Leading Public-Interest Technologist Sees National Research Resource as a Potential Foundation for an “AI Public Option”
As a chorus of transatlantic public interest groups calls for governments to build their own bedrock artificial intelligence systems, the Harvard Kennedy School’s Bruce Schneier says the National Artificial Intelligence Research Resource backed by key U.S. policymakers could lay the necessary groundwork.
"It’s a start, and [could] serve as a foundation for an AI Public Option," Schneier told Inside AI Policy referring to the NAIRR, a pilot for which is included in the Oct. 30 executive order on artificial intelligence.
The NAIRR has also been highlighted in a series of closed-door AI "insight forums" hosted by Senate Majority Leader Charles Schumer (D-NY) who has said there was agreement with top Republicans to spend at least $32 billion establishing it over the next five years.
The idea that the public sector should provide AI systems that can be used to generate all manner of applications to benefit society—not just those that would enrich the handful of powerful companies currently in control of the technology—is being explored by individuals from certain startup companies as well as groups like Public Knowledge in the U.S. and Chatham House in the United Kingdom.
Schneier is also a contributor to the new Public AI Network teasing out the associated policy details.
"A Public Option means building AI systems such as foundational large language models that would serve as an alternative to corporate-controlled AI," he said. "Like public roads and a postal system, a public AI option can guarantee universal access to the technology that is fast becoming fundamental for participation in the economy."
The idea has also taken hold within the UK’s Labor Party, with Labor for the Long Term, proposing the government spend billions to build a "Great British Cloud," and "BritGPT."
In the U.S., while some have said the NAIRR, as described in the executive order, represents an excellent approach to that kind of industrial policy, others are concerned that a plan the National Science Foundation produced for implementing the NAIRR would continue to rely on the major commercial cloud service providers who are primarily building foundational AI models for data storage and computing power.
Schneier has previously said establishing Public AI needn’t rely on governments "owning and operating the entire AI supply chain," putting resources toward building data centers and special, super expensive, computer chips. He sees value in building Public AI to set an example for how to responsibly deploy the technology in a variety of policy areas.
It "could set an implicit standard that services offered by private entities must surpass," he said. "Widely available public models and compute infrastructure would yield numerous benefits to the U.S. and to broader society. It would provide a mechanism for public input and oversight on the critical ethical questions facing AI development, such as whether and how to incorporate copyrighted works in model training, how to distribute access to private users whose insatiable appetites for AI integrations may outstrip cloud computing capacity, and how to license access for sensitive applications ranging from policing to medical use."
"It would serve as an open platform for innovation, on top of which researchers and small businesses as well as mega-corporations could experiment with novel training approaches and new user-facing applications," he said. "An AI Public Option, administered by a competent and accountable public agency, would offer greater guarantees about the availability, equitability, and sustainability of AI technology for all of society than would exclusively private AI development."
Categories: Articles