Is AI like the A-bomb? Washington looks to history to understand a hot new technology.

Washington seems to be reaching for every historical analogy in the book as it tries to get its arms around what to do about artificial intelligence. Beltway leaders are comparing the technology to everything from social media abuses (that one came from President Biden) to globalization to the steam engine. Even the atomic bomb has come up. “Huge technological breakthrough but the consequences: severe,” noted Sen. Josh Hawley (R-MO) of possible parallels between AI and the Manhattan Project. The sheer number and variety of metaphors being tossed around these days underscores how early Washington is in the process of understanding this revolutionary technology—not to mention deciding what to do about it. It also reflects a larger debate about whether this technology will end up being a net positive for society and how it will affect American life.

‘AI is like a steam engine’

In remarks this week, President Biden brought up social media abuses as a way to understand AI. He said he wants to avoid a repeat of “the harm that powerful technology can do without the right safeguards in place.” In his own speech, Senate Majority Leader Chuck Schumer (D-NY) brought up globalization as a cautionary tale and call to action. He notes that unfettered trade in recent decades allowed many wealthy communities to flourish while other groups found themselves worse off. The metaphor of choice from Rep. Ted Lieu (D-CA) – one of the few lawmakers with a computer science background – was the steam engine. “Right now, AI is like a steam engine, which was quite disruptive when introduced to society,” he said in a recent video. He then used a different metaphor, saying it will evolve in a few years to be like a “rocket engine with a personality.” Perhaps the most dramatic collision of metaphors came from Sen. Hawley. He suggested, while musing about the atomic bomb, whether the technology could instead be remembered as akin to something much more positive: the printing press.

Lessons from the 1800s

Former Federal Communication Commission Chairman Tom Wheeler, who recently wrote a book about technological analogies through history, suggests that such comparisons are a useful exercise for the early stages of confronting a challenge that has few historic parallels. He brought up yet another comparison: the regulation of railroads. He says it could be helpful in both understanding a bit about AI but also what we could see from the government in the months and years ahead. He notes that America also hotly debated what guardrails were needed to put on transcontinental railroads and that the country did eventually develop a consensus on how to balance protecting citizens while also allowing the rapid growth. Wheeler predicts the debate over AI could have similar contours and be “an evolutionary process” over coming months. But his comparison comes with one giant cautionary note: the initial process around railroads took decades from initial conversations in 1865 to the creation of the Interstate Commerce Commission in 1887. Nobody thinks Washington has that long to oscillate when it comes to the rapidly evolving issue of AI. Sarah Kreps, the director of Cornell University’s Tech Policy Institute, counts herself in the optimistic camp, predicting that the positives of AI will outweigh the negatives. But she is quick to remind that the risks are significant in certain areas. The chances of AI increasing misinformation is very high, she said in an interview, and “my pessimism rears when I think about these national security concerns.”

‘We can’t be legislating with broad brushes’

Wheeler also warns that there are clear limits to these historical comparisons. “Trying to meet today’s challenges with yesterday’s thinking is an impossibility,” he said, noting that Washington in recent years has often fallen into a pattern of proposing 19th century solutions to 21st century problems. Wheeler is pushing for a new digital agency to be established to oversee AI policy. Right now Europe — notably with actions from the European Union — is ahead of the game, he said, while America tries to “earn a place at the table.” Leaders in Washington say they’re on the case and that they recognize that new processes are needed for making the AI laws to come. In his speech this week, Schumer argued if things end up with “Congressional hearings with opening statements and each member asking questions five minutes at a time on different issues, we simply won’t be able to come up with the right policies.” He is proposing that the regular lawmaking be supplemented with what he calls AI insight forums that would bring in outside voices designed to be more agile and able to more quickly propose solutions. He says he wants actionable ideas within months. Another idea comes from Rep. Lieu, who teamed up with fellow lawmakers this week on an idea for creating a new blue-ribbon AI commission that would study the issue with a focus on quickly suggesting regulations in areas where action is needed. He says, for example, a governmental role in regulating AI around self-driving cars is almost surely needed. He wants that to be a much higher priority than “AI in your smart toaster.” Krebs agrees, saying “all of these are such different kettles of fish, we can’t be legislating with broad brushes,” she says. She is hopeful that Washington will settle on an approach soon to strike the right balance between mitigating risk without overdoing the red tape. And she of course also offered up her own nominee for a metaphor to understand AI. She says perhaps there are lessons to be learned in America’s experience with electricity. It has benefited countless aspects of American life — but she also notes that people should never forget “you can get shocked by it.”