Explained: The Artificial Intelligence Race is an Arms Race

February 2, 2020 Topic: Security Blog Brand: The Buzz Tags: Artificial IntelligenceInnovationRoboticsMilitary

Explained: The Artificial Intelligence Race is an Arms Race

Whoever wins it will have an advantage in every conflict around the world.

Having declared victory, the developers of Pluribus pulled its AI system from distribution on the grounds that they “don't want to hurt the online poker ecosystem.”

If you believe that to be the only reason then, as Allison would say, I have a bridge to sell you. This is because poker mimics something else that Chess and Go do not—life.

Artificial Intelligence and Global Primacy

So how does an AI system that plays poker at superhuman levels translate to ultimate victory in great power competition?

To answer, reflect on this claim: The only difference between a game of uncertainty and real-life is the degree of resolution.

A positive example could be designing a city. Your objective is to maximize human wellbeing and economic production. Many things you can control: placement of highways, schools, public transportation, utilities, parks, shopping complexes; plus there is a degree of uncertainty: emerging technologies, demographic change, disruptions to industry and entertainment.

By specifying these known and unknown criteria, an advanced AI system can teach itself to formulate the perfect city, running trillions of simulations in pursuit of that goal to achieve superhuman levels of urban design.

But this humanist function is not the primary driver of AI advancement. Rather, it is the great power rivalry between China and the United States.

To illustrate how this plays out, imagine that you are China’s leader. Your strategic objective over the next three decades is to replace the United States as the dominant power in the Indo-Pacific and then the wider world.

You know many details about your adversary, including how many planes, ships, submarines, and soldiers that the United States presently fields, along with their specifications and capabilities to varying degrees of confidence. You know the contents of published budget forecasts and force development plans.

There is also uncertainty: classified American capabilities and projects, shifting alliances, unforeseen international crises and so on.

With this mountain of information there will be one optimal way for the Chinese to invest to maximally exploit America’s weaknesses while mitigating their own shortcomings and risks. Human military planners will not hold a candle to an AI system that has gamed this out hundreds of trillions of times.

In this domain, AI even has an advantage relative to poker. Poker professionals live and die in the gladiatorial arena, constantly innovating and improving, held accountable by personal financial ruin. Defense planners, meanwhile, beaver away in middling Government jobs doing cyclical tasks, compromised in outcomes by politics, bureaucracy, vested interest and internal rivalry.

As incomprehensible as might seem today, in the years ahead AI systems will be making nearly all military decisions during conflict. This is inevitable. Imagine facing an enemy led by all of history’s greatest generals combined into one. Further that they had played out each forthcoming battle as if they had a lifetime to prepare. Add to this that they never tire or give into fear or distraction, always performing at their absolute peak. Under these conditions, no human decision-maker stands a chance.

Bottom line: in a world in which two equivalent superpowers are in conflict, the one with the slightly worse AI loses every battle.

Right now many readers may find this fantastical, but consider AI like the internet in scale of disruptive change, only this time in the exclusive hands of a single dominant power.

A solution for how the AI arms race can be safely managed is not yet apparent. Right now, as Allison emphatically suggests, the United States must place AI at the absolute core of national achievement, or surrender global primacy to China. The choice truly is that binary.

Crispin Rovere is an Australian public servant and professional poker player. Formerly he was a Ph.D. candidate at the ANU's Strategic and Defence Studies Centre (SDSC) and has published on nuclear policy. Crispin is the author of The Trump Phenomenon: How One Man Conquered America.

Image: Reuters.