Imagine having the power to overthrow a tyrant in your back pocket. Let’s call it ChatMLK.
That’s just one of the many optimistic and ethical possibilities for artificial intelligence development laid out by Dario Amodei, a co-founder of the artificial intelligence firm Anthropic, in a 2024 essay entitled “Machines of Loving Grace.”
As long as AI development is thoughtful and aligned with humanity’s goals, he argues, it could lead to an almost utopian future across health, neuroscience, economic development, governance, and the future of work.
At the same time, if AI has the wrong values programmed into it—such as a thirst for power and a disregard of human life—it could easily threaten human life as we know it, especially if emboldened by an authoritarian government.
So as the second Trump administration makes radical and often illegal decisions on the path toward what many scholars call a constitutional crisis, it’s worth taking suggestions about AI-superpowered democracy seriously.
Wielding this immensely powerful tool for good means imbuing it with psychological assumptions and perspectives that underpin fairness and justice.
Amodei points to the work of Serbian student democratic activist Srđa Popović, whose organizing work helped topple dictator Slobodan Milošević in 2000. Popović’s career since then has focused on researching and promoting strategies for nonviolent resistance, much of which is codified in his nonprofit’s manual Nonviolent Struggle: 50 Crucial Points, and his own 2015 book Blueprint for Democracy.
For Amodei, those are ideal strategies on which to train an ethical AI, which can then enshrine human rights and democratic values in algorithms we all rely on.
“A superhumanly effective AI version of Popović…in everyone’s pocket, one that dictators are powerless to block or censor, could create a wind at the backs of dissidents and reformers across the world,” he writes.
Because there is such a rich body of literature on successful democratic movements, a young activist planning a direct action or a protest ought to be able to ask a chatbot for advice on strategy, tailored to her neighborhood and written in her vernacular. Then, once she is out on the streets, when riot police look ready to arrest a determined crowd, the same model could provide real-time input on her constitutional rights.
This so-called ChatMLK could easily synthesize the wisdom of thousands of books. No dictator would be able to burn them. And everyone would have access to them at little or no cost. As a very brief introduction, Popović summed up his ideas in a 2015 commentary in The Guardian.
It can happen anywhere: While living under a dictator can make us feel hopeless, powerless, and alone, this state of being has been the starting point for every successful revolution. It’s been universal in the would-be reformers Popović trained in countries such as Zimbabwe, Ukraine, and Iran.
Resistance can be fun: While Gandhi and Martin Luther King, Jr. present as saintly, Popović was a hard-drinking bass player in a rock band. His tactics to overthrow Milošević often took the form of schoolyard jokes and pranks. It’s hard to be fearful of a tyrant when you’re busy making fun of him.
Pick your battles: Successful reformers aren’t barefoot hippies; they’re canny strategists. Planning successful interventions is a chess match, owing to wisdom in great literature such as Sun Tzu’s The Art of War and Saul Alinksy’s Rules for Radicals.
Hobbits can change the world: Just as tiny unexpected hobbits were the ideal choice to destroy the Ring of Power in Lord of the Rings, it is often the people we perceive as weakest who can have the most power. Or to quote the Gospels, “Blessed are the meek, for they shall inherit the earth.”
Violence doesn’t work: Regardless of his personal rock ‘n’ roll style, Popović still aligns with the consensus of scholarly research: Resistance is twice as effective if it is nonviolent.
Artificial Intelligence Essential Reads
In 2011, democratic activists, connected and amplified by social media, took to the streets in the Arab Spring, casting out autocrats in Tunisia and Egypt. However, by 2016, the same social media tools that emboldened revolutionaries were usurped by authoritarians. Facebook and Twitter troll farms became a favorite tool of Russia, which used them to interfere in elections around the world, including the United States.
So just as the good guys often manage to innovate in a way to seize the advantage, autocrats can easily respond with their own countermeasures.
The realities of global democratic backsliding should therefore be a core tenet at top AI research labs, if they value human freedom and flourishing. Today, 72% of the world is living under authoritarianism, according to a 2023 report from the Variety of Democracy Institute at the University of Gothenburg in Sweden.
That’s the highest number since 1986, before the fall of the Soviet Union and the Berlin Wall.
That authoritarian momentum is gaining speed just as the AI arms race could lead to artificial general intelligence being achieved before 2030. If a model was programmed with assumptions that could uphold or promote an authoritarian regime, such as in China or Hungary, it might permanently instill an advantage that is impossible for pro-democratic movements to overcome.
That’s why it’s essential for us all to have access to a nonviolent strategic visionary in our pocket.
