At the Reagan National Defense Forum last month, United States Commerce Secretary Gina Raimondo gave seemingly conflicting statements on the purpose of AI regulation, her stance on the U.S. relationship with China, and the continued development of artificial intelligence. Rather than focus on enabling U.S. development of AI, she threatened companies with lost revenue due to regulations while saying the United States should maintain its AI lead. She compared the U.S. relationship with China with that of the Soviet Union while claiming she doesn’t want “tension.” This comes after her department was greatly empowered by Biden’s recent executive order on AI—an important role that requires trust and a level head.
Raimondo’s statements came with a jibe directed at tech companies who lost money to her capricious regulations. “I know there are CEOs of chip companies in this audience who are a little cranky,” she said, “[be]cause they’re losing revenue. Such is life”. Not only is it unbecoming that of all government officials, the Commerce Secretary would be this actively hostile to industry, but that hostility also risks creating an uncertain environment for AI investment in the United States—a grave error for an industry Raimondo correctly cites as one of the most important for our national security.
Secretary Raimondo, who heads the department that has already once banned the export of chips designed to meet her requirements, then complained that it was “not productive” for Nvidia to respect export controls again by creating new chips below the latest thresholds. These thresholds, designed to restrict extremely high-performance chips, are primarily objective metrics of compute power measured in FLOPS, or “floating point operations per second.” The Commerce Department restricts chips with performance above these thresholds, indicating that those under it are deemed acceptable for export by the United States. Despite this, the strategy of the Commerce Department does not seem to be setting this threshold where they want it to be — or they simply don’t have the understanding to develop consistent thresholds and are setting reactionary regulations. Raimondo claimed that if manufacturers design a chip too close to the threshold—even if it’s still below it—“I’m going to control it the very next day.” This statement hinted at a potential export ban on Nvidia’s upcoming H20 chip, designed in good faith to comply with current regulations.
Such a ban would constitute enormous waste, both of design time and manufacturing, but that doesn’t seem to faze Raimondo, who boasts that “Every time I take an action, it denies them revenue.” Deny them the revenue it has, with the impacts of one regulation on the U.S.-based Nvidia estimated at $400 Million. That’s money that, were it not for the government’s intervention, could be reinvested into furthering chip development and increasing the U.S.’s lead in this technology. It’s difficult to ascribe any motivation to this destructive ambiguity from regulators short of active hostility to business from the Commerce Secretary.
Secretary Raimondo is not only inconsistent on export policy but also on foreign relations as a whole. At the same fireside chat, just minutes after recommending “Cold War” tactics on export controls, Raimondo turned to discuss ties with China, saying, “We have no interest in tension”—not quite the message sent by Cold War rhetoric. Raimondo phrased her export bans as “denying an entire country” access to technology, saying, “We’re going to continue to go in that direction.” This denial of technology to China also denies technology to smaller U.S. companies that can’t afford to train their Large Language Models but can use large “open weights” models released by Chinese researchers. While this may not be a major factor in the current day, with top open-weight models like Mistral and Llama 2 coming out of Western nations, any blow to open-source AI stifles innovation and benefits Sam Altman’s OpenAI, which will have less competition in the space of very large models like ChatGPT.
Raimondo cites an agreeable goal for this policy, stating, “We’re a couple years ahead of China, no way are we gonna let them catch up.” It is, indeed, essential for the United States to retain its lead in this vital technology over China, however useful their contributions may be. But not being caught requires continuing development. Raimondo also railed against the “effective acceleration” (“e/acc”) movement, mainly composed of software and AI engineers who believe in advancing AI development. She said, “There’s a view in Silicon Valley, you know, this ‘move fast and break things,’ effective acceleration. We can’t embrace that with AI, it’s too dangerous.” If Raimondo truly believes this technology is too dangerous, it’s all the more critical to maintain our lead over China—and if it’s not, then we shouldn’t slow down the development of a technology that is already improving outcomes in fields like healthcare.
Perhaps if Raimondo believes in maintaining our lead in AI, she should spend less time railing against the U.S. companies and engineers and leave them room to invest more time and money into staying ahead — something best done without vague threats of regulation and lost revenue hanging over the heads of American companies.
About the Author
Dylan Dean is a software engineer with degrees in Electrical and Computer Engineering and an advocate for the decentralization of emerging technologies.
Image: Lev Radin / Shutterstock.com.