Yves here. We are taking the liberty of republishing a post by Gary Marcus, who has been one of a handful of AI experts to make early and accurate calls about the shortcoming of the technology, particularly of large language models. Marcus is urging voters to call their Congressional representatives to stop the White House plan to bar state regulation of AI. Please take up his warning and tell your legislators in no uncertain terms to oppose this “get out of jail free” card for powerful AI players who can and should be regulated just as other potentially hazardous consumer and business services are. Please also circulate this post to AI concerned or skeptical friends and colleagues.
There have already been many examples of chatbots encouraging suicide and violence even before getting to the harm that can come from garden variety inaccurate information.
By Gary Marcus, professor emeritus of psychology and neural science at New York University. Originally published at his website
I write with some urgency.
Part I: Recent History
I warned here of a proposed Federal law, introduced by Congress, with the intention of cutting down the rights of individual states to regulate AI. Many other organizations spoke out as well.
At that point, there was a happy ending at the last minute. At the last minute Congress, led by Republican Senator Marsha Blackburn decided, essentially unanimously, to not enact such a monstrosity.
Good on them. A moratorium on Statewise regulation of AI was and is, a terrible idea, especially in a world in which there is no coherent Federal response to AI; children, in particular, are left incredibly vulnerable, to garbage like this:
As I put it in earlier today on X, re a call for federal AI standards in lieu of state standards only makes sense if there are (sensible) federal standards, but Congress has not put anything serious forward to a vote.
The actor Joseph Gordon-Levitt spoke eloquently about the same yesterday (full link here), referring a proposed last minute effort to stick a moratorium on statewise regulation into the National Defense Authorization Act:
Part II: Breaking news
But even in the last hours, the stakes have risen. I regret to inform you that, per an important scoop from The Information, The White House, clearly now tight with big tech, is preparing to circumvent Congress altogether:
Call your Congress people, ASAP; don’t let the White House cut them and the states you live in at the knees. I urge you in the strongest possible terms to follow the links that Joe Gordon-Levitt supplied at https://linktr.ee/StopTheAIBan, so that states can protect their citizens from the many dangers of AI, short-term and long. Airlines have regulations, cars have regulations, bakeries have regulations; AI bots that tell kids to kill themselves don’t.
We need to change that.
