For artificial intelligence to begin approximating human know-how, scientists will need to create models of how people think — such as simulations of your and my actual brain. That's when the trouble may begin.
The big picture: In this future — which could be plausible within decades — we will voluntarily upload these virtual versions of our brains onto platforms like Facebook or Elon Musk's aptly named Neuralink, which may conduct experiments on them. When they do, it will be only a little removed from fiddling with the real us.
Brain uploading, or whole-brain emulation, is one way to simulate intelligence. Although the science remains well out of reach today, Ray Kurzweil, Google’s director of engineering and an AI pioneer, argues it will be possible within decades.
One way brain simulations will be used is to better personalize commercial products and services. But in getting there, companies will raise numerous Frankenstein's monster scenarios of tinkering with people's very essence.
We risk a future "in which a handful of private companies own and monetize a map of our lives, ourselves, and how we think and feel at any given moment," said Meredith Whittaker, executive director of AI Now and a research scientist at New York University.
How about this freakish thought: By understanding your virtual brain, companies could advertise exactly what you want when you want it — perhaps useful, but incredibly intrusive.
"They could know us better than we know ourselves," Whittaker told the O’Reilly AI Conference in San Francisco last Thursday.
What's happening now: Even without a perfectly accurate copy of your brain, scientists are designing systems that try to imitate how humans make decisions. Basic behavior modeling is the bread and butter of social networks like Facebook, often used to improve advertising.
For now, the profiles Facebook creates from users’ browsing habits can be laughably off-base. But data taken directly from brains says a lot more about people than inferences from their likes and shares, powering much more accurate predictions.
Companies could be compelled to share this data with the government in certain cases, Whittaker said, or could build a business selling it to employers or health insurers.
