Three Short Arguments in AI Policy

The Case For Open Source

Machine learning is a classic example of upfront research and development costs. It can cost tens of millions of dollars in hardware alone to train a large language model, in this case LLAMA-2. Once this model is trained, it can be adapted for a variety of purposes. Incentivizing open-source development increases efficiency in training by reducing double-spending. It lowers the barrier of entry for talented engineers without independent wealth or institutional affiliation. By making software open to public scrutiny, it makes fixing security issues and preventing unintended behavior far easier. Consequently, proactively funding open-source organizations and incentivizing existing AI organizations to open-source models benefits everyone.

The Current Market Is Neither Free Nor Efficient

Recent scandals at Google shows that outdated government policy is actively hindering the most important technological developments in AI.

From a Pirate Wires exclusive:

“Three entire models all kind of designed for adding diversity,” I asked one person close to the safety architecture. “It seems like that — diversity — is a huge, maybe even central part of the product. Like, in a way it is the product?”

“Yes,” he said, “we spend probably half of our engineering hours on this.”

Google’s Gemini paper directly references the Biden Executive Order on AI as motivating its content policies:

External groups were selected based on their expertise across a range of domain areas, including those outlined within the White House Commitments, the U.S. Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence, and the Bletchley Declaration

Creating an open and dynamic environment for AI not only involves preventing further damage, but curtailing existing destructive measures.

Everything is a ‘Dual-Use’ Model

Securing our national defense require encouraging the development of technologies with both military and civilian uses. These ‘dual-use’ technologies are ones we are already comfortable using in everyday life. A transmitter on a phone or laptop is little different than one on a remote bomb. The wires, microchips, and batteries are only slight variations of each other. The same nitroglycerine can be used as explosives or as life-saving heart medication. Cars, planes, and ships all have vast uses in military conflict, but are all crucial to the way of life of every American.

Machine learning comes in a long line of technologies that drastically improves civilian and military effectiveness. This is why policies restricting “dual-use” machine learning models are self-destructive economically and strategically. Obviously, prohibiting the civilian use of electronics, medication, or cars because they have military uses would be economically disastrous. The free and open development of machine learning technology is crucial not only to the private sector applications of machine learning, but to the national security and competitiveness of the United States military. For centuries, free and open civilian research and entrepreneurship has enabled more efficient production, research, and leadership in the military. The accumulated knowledge, organizational processes, and practical training of American civilians gained from the private sector enables efficient and effective military development. Consequently, policies which constrict and alienate private sector research directly harms national security.