AI is a sharp tool, let’s not cut ourselves.

Last week on LinkedIn, I ventured that when A.I.s can “peer”-review research papers, and you no longer have to, your dean will assign you to three more committees, and maybe an extra course. My half-joking point was that AI won’t throw professors out of work. AI will cause professors to be assigned still more work!

An astute commenter replied that robot reviewers will further worsen social atomization, by depriving researchers of awareness of unpublished works, thus robbing researchers of contact with possible new ideas, colleagues and collaborators. This can slow scientific progress, even assuming the robot reviewers are “good reviewers.”

A fair point, and one that I’ve not seen in other commentaries on AIs’ impacts.

I also ran a survey on LinkedIn, asking which of the following poses the greatest threat to humanity: Malice, indifference, or mistakes on the part of the AI program itself; heat from the servers supporting AI applications; or humans making mischief with AI. The response rate was pitiful, yet the result was somewhat telling. The first two options each garnered 17% of responses, versus 71% for the third option.

The “atomization” possibility - the isolation of researchers from each other - strongly suggests the survey should have offered a fourth distinct option: The unintended consequence of marketing a well-meaning, seemingly productivity-enhancing AI application. Is that a realistic fear? Yes. Just think of the harms caused by social media.

I don’t deny the great benefits of AI and social media. Nor would I deny that the latter has worsened teen bullying, online scams, and political polarization, and robbed us of precious time hanging out face-to-face with friends. Real friends, not online friends.

The upshot? All of us should carefully but creatively imagine the downsides of using AI to cut humans out of a loop, before we do it. And professors will need to spend more time reading online pre-prints!