A U.S. congressman has begun advocating for a federal division to manage using synthetic intelligence, postulating a dystopian future the place AIs will make key choices and autonomous weapons roam America.
Rep. Ted Lieu (D-CA) authored an opinion piece in The New York Occasions on Monday, arguing that AI has emerged as a strong device that can be utilized to learn humanity — or deceive it, and worse. In reality, the instance he cited, reproduced above, wasn’t written by Lieu, however by ChatGPT, the AI chatbot developed by OpenAI. (OpenAI obtained a multibillion-dollar funding from Microsoft on Monday, amid person experiences that it’s unveiling a paid model.)
Lieu, who earned a B.S. diploma in Laptop Science from Stanford, famous that AI is now current in all the things from sensible audio system to Google Maps. However the place AI fails, individuals will be damage: The editorial factors out {that a} driver blamed Tesla’s self-driving mode for an eight-car pileup on the San Francisco Bay Bridge.
In response to Lieu, Congress is just incapable of passing laws that may regulate AI: The know-how strikes too quick, and legislators lack the required data to set legal guidelines and pointers. As a substitute, “[w]hat we’d like is a devoted company to manage AI,” Lieu wrote. “An company is nimbler than the legislative course of, is staffed with consultants and may reverse its choices if it makes an error. Creating such an company might be a tough and large enterprise as a result of AI is difficult and nonetheless not properly understood.”
Lieu cited different companies, such because the Meals and Drug Administration (FDA) as proof that the federal government might nonetheless regulate a brand new and rising know-how. However, he stated, the method couldn’t occur in a single day. As a substitute, Lieu stated he would advocate the formation of a non-partisan AI Fee to offer suggestions on how such a federal company may very well be shaped, what it ought to regulate and what requirements might apply.
The Nationwide Institute of Requirements and Expertise has already revealed an AI Danger Administration Framework, a non-binding doc that Lieu proposes that authorities construct upon and add compliance mechanisms to. “We could not want to manage the AI in a wise toaster, however we should always regulate it in an autonomous automobile that may go over 100 miles per hour,” Lieu wrote.
Already, ChatGPT has nearly handed the bar examination, scoring a 50.3 p.c right response fee. (A rating of 68 p.c is required to go.) Artists are involved that AI artwork could threaten their very own commissions. Is it potential Lieu is hoping to manage AI earlier than it comes for his job, too?