A U.S. congressman has begun advocating for a federal division to manage the usage of synthetic intelligence, postulating a dystopian future the place AIs will make key selections 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 instrument that can be utilized to profit 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 acquired a multibillion-dollar funding from Microsoft on Monday, amid consumer reviews that it’s unveiling a paid model.)

Lieu, who earned a B.S. diploma in Pc Science from Stanford, famous that AI is now current in every part from good audio system to Google Maps. However the place AI fails, individuals will be harm: 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 solely incapable of passing laws that may regulate AI: The know-how strikes too quick, and legislators lack the required information to set legal guidelines and pointers. As an alternative, “[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 selections if it makes an error. Creating such an company will likely be a tough and large enterprise as a result of AI is difficult and nonetheless not effectively understood.”

Lieu cited different businesses, 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 an alternative, Lieu stated he would advocate the formation of a non-partisan AI Fee to supply suggestions on how such a federal company could possibly be shaped, what it ought to regulate and what requirements might apply.

The Nationwide Institute of Requirements and Expertise has already printed an AI Threat Administration Framework, a non-binding doc that Lieu proposes that authorities construct upon and add compliance mechanisms to. “We might not want to manage the AI in a wise toaster, however we must always regulate it in an autonomous automotive that may go over 100 miles per hour,” Lieu wrote.

Already, ChatGPT has virtually handed the bar examination, scoring a 50.3 % appropriate response price. (A rating of 68 % is required to go.) Artists are involved that AI artwork might threaten their very own commissions. Is it doable Lieu is hoping to manage AI earlier than it comes for his job, too?

Source link