The warning came as Altman planned a six-continent tour to talk about AI with policymakers and the public.
ChatGPT OpenAI CEO Sam Altman told US lawmakers that government regulation of artificial intelligence is “extremely important” because of the potential risks it poses to humanity .
Altman took advantage of his appearance on the U.S. Senate Judiciary Subcommittee on Tuesday to urge Congress to bring new ideas to big technology despite deep political divisions that have blocked legislation aimed at regulating the Internet for years. appealed for the imposition of rules.
“If this technology doesn’t work out, it could go very wrong,” Altman, who has become a global face of AI, said at the hearing.
“OpenAI was founded on the belief that artificial intelligence has the potential to improve almost every aspect of our lives, but it also creates serious risks,” he said. said, but given concerns about disinformation, job security and other hazards, “we believe that to de-risk an increasingly powerful model, governments Regulatory intervention will be essential.”
Altman proposed the creation of a US or global agency with the power to license the most powerful AI systems, “de-license them and ensure compliance with safety standards.”
Altman’s San Francisco-based startup shot to prominence late last year after releasing ChatGPT, a free chatbot tool that answers questions with compelling, human-like responses. .
However, the initial concern about how students can use ChatGPT to cheat assignments is that the latest “generative AI” tools mislead people, spread falsehoods, infringe copyright protection, and some expanded to broader concerns about its ability to interfere with the work of
Lawmakers underscored their deep concern over the development of AI, with a prominent senator holding hearings on the Capitol to read texts written by bots in a computer-generated voice that closely resembled his own voice. .
Senator Richard Blumenthal said, “If you were listening at home, you might have thought that voice was mine and that it was coming from me, but in reality it was not mine. He said.
Artificial intelligence technologies are “not just research experiments.
“What would have happened if I had asked that, and if it offered content that would support the surrender of Ukraine or the surrender?” [Russian President] What about Vladimir Putin’s leadership? ”
global action needed
While recognizing the enormous potential of AI tools, Altman suggested the US government might consider a combination of licensing and testing requirements before releasing more powerful models.
He also recommended greater labeling and global collaboration in setting rules around technology.
“I think the US should lead here and do things first, but we need something global to be effective,” he added.
Republican Sen. Josh Hawley said the technology could have significant implications for elections, jobs and national security, and said the hearings were “an important first step in understanding what Congress should do.” rice field.
Blumenthal pointed out that Europe has already made considerable progress on the AI law due to be voted on by the European Parliament next month.
The European Union bill is a voluminous piece of legislation that could ban biometric surveillance, emotion recognition and certain police AI systems.
Co-founded by Altman in 2015 with the backing of tech billionaire Elon Musk, OpenAI has evolved from a non-profit research institute with a safety-focused mission into a business. The company’s other popular AI products include the image generator DALL-E. Microsoft has invested billions in the startup and integrated its technology into its own products such as the search engine Bing.
This month, Altman also plans a world tour of capitals and major cities on six continents to talk about AI with policymakers and the public.
At the Capitol, politicians also heard warnings that the technology was still in its infancy.
Another panelist, New York University professor emeritus Gary Marcus, said, “We’re going to see more genies wanting more bottles.”
“We don’t have machines that can actually… improve themselves. said.
