Sanders-Casar Ban Artificial Superintelligence Act. I [Gary Marcus] Oppose It (garymarcus.substack.com)

In response to the proposed Sanders-Casar Ban on Artificial Superintelligence, AI expert Gary Marcus has voiced strong opposition, arguing that the bill's scope is excessively broad. While Marcus acknowledges the value in pausing certain AI advancements to improve regulation and alignment, he contends that an outright ban on superhuman AI research would undermine the U.S. position in the global AI landscape and ignore current risks in AI development. He stresses that the legislation is overly focused on hypothetical future dangers while neglecting immediate concerns, which could allow authoritarian figures to exploit regulatory loopholes. Instead of a permanent prohibition, Marcus advocates for a more nuanced approach that permits the development of narrow AI tools while setting stringent criteria for projects that aim for higher intelligence and autonomy. He suggests adopting principles that would require developers to demonstrate robust alignment and oversight to an independent authority, thereby fostering responsible innovation rather than stifling it. This revised framework could help balance the urgency of safety with the need for continued advancement in AI research, ensuring that progress is made without compromising safety and ethical considerations.