(Recommended listening: Low - Violence)

Last year, I personally called AI companies to warn their security teams about Sam Kirchner (former leader of Stop AI) when he disappeared after indicating potential violent intentions against OpenAI.

For several years, people online have been calling for violence against AI companies as a response to existential risk (x-risk). Not the people worried about x-risk, mind you, they’ve been solidly opposed to the idea.

True, Eliezer Yudkowsky’s TIME article called on the state to use violence to enforce AI policies required to prevent AI from destroying humanity. But it’s hard to think of a more legitimate use of violence than the government preventing the deaths of everyone alive.

But every now and then some smart ass says “If you really thought AI could kill everyone, you’d be bombing AI companies” or the like.

Now, others are blaming the people raising awareness of AI risk for others’ violent actions. But this is a ridiculous double standard, and those doing it ought to know better.

AI poses unacceptable risks to all of us. This is simply a fact, not a radical or violent ideology.

Violence comes to AI Safety

Today on Twitter, as critics blamed the AI Safety community for the attacker who threw a Molotov cocktail at Sam Altman, I joined a chorus of other advocates for AI risk reduction in -- again -- denouncing violence. This was the first violent incident I’m aware of taken in the name of AI safety.1

Violence is not a realistic way to stop AI. Terrorism against AI supporters would backfire in many ways. It would help critics discredit the movement, be used to justify government crackdowns on dissent, and lead to AI being securitized, making public oversight and international cooperation much harder.

The first credible threat of political violence motivated by AI safety was the incident with Sam Kirchner (formerly of Stop AI) I mentioned at the outset. This incident was surprising, since from its conception, Stop AI held an explicit policy of nonviolence, and members of the group liked to reference Erica Chenoweth and Maria Stephan’s book Why Civil Resistance Works: The Strategic Logic of Nonviolent Conflict.

Research like Chenoweth’s suggests that nonviolence is indeed generally more effective. It’s a little bit unclear how to apply such research to the movement to stop AI, as her studies involved movements seeking independence or regime change rather than more narrow policy objectives. But if anything, I’d expect nonviolence to be even more critical in this context.

When do movements turn violent?

So if nonviolence is often strategic, when do movements turn to violence? Perhaps surprisingly rarely.

People say anti-AI sentiments and movements -- especially those that emphasize the urgent threat of human extinction -- are bound to breed violence. I think this is ignorant and actually makes violence more likely. Environmentalism has been a much larger issue for a much longer time, and “eco-terrorism” is basically a misnomer for violence against property, not people (more on that later).

There are many political issues in the USA that we never even consider as potential bases of violent movements. Even if there are occasional acts of political violence like the murders of Democratic Minnesota legislators or Conservative pundit Charlie Kirk, we don’t generally view them as indicting entire movements, but as the acts of deranged individuals.

My hunch would be that movements generally turn violent because of violent oppression against their members, not simply for ideological reasons. Although there are of course counter-examples, such as bombings of abortion clinics, where attackers justified their actions as preventing the murder of unborn children, or ideologies preaching violent revolution, such as at least some varieties of Communism.

Does violence include property damage?

An important question for “nonviolent” activists is whether they include violence against property in their definition of “violence”. Stop AI does. I assume Pause AI does as well, but it’s a moot point since they also reject illegal activities entirely.

The question deserves a bit more discussion, though, as it’s a common point of contention and legal and dictionary definitions differ. First, there is clearly an important distinction between violence against property and violence against people. An argument in favor of using “violence” to only mean “against people” is that we don’t have another word for that important concept. Still, I favor a broader definition that includes attacks on property, for a few reasons:

1. Many other people use this definition, and I think the damage that being perceived as violent can cause to a movement can’t be mitigated by a semantic argument.

2. Attacks on property can escalate. You are commonly allowed to use proportionate violent force against people to defend your property.

3. Attacks on property can hurt people. Setting fire to buildings, as activists associated with the Earth Liberation Front have done, seems hard to do without some risk of hurting people.

That being said, I think there’s a bit of a grey line between “violence against property” and vandalism. I’d say violence must involve the use of “force”. For example, I think most people wouldn’t consider graffiti an act of “violence”.

“Ecoterrorism”

I think the example of eco-terrorism is instructive. The vast majority of the environmentalist movement is non-violent. However, a small number of activists have advocated for and enacted tactics such as tree-spiking that have injured people.

Hence we now have the term “ecoterrorist”. The very existence of this phrase is misleading. I remember a while back I was curious — who were these ecoterrorists? What had they done? Why hadn’t I heard about it, the way I heard about other terrorist attacks. Well, when you look into it, it’s arson and tree-spiking and that’s about it. I seem to recall reading about one example where actions intended to destroy property actually ended up killing people, but wasn’t able to easily dig it up.

Still, these few actions were enough to add this word to our lexicon, and create an image of environmentalists as more radical and anti-social than they really are.

Conclusion

I’m struggling to find a good way of ending this post.

I believe the actions of AI companies are recklessly and criminally endangering all of us, and the public will be increasingly outraged as they discover the level of insanity that’s taking place. Similarly to Martin Luther King Jr.’s comment that “a riot is the language of the unheard”, I do understand why this emotional outrage might provoke a violent response.

But I hope the movement doesn’t spawn a violent element and that these recent examples are isolated incidents. To make that more likely, we should continue to vocally espouse nonviolence, and denounce those who would encourage violence among us.

But ultimately, movements are fundamentally built through voluntary participation, and nobody can entirely control their direction. The response should be to try and steer them in a productive direction, not to avoid engaging with them.