Don't fall for rhetorical answers

"Just let humanity die" may relieve your terror, but it's not a real answer. Just because an answer gives you a conversational out doesn't mean it solves the problem.

I believe that possibly the most important role of PauseAI is to hold space for people to believe together that international coordination can work.

Why? Well, despite most people agreeing that international coordination to regulate frontier AI would be ideal, it’s an extremely painful and exposed position for many to hope for or work toward. “What if it doesn’t work?” It’s too immediately vulnerable for many people to feel like they are out of control of their fate, out on a limb, so they reach for rhetorical answers that make them feel in control and that people around them will back up. Paradoxically, “it’s over” can be this answer!

When people are scared, they want mental peace. They can get mental peace from adopting a position that:

1. other people allow them to hold (it's in the Overton window) and

2. excuses them from having to think about the matter further

Now, one can fulfill these criteria with a valid solution that leads to a positive real world result. But a true solution is often much, much harder to find than what I’m calling a “rhetorical answer”. A rhetorical answer is a mental trick that allows the thinker to treat a matter as already solved without changing anything about the external situation because it gives the person a conversational out of worrying about the problem.

Examples:

• Thought-terminating cliches (i.e. “it is what it is”)

• Flippant or callous refusals to engage on substance (i.e. “mmm, bacon” in response to an argument for vegetarianism)

• Biting a rhetorical bullet (i.e. “if AI can kill us all then we deserve it” breaks the assumption that we need to act on our own behalf)

• “I don’t have to care”/”You can’t stop me”— this is almost a direct appeal to the rules of the rhetoric game, as if to say “I can win on a technicality by biting the bullet of being a bad person”— (i.e. “idgaf about anyone else as long as I have power for my server”)

• Asserting that perceiving the problem reveals negative character traits, like close-mindedness, bypassing the truth claim entirely (i.e. “you sound like a Jim Crow segregationist trying to keep AI writing out of your field— what are you so afraid of?”)

• Asserting that action would be futile (i.e. “it’s inevitable”, “it’s over”, “those grapes were probably sour anyway”)

Because AI danger is just entering the set of publicly thinkable sentiments (the Overton window), empty rhetorical answers like “AI is progress, and I’m not afraid of progress” are harder to spot than if you heard “of course the price of air travel is a 20% chance of ending humanity— what are you, a Luddite?” When an issue is squarely in the Overton window, we are more at ease with its existence. The more mainstream an issue is, the more nuance the public conversation can hold, and the more it can be met on its own on-going terms. Everyone feels exposed on AI. Even people who have been thinking about AI for a long time don’t know where they fit in broader society on this rapidly evolving topic. So it is extremely hard to feel settled in a position on AI danger that doesn’t promise a complete answer.

It’s nearly impossible to find an implementable solution to the entirety of the AI danger problem that also feels likely to happen AND is already in the Overton window. It’s no wonder people are turning to rhetorical answers to resolve cognitive dissonance and find some peace. This is why PauseAI works to bring an AI Treaty into the Overton window and create social support for championing international cooperation for AI Safety, because those are the social conditions we need to stay present with the complexities of the problem and pursue a meaningful solution.

I’m not saying to get worked up about every fringe thing someone starts claiming is a problem. I’m a fan of letting go of overresponsibility on AI danger as well as condemning rhetorical answers to it. But you can’t let yourself believe that, just because you can give a rhetorical answer waving a problem away and feel a little mental relief, that means the real world issue is solved. You can’t let yourself fall for rhetorical answers even though society allows it on the issues, like AI danger, that we are only beginning to wrap our heads around. And when an issue really matters, like AI danger, we have to work to create the conditions for the people around us to see through rhetorical answers to find real answers.