Congress Is Racing to Regulate AI. Silicon Valley Is Eager to Teach Them How.

The Washington Post / Cat Zakrzewski, Cristiano Lima

17:40 JST, June 19, 2023

When Rep. Jerry McNerney took over the House caucus dedicated to artificial intelligence in 2018, his colleagues were not all that interested.

“There was difficulty getting members to attend our meetings,” the California Democrat said, estimating that a typical session would draw about 18 or 20 lawmakers from the 435-person body.

McNerney’s counterparts across the Atlantic felt the lack of enthusiasm, too. Brussels was expanding efforts to regulate the technology in 2020, but when Dragos Tudorache, a Romanian member of the European Parliament who co-leads AI work, contacted the U.S. caucus, there seemed to be little political momentum.

That’s changed. The overnight success of AI-powered ChatGPT has triggered a frenzy among Washington lawmakers to draft new laws addressing the promise and peril of the burgeoning field. When Tudorache visited Washington last month, he witnessed a tumult of activity around AI and attended a bipartisan briefing with OpenAI CEO Sam Altman.

“There is a different mood,” Tudorache said in an interview.

But tackling the swiftly evolving technology requires a sophisticated understanding of complicated systems that back AI, which sometimes confound even experts. Congressional salary caps that pale in comparison to Silicon Valley’s sky-high paychecks make it difficult to retain staff technologists, putting lawmakers at a disadvantage in getting up to speed – a goal that has become increasingly urgent as the European Union has leaped ahead of Washington, advancing robust AI legislation just this week.

To catch up, members of Congress and their staffs are seeking a crash course on AI. With Senate Majority Leader Charles E. Schumer (D-N.Y.) preparing to unveil a plan Wednesday for how Congress could regulate AI, lawmakers are suddenly crowding into briefings with top industry executives, summoning leading academics for discussions and taking other steps to try to wrap their heads around the emerging field.

Lawmakers’ gaps in technical expertise have provided an opening for corporate interests. Executives motivated to develop AI without hindrance are flocking to Washington, eager to lend a hand in lawmakers’ education – and influence policy. Schumer said his office has met with close to 100 outside experts, including “CEOs of companies who do AI, scientists, AI academics, leaders in the industry of many different viewpoints, and critics of AI” – among them Microsoft president Brad Smith and Tesla CEO Elon Musk.

This charm offensive has left some consumer advocates uneasy that lawmakers might let the industry write its own rules – which some executives are outright recommending. In an interview this spring, former Google CEO Eric Schmidt argued that the industry, not the government, should be setting “reasonable boundaries” for the future of AI.

“There’s no way a non-industry person can understand what is possible. It’s just too new, too hard. There’s not the expertise,” Schmidt told NBC. “There’s no one in the government who can get it right. But the industry can roughly get it right.”

Other industry leaders are taking a different tack, blitzing Congress with their vision for how Washington should regulate their companies. Altman in May had private meetings and a dinner with lawmakers, where he demonstrated – to their amusement – how ChatGPT could write a speech for them to deliver on the chamber floor. Smith has given legislators a lesson on the technical stack that underpins generative AI models like ChatGPT, including computing infrastructure and applications. And Smith recently unveiled his blueprint for AI regulation at a speech in Washington attended by half a dozen lawmakers.

The stereotyped view that Congress doesn’t understand technology – bolstered by high-profile gaffes in key tech hearings – is “outdated,” Smith said, adding that he is “optimistic” about Congress’s ability to keep pace with AI advances.

Regular briefings have imparted a more formal education. Senate and House leaders have hosted AI discussions with MIT professors, where they reviewed the basics of how AI works and examined challenges with the technology, including how it can exacerbate existing biases.

At a Tuesday briefing with MIT professor Antonio Torralba organized by Schumer’s office, some lawmakers asked basic questions, including how AI learns and where it gets data, said Sen. Jacky Rosen (D-Nev.), a former computer programmer who left the session early.

“They are putting a lot of time and effort into coming up to speed on AI,” said Aleksander Madry, a MIT professor who spoke at a briefing in April arranged by House Speaker Kevin McCarthy (R-Calif.). Madry has since gone on a professional leave and is working at OpenAI.

Sen. Ted Cruz (R-Tex.) has expressed skepticism about these efforts, suggesting that his colleagues’ tech acumen was irredeemably deficient.

“To be honest, Congress doesn’t know what the hell it’s doing in this area,” Cruz said, donning ear buds as he video-conferenced into a Politico tech summit. “This is an institution [where] I think the median age in the Senate is about 142. This is not a tech-savvy group.”

Sen. Mark R. Warner (D-Va.), who previously worked as a venture capitalist, brought in researchers and industry leaders to speak to senators after Schumer’s all-member briefing. His guests included a mix of experts, including Microsoft’s chief scientific officer, Eric Horvitz, Center for Security and Emerging Technology executive director Dewey Murdick and deputy national security adviser Anne Neuberger, according to Warner spokeswoman Rachel Cohen.

“Lots of us are all on different paths of our learning curve,” Warner told reporters Tuesday.

The uptick in AI briefings and strong attendance is a major shift for Congress, where a handful of members – some of whom hold degrees in computer science – have long struggled to capture the attention of their peers. Congress hosted its first hearing on AI in 2016, according to Cruz, who said he chaired the session. House lawmakers launched an AI caucus in 2017, and their Senate counterparts launched a similar initiative in 2019.

The rise of generative AI has finally awakened interest in such efforts. Rep. Mark Takano (D-Calif.) said AI is going to impact “every jurisdiction of Congress,” and argues that lawmakers need to respond by reviving the Capitol’s tech think tank, the Office of Technology Assessment, which lawmakers defunded during partisan battles in the 1990s. Takano plans to introduce a bill next month to fund the office, along with Sen. Ben Ray Luján (D-N.M.), who sits on the Commerce Committee.

“What is missing in Congress is a repository of expertise that is more in an anticipatory mode, that has quicker turnarounds, that can deliver responses more quickly,” Takano said. “We want to have expertise that is not tainted or connected to commercial interests.”

Some argue that concerns about the lack of technical expertise on Capitol Hill have been overblown, saying lawmakers have already introduced bills that could address most issues with generative AI, including data protection and algorithmic audit bills.

“Congress’s job is not necessarily to know the ins and outs and nuts and bolts of every single technology that they regulate,” said Anna Lenhart, who worked on tech policy for Rep. Lori Trahan (D-Mass.). “Their job is to understand the impact of technology on society, the risks and the benefits.”

Lawmakers can seek tech assessments from the Government Accountability Office and Congressional Research Service. Zach Graves, the executive director of the Foundation for American Innovation, said GAO’s resources have made gains in recent years, resulting in better preparation for tech hearings, such as the one with Altman.

“They clearly did a lot more of their homework,” Graves said.

Still, some worry that the recent flurry of corporate lobbying on AI has pushed lawmakers uncomfortably close to the industry they’re aiming to regulate.

Unlike clashes with the CEOs of Facebook and Google, lawmakers’ chummy hearing with Altman was a reflection of how effective intimate events, like his private dinner, have been, said Sarah West, the managing director of the AI Now Institute and a former senior adviser on AI at the Federal Trade Commission.

West said executives like Google’s Schmidt are fueling the perception that AI is too difficult for Congress to grasp.

That, she said, is “a convenient narrative that positions accountability out of the hands of the people that the public has vested it in – and into the hands of the industry that is benefiting.”