Mac Dejuran
Published 18 minutes ago: May 17, 2023 8:52 AM
OpenAI CEO Sam Altman made his congressional debut today on the Senate Judiciary Subcommittee. He launched a charm offensive to persuade lawmakers to push for a rash bill that would give artificial intelligence (the company’s first product, ChatGPT) a broad runway for rapid progress. Altman’s rosy testimony describing the social virtues of generative AI ran counter to other expert witnesses who expressed more skepticism and brought up discrimination and other unintended harms of AI. .
Altman said of the public understanding of ChatGPT and text generation AI: “People were fooled by PhotoShop for a while. Then they quickly got a better understanding of the altered images. This will be the case too, but it will be on steroids.” .”
Lawmakers who questioned Mr. Altman on Tuesday said the danger was serious. Missouri Senator Josh Hawley AI said in an opening statement that it could go one of his two routes: a new printing press or an atomic bomb.
“We may be witnessing one of the most important technological innovations in human history,” Hawley said. “At least akin to the invention of the Internet at scale.”
The hearing comes at a critical tipping point for large-scale language models like ChatGPT, as lawmakers and regulators struggle publicly to stay ahead of rapidly evolving technology. It was held. Together, Google, Altman’s Open AI, Microsoft, Meta and others are sprinting to determine who will be the biggest winners in the new AI arms race.
Sam Altman supports regulation of ChatGPT and AI in general…as long as it’s his preferred kind of law
Over the years, executives at big tech companies have learned that it is foolish to speak out against the threat of regulation. Instead, the more popular strategy followed by Armutan is to insist on supporting the kind of law they prefer. On Tuesday, Altman told lawmakers that “regulation of AI is essential,” but that such regulation must balance safety with ensuring broad access to the technology for the general public. His ChatGPT, which debuted in November, already has 100 million users, according to OpenAI.
During his testimony, Altman urged lawmakers to pursue new safety requirements that allow products to be tested before they go on sale. Altman suggested that new testing and licensing requirements for AI developers could help set a level playing field for the competition. The CEO was open to Senator Blumenthal’s recommendation to consider “AI nutrition labeling” and other transparency proposals, but still believed the benefits of AI “outweigh the risks.” said and warned. Altman says safety requirements must be flexible enough to accommodate unexpected new advances in technology.
“I think government regulatory intervention will be important,” Mr. Altman said.
Other expert witnesses, such as former New York University professor Gary Marcus, are expected to take a more cautious stance on the technology, warning of the recent rise in potentially dangerous “AI hype.” . Marcus on Tuesday urged lawmakers to approach AI safety with a deep sense of urgency and warned them not to repeat the same mistakes they made years ago when they failed to regulate social media.
“We are facing a perfect storm of corporate irresponsibility, widespread deployment, lack of proper regulation and inherent unreliability,” Marcus said at the hearing.
Marcus and Altman were joined by IBM Chef Privacy Trust Officer Christina Montgomery, who urged lawmakers not to regulate AI as a technology, but instead to regulate some of the technology’s most harmful use cases. advised lawmakers to consider
AI will ask for jobs, but how much?
All three witnesses who testified on Tuesday agreed that AI could disrupt and transform the workplace, but the scope and duration of that change are debatable. Altman said he believes his company’s products and other AI services will have a “significant impact on employment,” but said he wasn’t sure exactly how it would play out. . At the moment, Altman said, GTP-4, OpenAI’s latest large-scale language models, and other AI systems are good at completing tasks, but not very adept at completing the full job. “I believe there are far better jobs out there on the other side,” he said. He said.
Marcus went one step further and said that artificial general intelligence could threaten most jobs. But that day may be 50 years away. AI skeptics said OpenAI’s model is far from achieving artificial general intelligence.
Altman voices support for new agency to oversee AI
Altman and Marcus clashed at times during the hearings, but were united in their vision of a new government agency staffed with AI experts. A virtual agency of AI experts will be tasked with overseeing the development of the technology and setting standards for its use. In response to a question from South Carolina Senator Lindsey Graham, Altman said he supports government agencies that can grant operating licenses to AI companies and strip them of their licenses if they violate standards. .
Marcus took that idea a step further, advocating a ministerial-level organization that could deal with the damage caused by AI on a global level. Marcus said the global reach and international interest in AI will require some kind of international body to set common standards. “Probably above my salary level,” he said, as to how the organization navigates intense geopolitical tensions between countries like the United States and China. IBM’s Montgomery denied that endorsement, saying government oversight of AI systems should be left to current regulators such as the FTC and FCC.
“At this point, we don’t want to loosen regulations to address real risks,” Montgomery said. “We have existing regulators who have made it clear that they have the ability to regulate in their respective areas.”
Altman tells Congress: ‘Can’t the people sue us?
Witnesses and lawmakers alike speculated positively about the possibility that the hypothetical new agency would exercise enforcement powers, but it was far less clear what could be done to hold AI companies to account at this time. rice field. Hawley, Minnesota Senator Amy Klobuchar, and others question whether Section 230 of the Communications Decency Act, social media’s primary liability protection for user content on platforms, applies to AI-generated content. watched.
“We still don’t know what the right answer is,” Altman said. “I don’t even think Article 230 is the right framework.”
Altman seemed unclear as to what framework actually exists for consumers to hold themselves or other AI companies responsible for damages.
“Can’t people sue us?” Altman asked Holly unconvincingly.
OpenAI does not rule out future reliance on advertising
In this testimony, Altman sought to separate himself and OpenAI from social media companies such as Facebook and Instagram, which are under regulatory scrutiny as addictive, ad-driven platforms. Altman told lawmakers that OpenAI does not currently operate on an advertising-based business model and is actively trying to build models that do not specifically maximize engagement.
“There is such a shortage of GPUs that the fewer people using them, the better,” Altman said with a small chuckle. “We’re not an advertising model, and we don’t want people to use it.” [GPT-4] gradually. “
These remarks were intended to reinforce Altman’s narrative that OpenAI is an idealistic research company that puts human prosperity ahead of huge profits. But when later asked by New Jersey Senator Cory Booker whether the company promised never to pursue an ad-based business model, Altman had to warn of the difference. “I’m not saying never,” Altman said. “Some people may want to provide services, but other services do not work.”
Lawmakers will want to appear prepared for a new age of technology
As with any technology hearing, some lawmakers may appear ill-prepared and unfamiliar with the technology in question. At the same time, widespread experimentation with ChatGPT-style chatbots and the recent surge in public concern over the harm of AI have motivated at least some of the senators in attendance to investigate the issue and speak out aggressively. There are good reasons to think that Legislators on both sides of the political spectrum, at previous hearings on cryptocurrencies and social media, appeared incapable of pressuring executives about the technology’s impact and incapable of drafting meaningful legal safeguards. I try to avoid repeating things like that.
In recent weeks, about six new bills and legislation on AI have been introduced, led by Senator Michael Bennett of Colorado and Rep. Ted Liu of California. On the regulatory side, the Federal Trade Commission has issued multiple statements making clear its intent to use existing law to punish AI companies. Chairman Rina Khan earlier this month further hinted at her proactive approach to AI in a New York Times editorial titled concisely, “We Must Regulate AI.”
On the administration side, the White House has so far expressed concern about areas of potential abuse, but has invested in new AI research and has had friendly discussions about AI with executives at major tech companies. I’ve been trying to cross the middle ground.
Details from Gizmodo Australia
