for people to opt out of having their data used for AI training, says Cian O’Donovan, the study’s lead author and an innovation specialist at University College London. For instance, the UK government is proposing such a policy for copyright holders.

To regulate effectively, “we need to do an enormous amount of research, for instance to understand how AI systems can be applied to make cyber defenses and information ecosystems more robust”, says Robert Trager, director of the Oxford Martin AI Governance Initiative at the University of Oxford, UK.

The study, published¹ on 1 April on the preprint server Zenodo, surveyed researchers globally, including in the United States, India and China, and compared their answers with responses from the UK public previously collected by the UK Office of National Statistics.

Risks versus benefits

When asked how AI could have a positive impact on people’s lives, 75% of researchers think the technology will increase people’s access to learning, and 57% think it will improve access to health care (see ‘How will AI change your life?‘).

The most striking finding is that fewer than one-third of AI scientists think that the technologies should be developed as quickly as possible, says Trager. “They seem to want a more considered approach to development to mitigate risks.”

How people’s personal data are used to train AI models particularly concerns researchers. Millions of people’s books and scientific papers have been pirated and used to train AI models, raising concerns about intellectual property. Only 25% of AI researchers say that firms should be allowed train their models on publicly available data. And almost half say that people should give explicit permission for AI companies to use their data in training — a higher proportion than members of the public (see ‘Data concerns’).