That’s what the headline on the WSJ article claims. The headline misleads, as the article itself goes on to say.
Experts call…tendency of machines to excessively agree with and flatter their users AI sycophancy. Vance chose spiritual terms. “I think that’s kind of satanic.”
Certain behaviors of AI can be satanic; that does not make AI satanic.
The larger issue for me, though, is what AI gets used for. It certainly can be abusive—or satanic, in the extreme—if it’s used wrongly, and misusing it can be done subtly enough that the user might not recognize the abuse. Vice President JD Vance (R) cited a friend who used an AI package for marital advice and got back validation of his own selfishness. Children have used AI packages for advice and have been led to suicide.
Those are uses of the tool, though. AI also has been used to good effect for business moves, to identify patterns in complex and/or noisy data, to point ways toward disease controlling chemistry.
Even where AI sycophancy leads to failure, even egregious failure, though, it’s not that AI is misbehaved or satanic, it’s that its human programmers have programmed the sycophancy into it through flawed programming and flawed testing.
AI in itself is just another tool, neither good nor bad intrinsically. It’s necessary to discriminate between the tool and what the tool gets used for. Build the data centers, and get training in how to use it. Then take care in its use, just as is necessary for any other tool, whether a power saw or a firearm.
