Artificial Intelligence & Machine Learning , Next-Generation Technologies & Secure Development , Standards, Regulations & Compliance
OpenAI CEO Calls for Regulation But No Pause on Advanced AI
Sam Altman Tells Senate Judiciary Panel Regulations Should Be GlobalThe chief executive of the company behind high-profile generative artificial intelligence tool ChatGPT called for global regulations, including licensing, in testimony before a U.S. congressional panel.
See Also: Achieve Machine Learning Results Faster
The United States should consider licensing and registration requirements for the development and release of powerful AI models, said Sam Altman, CEO of OpenAI. He also called for safety standards and for models to pass tests and independent audits.
His call, delivered to a Senate Judiciary subcommittee, reflects growing concern among the public and regulators alike. The Biden administration is investigating possible mechanisms to ensure an "AI accountability ecosystem" (see: Feds Call for Certifying, Assessing Veracity of AI Systems). Lawmakers are close to approving regulations for the European Union, including new obligations for large language models that power tools such as ChatGPT. China in April proposed draft regulations that would make companies responsible for model outputs.
"As this technology advances, we understand that people are anxious about how it could change the way we live. We are, too," Altman told the panel of senators.
An executive from IBM somewhat echoed Altman, telling the subcommittee that the company supports "precision regulation." Christina Montgomery, IBM's chief privacy officer, said that means regulating use cases but not the technology itself. IBM, an early leader in artificial intelligence, announced earlier this month renewed investments in the technology.
Altman and Montgomery rejected a call made in March by tech luminaries for a half-year pause in the development of advanced AI systems. "If we pause for six months, then I'm really not sure what we do then. Do we pause for another six?" said Altman. "I'm not sure how practical it is to pause," Montgomery said.
Cybersecurity researchers have spotted low-level hackers testing the boundaries of generative AI to generate malware scripts and write phishing messages. Altman's written testimony says the company's latest large language model, GPT-4, is better than its predecessor at rejecting requests to produce disallowed content.
Effective regulation will require a global approach, Altman said, and he suggested following the model of the International Atomic Energy Agency. "There are paths to the U.S. setting some international standards that other countries would need to collaborate with and be part of, that are actually workable, even though it sounds on its face like an impractical idea," he said.
The Italian data protection authority banned ChatGPT in March over privacy concerns but allowed it to resume service for Italian IP addresses in late April. The Italian agency said OpenAI implemented changes, including giving Europeans the right to have their data excluded from training algorithms and adding an age verification button.
Altman and another witness, Gary Marcus, emeritus professor of psychology and neural science at New York University, also called for the creation of a new American regulatory agency to oversee AI. Critics of big technology firms for years have called for creation of a new tech regulatory agency, a proposal unlikely to gain traction during divided government despite the support of some tech-savvy members of Congress. The idea also struggled for mainstream acceptance during Democrats' period of legislative and executive branch control during the first two years of the Biden administration.