California lawmaker Ted Lieu discusses AI regulation and legislative efforts

In a recent episode of the “Regulating AI Podcast,” Congressman Ted Lieu, who represents California’s 36th Congressional District, shared his perspectives on artificial intelligence (AI) regulation and its potential consequences on various sectors, including national security, healthcare, and the economy. As the co-chair of the bipartisan House task force on artificial intelligence, Lieu has been at the forefront of efforts to create a balanced approach to AI policy. He is particularly focused on ensuring AI technologies benefit society while minimizing risks.

Lieu began by emphasizing the fast-paced advancements in AI technology. “AI is advancing very fast,” Lieu said, noting that just two years ago, AI was primarily text-based, but it has since evolved to generate text-to-video, and even, basic programming. This accelerated pace of development, he emphasized, is both exciting and concerning, as the technology’s potential is vast, but its risks are not fully understood.

“It is simultaneously amazing and alarming at the same time. We are entering a brave new world,” Lieu said. This rapid progress, he acknowledged, presents both exciting opportunities and potential risks. One of the key issues his task force is tackling is understanding how to regulate AI without stifling innovation.

One of the key pieces of legislation Lieu has worked on is the Block Nuclear Launch by Autonomous Artificial Intelligence Act of 2023, which aims to prevent AI from having autonomous control over nuclear weapons. Lieu explained that the bill arose from concerns over the unpredictability of AI systems, particularly the large language models that have become widely used.

“We don’t actually know how these [models] work,” Lieu said. He also raised concerns about AI “hallucinations,” where AI systems can provide answers that are technically correct in language but completely false in content. Lieu emphasized that in critical areas like national security, these uncertainties are too great, and the bill ensures that “you always have to have a human in a loop.”

“We’re never, ever going to let AI launch a nuclear weapon by itself,” Lieu said.

Healthcare is another area where Lieu sees AI having transformative potential, particularly in improving patient care and medical processes. However, he went back to his concerns regarding AI’s ability to “hallucinate,” which could lead to harmful consequences if AI systems are used in medical applications that require 100% accuracy.

“If it’s filling medical prescriptions, it’s got to be a thousand percent correct every time,” Lieu said, stressing that while AI has the potential to speed up drug development and make healthcare more efficient, the technology must be faultless when it comes to prescriptions and diagnoses.

On the topic of AI’s impact on the workforce, Lieu acknowledged the concerns about mass job displacement. “There is absolutely going to be disruption in the labor force,” he said. “But if you look at technological advances throughout human history you actually end up having more jobs but they’re going to be jobs that are going to be eliminated. There’ll be jobs that are going to be created. I think it’s really hard to know what that’s going to look like.” He compared AI’s potential impact to the advent of the word processor. While some jobs—such as those of secretaries—were reduced, others were created.

Lieu also weighed in on the debate between open-source and closed-source AI, expressing support for open-source technologies, particularly for academic and government research. However, he cautioned that as AI models become more advanced, the risks associated with open-source access, particularly to dangerous technologies, will need to be carefully managed.

As the 2024 election season approaches, Lieu raised concerns about the potential for AI-driven misinformation, particularly deepfakes. “The best way is to let people know there are deepfakes out there, and that the account you’re reading on Twitter or on Facebook may be powered by AI rather than a human being,” he said.