Posts
Last night the U.S. issued an export control directive for the two most advanced AI models publicly available - Mythos 5 and Fable 5.
This made these models a controlled technology that may not be made available as products to other nations, or even the foreign passport holding staff members of the company that makes these models (Anthropic). That includes nations allied with the USA.
I wasn’t surprised to see this coming. However I was surprised at how fast it has come.
The US’s export restrictions are not the first shot in this war. Restrictions on chips and expertise already existed that had an underlying impact on nations’ pace of AI development. AI companies have already attempted to prevent rivals using their commercial models to train their own.
However this new shot is the most blatant and lifts the global AI Cold War to a new level.
I have advocated for sovereign AI capability for over five years. AI isn’t like other consumer technologies that can be freely traded for cash globally, or even restricted products that nations trade with their friends and keep out of the hands of bad agents and hostile nations.
Sufficiently advanced AI has the potential to quickly change the global balance of power.
As such nations need their own sovereign capability, or face the risk of becoming the new ‘have nots’ of the global economy. Here in Australia, we just became one of those ‘have nots’.
We won’t remain competitive in this new Cold War by clinging to an ally’s shield, or by buying technology from other nations. AI doesn’t only project military power, it projects commercial and social power, allowing the holder of the most advanced technologies to outthink and outmaneuver it’s adversaries AND its allies to create lasting economic, social and military domination over other nations.
The nation, or company, with the best AI technology isn’t simply able to exploit it to win favourable deals and tilt wars in its favour. It is able to create a lasting advantage by developing even smarter AI models built at an accelerating pace by the AIs it has created.
There’s a runaway effect. Even putting aside the still real and deep concerns with AI - from power and water use (a competitor for human activities); to the ever present risk of developing AI that doesn’t align with human goals, but is too intelligent for us to grasp or oppose this, whether or not we define it as sentient - we’re reached the point where nations must compete to build sovereign AI capability that can support their nation to remain relevant and sovereign.
This is a turning point for the world. Most of us won’t notice or accept that it has occurred until long after its effects have been felt. However at some point future histories, whichever entities are recording them, will note that when commercial AI models because a sovereign export risk, the world started on a new path to a new and unknowable future.
Will the world come together to develop collective AI models that must be made available to all on an equitable footing? Such that we can all benefit from their prospects and work together to safeguard the technology against the known risks.
Or will do the world do what it has always done, and divide into competing factions, each attempting to build and control the most valuable resource in the world today - AI capability? Expending our other precious resources to gain advantage and prevent hostile actors from creating their own. Including proscribing the travel and choices of the people capable of building these models, until they are no longer required and the AI models construct their own more powerful replacements.
Australia will have to decide which side it stands on soon. Half-measures, as we’ve seen so far, paying lip service to AI’s importance without committing to domestic capability will only harm us more in the long-run.
