The Scoop
Proponents of US legislation that would formally establish a public AI resource and data hub, providing cheaper access to the technology, are making a last-ditch effort to push the plan through during this congressional session.
While the CREATE AI Act has bipartisan support, some conservative lawmakers are concerned about the funding needed to provide compute power to academics, nonprofits, and researchers. The bill would establish the National AI Research Resource to give access to more powerful AI tools that only the largest companies like Google and Microsoft currently enjoy.
While Republicans tend to shy away from supporting new government funding for academic programs, the pitch for this plan has been popular because of the recent national focus on beating China as the top AI developer.
Democratic Sen. Maria Cantwell is leading the effort to get it approved as part of a legislative package before the end of the year, according to people with knowledge of the matter. It would need to overcome conservative opposition, and have congressional leadership make it a priority to push the plan through. Cantwell’s office didn’t immediately respond to a request for comment.
The bill is also co-sponsored by Republican Senators Todd Young of Indiana and Mike Rounds of South Dakota. Conservatives have increasingly called for creating an AI “Manhattan Project,” a government-led program aimed at developing the most advanced technology ahead of China.
Earlier this week, 77 institutions, from large tech companies like Amazon to universities in Indiana, Louisiana, and Texas, signed an open letter asking Congress to pass the bill.
Russell Wald, executive director of the Stanford Institute for Human-Centered AI, has been making the rounds in DC this week, arguing that the program is critical to US national security because it will help train the next generation of AI researchers, feeding the innovation pipeline that has kept the country in the top spot of global AI ecosystems.
“Time is of the essence,” Wald told Semafor. While US companies are ahead of China on the most advanced AI models, he says China is putting more money into its academic ecosystem, which could help the country make its own AI breakthroughs. “China is supplying their people with the resources. We are not. So we need to take a minute to think about that,” he said.
In this article:
Know More
While the biggest tech companies now employ many of the best AI researchers, Wald and others have argued that the next breakthroughs could still come from academia, where the lack of commercial incentives allows unfettered exploration.
“One of the most important things about this broad base of public sector investment is that it is what industry builds on,” Wald said.
Earlier this week, Wald was with Fei-Fei Li, an AI pioneer and co-director of the institute, imploring legislators to vote on the bill. Li was an advocate who helped the idea get off the ground, and Wald has been working on it for five years.
It’s unclear whether the plan will pass during the lame duck session — there’s a version sitting in the House and another in the Senate — but it could be attached to a more urgent bill, considering its bipartisan support.
Reed’s view
There are a lot of brilliant minds doing AI research for tech companies these days, and many of them are making millions of dollars per year.
Still, academia is a draw for brilliant people because of the freedom to try anything. As much as some tech companies try to recreate the leeway of academia, the private sector just can’t replace it.
And, as ChatGPT has turbocharged the productization of AI, big companies are becoming less and less academic, as Semafor’s reported.
But in order to continue making AI breakthroughs, academia needs access to a wealth of data, and compute power at its disposal.
One way to look at this is if the private sector — or a government-backed “Manhattan Project” — could more swiftly carry the AI ball across the Artificial General Intelligence goal line. Maybe, but it probably makes sense to plan for another possibility: It will take more basic research and creativity to find the breakthroughs required.
Another possibility is that the private sector is able to scale up the capabilities of AI models, but is unable to make the technology reliable and trustworthy enough to release into the world.
In that case, we’d need a breakthrough to make it safe, which might require the kind of research that often comes from academia.
In essence, the bill can’t hurt the US. Funding basic scientific research, like the intent behind the CREATE Act, has always paid off for the US economy in the long run.
Room for Disagreement
The AI Now Institute of New York University has argued that while the NAIRR proposal “claims to democratize access to these resources as a way of contending with the concentrated power of the companies that control AI infrastructures, it would in reality almost certainly work to expand and entrench the power and control these companies have.”