ChatGPT’s Sam Altman tells Congress to stiffen AI rules to avoid ‘significant harm to the world’
OpenAI CEO Sam Altman called on Congress to implement regulations to address serious risks posed by artificial intelligence during a closely-watched appearance Tuesday on Capitol Hill.
Altman, whose Microsoft-backed firm created the immensely popular ChatGPT, admitted his worst fear was that advanced AI technology could “cause significant harm to the world” without proper guardrails in place.
“If this technology goes wrong, it can go quite wrong and we want to be vocal about that,” Altman said at a hearing of the Senate subcommittee on privacy, technology and the law. “We want to work with the government to prevent that from happening.”
Altman appeared before members of Congress for the first time as lawmakers step up calls for increased scrutiny and regulation of AI technology. Prominent critics, including billionaire Elon Musk and the “Godfather of AI” Geoffrey Hinton, have warned of various potential risks, including major job losses, the spread of misinformation and election interference.
Congress could police rapid advancements in AI through several methods, including the adoption of licensing and testing requirements for AI “above a threshold of capabilities” as well as standardized safety requirements and global collaboration on guidelines, Altman added.
Companies should also be held responsible for the safety of their AI products before release, said Altman, who touted OpenAI’s rigorous testing process for ChatGPT.
“GPT-4 is more likely to respond helpfully and truthfully and refuse harmful requests than any other widely deployed model of similar capability,” Altman said. “However, we think regulatory intervention by governments will be critical to mitigate the risks of increasingly powerful models.”
The US should take a leading role in AI oversight to ensure future products are developed with democratic values in mind, Altman added.
ChatGPT has already established a massive base of more than 100 million users since its launch late last year. The free service has wowed the public with its ability to generate human-like responses to an endless variety of prompts, even as critics have called out its propensity to spit out false information.
Sen. Richard Blumenthal (D-Conn.), the panel’s chair, opened the hearing with an audio recording of his own voice discussing the potential pitfalls of AI – only to reveal that he hadn’t actually said the words.
Instead, Blumenthal had prompted ChatGPT to write a statement and used an audio “deepfake” trained on his own speeches.
The Democrat expressed concern that AI could enable bad actors to trick the public with fake messages by public figures. Blumenthal noted the possibility of AI deepfake audio that could portray him or others as backing Russian president Vladimir Putin or calling for Ukraine’s surrender
“Weaponized disinformation, housing discrimination, harassment of women and impersonation fraud, voice cloning, deepfakes – these are the potential risks despite the other rewards,” Blumenthal said.
Blumenthal floated the possibility of establishing “scorecards and nutrition labels” that would rate AI products released to the public on their level of “safety and trustworthiness.
Sen. Josh Hawley of Missouri (R-MO), the panel’s ranking member and a prominent critic of Big Tech firms, argued that advancements in artificial intelligence could yield massive benefits for humanity, such as the creation of the printing press – or bring major negative consequences, such the invention of the atomic bomb.
“Artificial intelligence will be transformative in ways we can’t even imagine, with implications for Americans’ elections, jobs, and security,” Hawley said. “This hearing marks a critical first step towards understanding what Congress should do.”
Altman maintained a positive tone about AI’s future throughout the hearing, arguing it could help address “some of humanity’s biggest challenges, like climate change and curing cancer.”
He expressed optimism about the technology’s impact on the future of work, though he acknowledged an exact outcome was “difficult to predict.”
“GPT-4 will, I think, entirely automate away some jobs and it will create new ones that, we believe, will be much better,” Altman said.
