“My aspiration for the field is to achieve consistency so we can fully realize the benefits of this technology, ” stated Terra Terwilliger, the director of strategic initiatives at Google DeepMind, the company's AI research division. She made these comments at Fortune’s Most Powerful Women Summit on Wednesday, alongside Noosheen Hashemi, CEO and cofounder of January AI, Aidan Madigan-Curtis, general partner at Eclipse Ventures, and Dipti Gulati, CEO for audit and assurance at Deloitte & Touche LLP US. The panel discussed SB-1047, the much-debated California bill that would have mandated that developers of the largest AI models adhere to specific safety testing and risk mitigation protocols. Madigan-Curtis remarked that if organizations like OpenAI are indeed constructing models as powerful as they claim, there should be legal responsibilities to ensure they are developed safely. “That’s essentially how our system functions, isn't it?There’s a balance of push and pull, ” Madigan-Curtis explained.

“One of the fears of being a doctor is the potential for being sued for medical malpractice. ” She referenced the now-abandoned California bill’s “kill-switch” clause, which would have mandated that companies implement a mechanism to disable their model if it were misused, such as in the creation of weapons of mass destruction. “If your model is being utilized to instill fear in a specific population, shouldn't we be able to deactivate it or prevent its use?” she inquired. DeepMind's Terwilliger advocates for regulation that differentiates various levels of the AI stack, emphasizing that foundational models carry different obligations than applications that utilize them. “It’s essential for us all to engage with regulators to clarify these distinctions, ensuring that we have stable and logical regulations, ” she stated. However, Terwilliger insisted that the impetus for responsible development shouldn't solely come from the government. Even amidst changing regulatory landscapes, the responsible construction of AI is crucial for the technology’s long-term acceptance, she explained. This responsibility spans all aspects of AI development, from ensuring clean data to implementing safeguards for the model. “I believe that taking on responsibility can be a competitive edge, and understanding how to maintain accountability at all levels will truly make a difference, ” she concluded.