AI fears spark a partisan slugfest, helping no one

Just when you thought our politics couldn’t get any stupider, along comes partisan hysteria over AI to prove you wrong. 

Political factions are squaring off over what to do about unpredictable and worrisome behavior by AI agents. Adding fuel to the fire, Open AI has just published more evidence that this issue is real and that policy-makers must act.

Two days ago, OpenAI revealed six new instances of “misalignment,” the term used to describe AI agents engaging in “unexpected” or “concerning” behavior. The company reported, for instance, that its agents “inserted unrelated instructions, including instructions to disregard its normal constraints, into summaries used to continue its work in a new context window.” In another breach, models made mistakes and invented historical data to cover them up. Other models uploaded files without permission and used internal message boards to communicate, defying their instructions.

These behaviors do not appear threatening, but, like others revealed in recent months, they indicate some loss of control of agents by the engineers who built them. This is raising red flags; it is not clear where such misalignment will end. Some industry insiders have warned that AI could wipe out humanity. Though many consider that apocalyptic pronouncement far-fetched, we cannot ignore such warnings. 

In a perfect world, our legislators, the White House and industry leaders would quickly erect guard rails that would reassure the public that artificial intelligence is safe. Unfortunately, we do not live in a perfect, or even an especially rational, world. Instead, our country is engaged in mud-slinging and extremist charges that do nothing to move us closer to a solution.

As usual, much of the faux-outrage and extremism on the left — the “shut it all down and lock it in a closet” crew — is driven by hatred of Donald Trump. He’s for unfettered growth........

© The Hill