OpenAI CEO Sam Altman testifies in ‘historic’ Senate hearing on AI safety

Sam Altman, the CEO of OpenAI and co-founder of Worldcoin, recently testified before Congress alongside IBM’s chief of trust, Christina Montgomery, and NYU professor Gary Marcus.

The Senate Judiciary Privacy, Technology, & the Law Subcommittee session represented Altman’s first official appearance before Congress, giving senators the opportunity to question the OpenAI CEO concerning his company’s views on regulation.

‘If this technology goes wrong, it can go quite wrong.’

OpenAI CEO Sam Altman admitted during a Senate hearing that his ‘worst fears’ are causing ‘significant harm to the world’ via developing tech. pic.twitter.com/RROe9DnR2U

— NowThis (@nowthisnews) May 16, 2023

Dubbed a “historic” session by Illinois Senator Dick Durbin, the proceedings focused on understanding the potential threats posed by generative artificial intelligence (AI) models such as ChatGPT and how lawmakers should approach regulation.

Altman’s comments — which were described by congressional members and fellow speaker Marcus as seeming sincere and genuine — appeared to take several Senate members by surprise.

He advocated for the establishment of a federal oversight agency with the authority to issue and revoke development licenses, stated that he believed creators should be compensated when their work is used to train an AI system and agreed that consumers who suffer harm using AI products should be entitled to sue the developer.

Altman shrugged off questions related to the recent “AI pause” letter calling for a six-month moratorium on the deployment of systems more powerful than GPT-4, the AI system underpinning ChatGPT, by stating that OpenAI had spent longer than six months evaluating GPT-4 before deployment. He said the company had no plans to deploy another model within the next six months.

Marcus, a signatory of the pause letter, admitted he agreed more to the spirit of the letter than its contents, but the NYU professor urged Congress to consider global oversight as well as federal regulation — a sentiment Altman agreed with.

Throughout the hearing, the three guest speakers aligned on most topics. This included support for privacy protections, greater government oversight, third-party auditing and how soon the United States government should seek to regulate the industry (immediately).

Senator Lindsey Graham asks the witnesses at the hearing on artificial intelligence regulation if there should be an agency to license and oversee AI tools.

All say yes, but IBM Chief Privacy & Trust Officer Christina Montgomery has stipulations: pic.twitter.com/UD7R8N7s23

— Yahoo Finance (@YahooFinance) May 16, 2023

However, one of the lone sources of discord came from IBM’s Montgomery, who disagreed with the notion that a new federal agency would be necessary to enforce regulations in the AI industry.

Her statements indicated that IBM is in favor of a surgical approach to regulation using currently existing regulatory bodies to focus enforcement on specific use cases.

While all three speakers agreed that AI could be harmful and required safety interventions, Marcus made it clear he believes nobody currently understands or can predict how harmful existing AI products can be or will become. He advocated for a cautious approach involving greater transparency.

‘We have built machines that are like bulls in a china shop: powerful, reckless, and difficult to control.’

NYU Professor Gary Marcus sounded the alarm during a Senate hearing Tuesday on the emerging threats of AI. pic.twitter.com/d2b2kdJBWK

— NowThis (@nowthisnews) May 16, 2023

The speakers also agreed with members of Congress that the U.S. needed a national privacy law similar to those in Europe. Altman, however, disagreed with the notion that consumers should be able to opt out of having their publicly available web data included in training data sets.

Altman also wouldn’t go on the record to state that OpenAI was opposed to offering an ad-based version of its GPT products. The CEO simply stated that he “wouldn’t say never.”

This, despite saying earlier in the hearing that OpenAI’s products adhered to consumer privacy standards because the company didn’t build user profiles for the purposes of serving tailored advertisements.

Citing his experience as a Stanford graduate and lawmaker working with decentralized finance and Web3 companies, New Jersey Senator Corey Booker brought up the grander privacy issue of centralization. He asked what the implications for centralization and monopolization would be for the industry, a question Marcus responded to with the grave warning that the country could be risking giving control over the public perception to a small number of actors — the leading AI companies with enough money to compete with Microsoft, Google and Amazon.

Altman, whose Worldcoin project combines a decentralized cryptocurrency asset on the Ethereum blockchain with identity authentication via iris-scanning technology, explained that OpenAI merely provided a platform and that the democratization of OpenAI’s products occurred when developers, companies and end users adapted the GPT API for “fantastic” uses.