“Arms race” is the wrong mental model for AI. There are better ones.
You’ve probably heard advances in AI described as a classic “arms race.” The basic logic is that if you don’t rush to develop advanced AI, someone else, perhaps more reckless and less safety conscious, will. So it’s better for you to build a super-intelligent machine than to let someone else cross the finish line first. (In American debates, the other party is usually China.)
But, as I wrote before, this is not an accurate picture of the AI situation. AI is like an atomic bomb he has one purpose he is not one thing so there is no “finish line”. It’s a more general purpose technology, like electricity. Moreover, if your lab takes the time to solve AI safety issues, other labs may adopt the improvements, which will benefit everyone.
And, as AI Impacts principal investigator Katja Grace said in Time, “In a classic arms race, political parties can always win in theory. But with AI, the winners may be advanced AI itself [if it’s unaligned with our goals and harms us]. This can cause you to rush the losing hand. ”
I think it’s more accurate to see the AI situation as a “tragedy of the commons.” This is what ecologists and economists call the situation in which many parties have access to a finite and valuable resource and overuse it, destroying it for everyone.
A perfect example of a commons: the ability of the Earth’s atmosphere to absorb greenhouse gas emissions without turning into a climate disaster. Individual firms can argue that it is pointless to underuse their capabilities – someone will just use them instead – but all actors acting in their rational self-interest destroy the whole earth.
That’s what AI is. Commons here is society’s ability to absorb the effects of AI without falling into disaster. Any company can argue that it doesn’t make sense to limit the amount or rate of adoption of increasingly sophisticated AI. The argument holds that if OpenAI doesn’t do it, it will just be Google or Baidu. But if all companies acted like this, the social consequences could be tragic.
“Tragedy” sounds bad, but seeing AI as a tragedy of the commons should actually give you a sense of optimism. Because researchers have already found a solution to this kind of problem. In fact, political scientist Eleanor Ostrom won the Nobel Prize in Economic Sciences in 2009 for doing just that. So let’s dig into her research and see how it helps her think about AI in a more solutions-focused way.
Eleanor Ostrom’s Solution to the Tragedy of the Commons
In a 1968 essay, chemistry, ecologist Garrett Hardin popularized the idea of ”the tragedy of the commons.” He argued that humans compete fiercely for resources and eventually destroy them. The only way around it is full government control or full privatization. “Doom is the destination to which all men rush in pursuit of their own best interests,” he wrote.
Ostrom didn’t buy it. Studying communities from Switzerland to the Philippines, she found more and more examples of people coming together to manage common resources like pastures. Ostrom has found that the Commons tragedy can and does be avoided if communities specifically adopt her eight core design principles:
1) Clearly define the community that governs the resource.
2) Make sure the rules strike a reasonable balance between resource usage and maintenance.
3) Involve everyone affected by the rule in the process of creating the rule.
4) Establish mechanisms to monitor resource usage and behavior.
5) A progressively tougher set of sanctions for rule-breakers.
6) Establish procedures for resolving disputes that arise.
7) Make sure the authorities recognize the right to organize communities and set rules.
8) Encourage the formation of multiple governance structures at different scales to enable different levels of decision-making.
Applying Ostrom’s Design Principles to AI
So how can we use these principles to understand what AI governance should be?
In fact, people are already promoting some of these principles in relation to AI. They may just be unaware that it’s built into the Ostrom framework.
Many argue that AI governance should start by tracking the chips used to train frontier AI models. Writing for Asterisk magazine, Avital Barwit outlined a potential governance regime: “Fundamental elements include tracking the location of advanced AI chips and proving to those who use AI chips in bulk that the models they train meet certain standards for safety and security. Chip control addresses Ostrom’s fourth principle: establishing mechanisms to monitor resource usage and behavior.
Some say AI companies should be held legally responsible if they release systems that cause harm to the world. As tech commentators Tristan Harris and Aza Ruskin argue, liability is one of the few threats these companies actually pay attention to. This is Mr. Ostrom’s Principle 5, tougher sanctions for rule-breakers.
And despite a chorus of tech executives who argued that AI needs to be developed quickly to keep pace with China, it said it needed international cooperation, much as it ultimately achieved with nuclear non-proliferation. Some subtle thinkers would argue. That is Ostrom’s Principle 8.
If people are already applying some of Ostrom’s ideas, perhaps without realizing it, why is it important to explicitly point out the connection with Ostrom? There are two reasons.One is that you have not applied all Her principle is still
Another thing is that the story is important. Myths matter. AI companies love the narrative of AI as an arms race. That justifies the rush to market. But it puts us all in a pessimistic position. There is power in telling ourselves another story that AI is a potential tragedy of the commons, but that tragedy is only a potential one and we have the power to avoid it. .
