The US government releases first guardrails on AI

Artificial intelligence is officially a national priority for the United States Government. 

On October 24th, President Joe Biden signed the first national security memorandum detailing how the Pentagon, intelligence agencies, and other national security institutions should use and protect AI. 

As White House national security adviser puts it, “We have to get this right, because there is probably no other technology that will be more critical to our national security in the years ahead.”

This 38 page order is meant to put the government on the right track to doing so, despite the fact that AI development has been driven almost entirely by the private sector. 

As you’d expect with a national security document, the order mainly discusses how “the United States Government must harness powerful AI, with appropriate safeguards, to achieve national security objectives.”

A lot of ink is spilled on the importance of developing an international AI governance framework, but even more is used to describe the importance of “ensuring that the United States remains the most attractive destination for global talent and home to the world’s most sophisticated computational facilities.”

Simply put, the US government now has a vested interest in attracting the top technical talent in the world and building as much computational power as possible. So, they will now personally recruit top AI talent, invest in AI semiconductor infrastructure (which, to their credit, they are already doing with the TSMC plant in Arizona), and distribute “computational resources, data, and other critical assets for AI development to a diverse array of actors that otherwise would lack access to such capabilities — such as universities, nonprofits, and independent researchers.”

The document also spends a lot of time on the need for the US government to be involved in testing frontier AI models, especially as it pertains to nuclear, cybersecurity, biosecurity, and chemical weapons risk. So, for example, the Energy Department is now required to submit an annual report to the President on AI’s “radiological and nuclear risk” and the AI Safety Institute is required to develop their own benchmark for evaluating AI’s capabilities and limitations. 

Overall, the order is a typical government document on technology, much more focused on the big picture than granular details. However, the very fact that the government is taking such an interest in AI is extremely telling of just how good (and rapidly advancing) the frontier models are getting. And if former OpenAI researcher Leopold Aschenbrenner is to be believed, this national security interest shows that superintelligence is very near. 

Whether or not that is true remains to be seen. For now, all that is certain is that the US government is committed to turning the country into an AI powerhouse.