Atomic Bombs, Garth Brooks, and AI: 9 Key Moments From Congress' Hearing With the Creator of ChatGPT

US politicians just can’t quite seem to get over their obsession with using AI chatbots to draft legislation and write speeches. Connecticut Sen. Richard Blumenthal continued that trend on Tuesday and kicked off the hearing with a recording of him warning about the dangers of unregulated AI. Only the recording wasn’t actually the senator but rather a voice cloning software reading a prompt generated by ChatGPT.

“This is not the future we want,” the AI-generated Blumenthal impersonator said.

The OpenAI CEO admitted early on that his, and other AI products, may bring about a slurry of unintended consequences and welcomed new regulatory standards enacted by Congress to set rules of the road for him, and maybe more importantly, his competitors.

“We think that regulatory intervention by governments will be critical to mitigating the risks of increasingly powerful [AI] models,” Altman said. Altman caveated that, however, and urged lawmakers to focus on reigning in large companies and avoid stifling smaller startups.

This week’s Senate hearing on AI at times felt as much about correcting for past tech hearing disasters as it did coming to any meaningful conclusions about AI regulations. Multiple lawmakers, including Sen. Blumenthal and New Jersey Sen. Cory Booker, admonished colleagues for dropping the ball when it came to preventing social media companies from becoming vehicles for misinformation and harmful content. AI potentially presents a case for redemption.

“Congress failed to meet the moment on social media now we have the obligation to do it on AI before the threats and risks become real,” Blumenthal said.

Missouri senator and Capitol Hill attack booster Josh Hawley believes large language models like GPT4 could evolve in one of two extremely different directions.

“We could be looking at one of the most significant technological innovations in human history,” Hawley said. “My question is what kind of innovation is it going to be? Is it going to be like the printing press that diffused knowledge and power and learning…or is it going to be more like the atom bomb.”

Others, like South Carolina Sen. Lindsey Graham echoed the atom bomb rhetoric despite all three witnesses donwplaying the more cataclysmic doomsday AI scenarios.

Altman may not have compared AI to a nuclear weapon like some lawmakers but he did acknowledge he worried his product could cause “significant harm to the world.” Those harms could present themsvles by way of mass disinformation, scams, large-scale job displacement, and a muddying of what’s considered true online.

Altman assured senators he believes AI’s “benefits greatly outweigh the risks,” but acknowledged government intervention and regulation could be critical to safeguarding against some of its worst byproducts. Pressing pause on new LLMs, as some have suggested, was quickly written off as unpractical.

Tennesse Senator Marsha Blackburn sparked an exchange over AI models’ use of copyrighted material with an unexpected example: AI Garth Brooks. The Senator told Altman she had heard an AI trained on the country singer produce a convincing adaptation of his work. Blackburn used the example to demand artists receive compensation from Altman’s company if their works helped train an AI model. Altman seemed open to the idea of compensating artists but wouldn’t land on any specific path forward.

“We lived through Napster,” Blackburn said. “That was something that really cost a lot of artists a lot of money.”

Altman acknowledges rapidly growing concerns over AI’s potential to deceive users through sophisticated chatbots and deepfakes. Though he admitted he was “concerned,” Altman said society expressed a similar type of phenomenon with Adobe’s Photoshop image editing tool.

“I think people are able to adapt quickly,” Altman said. “When Photoshop came onto the scene a long time ago, for a while people were really quite fooled by photoshopped images but then pretty quickly developed an understanding that images might be photoshopped. This will be like that but on steroids.

Both Altman and a handful of the lawmakers sitting on the Subcommittee on Privacy, Technology & Law seemed to agree on the need for a new expert-led AI agency capable of monitoring the development and deployment of AI. That proposed agency, which would stand apart from other regulatory bodies like the FTC and FCC, could, in theory, issue licenses to responsible AI companies and revoke those same licenses from companies deemed to have behaved badly. Gary Marcus, an AI researcher and ex-NYU professor testifying at the hearing, supported an AI agency and said it should work on closely with international partners to set global AI standards.

Lawmakers and experts speaking during the hearing may have agreed on some of the societal harms posed by AI but there was far less agreement on what affected users could currently do to hold OpenAI or other AI companies accountable for harms. Lawmakers and experts disagreed, for example, over whether or not chatbots should receive the same immunity projections as social media companies under Section 230 of the Communications Decency Act. When asked directly by Hawley what recourse users had, Altman seemed perplexed.

“Can’t people sue us?” Altman asked. The answer to that seemingly simple question remains concerningly unclear. “I don’t know yet, exactly what the right answer here is.”

Want to know more about AI, chatbots, and the future of machine learning? Check out our full coverage of artificial intelligence, or browse our guides to The Best Free AI Art Generators and Everything We Know About OpenAI’s ChatGPT.