海角大神

Part-time student, full-time congressman. Rep. Don Beyer is learning AI.

AI technology is a rising focus of legislation in Congress. Rep. Don Beyer is heading back to college for a master's in machine learning, hoping to stay informed ahead of legislative debates.

First Look

Don Beyer鈥檚 car dealerships were among the first in the U.S. to set up a website. As a representative, the Virginia Democrat leads a bipartisan group focused on promoting fusion energy. He reads books about geometry for fun.

So when questions about regulating artificial intelligence emerged, Mr. Beyer took what for him seemed like an obvious step, enrolling at George Mason University to get a master鈥檚 degree in machine learning. In an era when lawmakers and聽Supreme Court justices sometimes concede they don鈥檛 understand聽emerging technology, Mr. Beyer鈥檚 journey is an outlier, but it highlights a broader effort by members of Congress to educate themselves about artificial intelligence as they consider laws that would shape its development.

Frightening to some, thrilling to others, baffling to many: Artificial intelligence has been called a transformative technology, a threat to democracy, or even an聽existential risk for humanity. It will fall to members of Congress to figure out how to聽regulate the industry聽in a way that encourages its聽potential benefits聽while mitigating the worst risks.

But first they have to understand what AI is, and what it isn鈥檛.

鈥淚 tend to be an AI optimist,鈥� Mr. Beyer聽told The Associated Press following a recent afternoon class on George Mason鈥檚 campus in suburban Virginia. 鈥淲e can鈥檛 even imagine how different our lives will be in five years, 10 years, 20 years, because of AI. ... There won鈥檛 be聽robots with red eyes聽coming after us any time soon. But there are other deeper existential risks that we need to pay attention to.鈥�

Risks like massive job losses in industries made obsolete by AI, programs that retrieve聽biased聽or聽inaccurate results, or聽deepfake images, video, and audio聽that could be leveraged for聽political disinformation,聽scams, or聽sexual exploitation. On the other side of the equation, onerous聽regulations聽could stymie innovation, leaving the U.S. at a disadvantage as聽other nations look to harness the power聽of AI.

Striking the right balance will require input not only from tech companies but also from the industry鈥檚 critics, as well as from the industries that AI may transform. While many Americans may have formed their ideas about AI from聽science fiction movies聽like聽鈥淭he Terminator鈥澛爋r 鈥淭he Matrix,鈥� it鈥檚 important that lawmakers have a clear-eyed understanding of the technology, said Rep. Jay Obernolte, R-Calif., and the chairman of the House鈥檚 AI Task Force.

When lawmakers have questions about AI, Mr. Obernolte is one of the people they seek out. He studied engineering and applied science at the California Institute of Technology and earned an M.S. in artificial intelligence at UCLA. The California Republican also started his own video game company. Mr. Obernolte said he鈥檚 been 鈥渧ery pleasantly impressed鈥� with how seriously his colleagues on both sides of the aisle are taking their responsibility to understand AI.

That shouldn鈥檛 be surprising, Mr. Obernolte said. After all, lawmakers regularly vote on bills that touch on complicated legal, financial, health, and scientific subjects. If you think computers are complicated, check out the rules governing Medicaid and Medicare.

Keeping up with the pace of technology has challenged Congress since the聽steam engine聽and the聽cotton gin聽transformed the nation鈥檚 industrial and agricultural sectors. Nuclear power and weaponry is another example of a highly technical subject that lawmakers have had to contend with in recent decades, according to Kenneth Lowande, a University of Michigan political scientist who has studied expertise and how it relates to policy-making in Congress.

Federal lawmakers have created several offices 鈥撀爐he聽Library of Congress, the聽Congressional Budget Office, etc. 鈥撀爐o provide resources and specialized input when necessary. They also rely on staff with specific expertise on subject topics, including technology.

Then there鈥檚 another, more informal form of education that many members of Congress receive.

鈥淭hey have interest groups and lobbyists banging down their door to give them briefings,鈥� Mr. Lowande said.

Mr. Beyer said he鈥檚 had a lifelong interest in computers and that when AI emerged as a topic of public interest he wanted to know more. A lot more. Almost all of his fellow students are decades younger; most don鈥檛 seem that fazed when they discover their classmate is a congressman, Mr. Beyer said.

He said the classes, which he fits in around his busy congressional schedule 鈥撀燼re already paying off. He鈥檚 learned about the development of AI and the challenges facing the field. He said it鈥檚 helped him understand the challenges 鈥撀燽iases,聽unreliable data聽鈥撀燼nd the possibilities, like improved cancer diagnoses and more efficient supply chains.

Mr. Beyer is also learning how to write computer code.

鈥淚鈥檓 finding that learning to code 鈥撀爓hich is thinking in this sort of mathematical, algorithmic step-by-step, is helping me think differently about a lot of other things 鈥撀爃ow you put together an office, how you work a piece of legislation,鈥� Mr. Beyer said.

While a computer science degree isn鈥檛 required, it鈥檚 imperative that lawmakers understand AI鈥檚 implications for the economy,聽national defense,聽health care, education, personal privacy, and intellectual property rights, according to Chris Pierson, CEO of the cybersecurity firm BlackCloak.

鈥淎I is not good or bad,鈥� said Mr. Pierson, who formerly worked in Washington for the Department of Homeland Security. 鈥淚t鈥檚 how you use it.鈥�

The work of safeguarding AI has already begun, though it鈥檚 the executive branch leading the way so far. Last month, the White House unveiled聽new rules聽that require federal agencies to show their use of AI isn鈥檛 harming the public. Under an聽executive order聽issued last year, AI developers must provide information on the safety of their products.

When it comes to more substantive action, America is playing catchup to the聽European Union, which recently聽enacted the world鈥檚 first significant rules聽governing the development and use of AI. The rules prohibit some uses 鈥撀爎outine AI-enabled facial recognition by law enforcement, for one 鈥撀爓hile requiring other programs to submit information about safety and public risks. The landmark law is expected to serve as a blueprint for other nations as they contemplate their own AI laws.

As Congress begins that process, the focus must be on 鈥渕itigating potential harm,鈥� said Mr. Obernolte, who said he鈥檚 optimistic that lawmakers from both parties can find common ground on ways to prevent the worst AI risks.

鈥淣othing substantive is going to get done that isn鈥檛 bipartisan,鈥� he said.

To help guide the conversation lawmakers created a new AI task force (Mr. Obernolte is co-chairman), as well as an AI Caucus made up of lawmakers with a particular expertise or interest in the topic. They鈥檝e invited experts to brief lawmakers on the technology and its impacts 鈥撀燼nd not just computer scientists and tech gurus either, but also representatives from different sectors that see their own risks and rewards in AI.

Rep. Anna Eshoo聽is the Democratic chairwoman of the caucus. She represents part of California鈥檚 Silicon Valley and recently聽introduced legislation聽that would require tech companies and social media platforms like Meta, Google, or TikTok to identify and label AI-generated聽deepfakes聽to ensure the public isn鈥檛 misled. She said the caucus has already proved its worth as a 鈥渟afe place鈥� place where lawmakers can ask questions, share resources, and begin to craft consensus.

鈥淭here isn鈥檛 a bad or silly question,鈥� she said. 鈥淵ou have to understand something before you can accept or reject it.鈥�

This story was reported by The Associated Press.