AI Experts Propose Steps to Regulate AI at Senate Hearing

JJohn July 26, 2023 3:36 AM

AI researchers shared their insights on AI regulation at a Senate Judiciary Committee hearing, emphasizing the urgency for measures that are swift yet mindful of potential risks. Testimonies from experts included Anthropic's Dario Amodei, Yoshua Bengio, and UC Berkeley's Stuart Russell.

Balancing speed and caution in AI regulation

At a Senate Judiciary Committee hearing, leading AI researchers articulated the urgency of regulatory measures for the nascent AI industry. They emphasized the fine balance between moving swiftly to protect against AI abuse, while also ensuring that undue haste does not hamper industry development. Participants in the discussion included AI experts such as Dario Amodei of Anthropic, Stuart Russell from UC Berkeley, and veteran AI researcher Yoshua Bengio. They collectively underscored the twin risks of moving too slow or too fast in regulating AI.

Dario Amodei, Anthropic's co-founder, proposed a two-pronged approach to AI regulation for immediate consideration. Firstly, he stressed the need to secure the supply chain, indicating that there are vulnerabilities in our hardware dependencies for AI research and provision, which could be jeopardized by geopolitical factors or issues related to intellectual property and safety. The second recommendation involved the creation of a rigorous testing and audit procedure like those for vehicles and electronics. He emphasized that defining risks and dangers is necessary to establish these measures.

Bengio's safety-focused approach to AI

AI researcher Yoshua Bengio proposed limiting the access to large-scale AI models and creating incentives for security and safety. Additionally, he emphasized the importance of alignment, which ensures that AI models behave as they are intended to. He also suggested monitoring who has the ability to access the scale of hardware necessary to create such models. Bengio reiterated the need to globally fund AI safety research and promote cooperation between nations rather than competition.

Russell's human-centric regulatory suggestions

UC Berkeley's Stuart Russell offered a series of suggestions for regulating AI technology. These included the creation of an unequivocal right for individuals to know whether they are interacting with a human or a machine, the prohibition of algorithms capable of deciding to kill humans, a mandated kill switch for AI systems that break into other computers or replicate themselves, and a requirement that systems which violate these rules be withdrawn from the market.

More articles

Also read

Here are some interesting articles on other sites from our network.