Would You Rather Let the World Burn or Be a Side Character?
There’s a common pattern when people start to believe that advanced AI will be developed soon, and that there’s a reasonable chance it’ll end the world. It goes like this: rather than feeling the heavy weight of the challenge at hand, coming together and figuring out what you can do to best do to avoid terrible outcomes, your monkey brain activates, and you start thinking about how to make it this about yourself.
An insanely egregious instance of this
One version of this is the classic “well, if someone’s gonna build it, may as well be me” line. As Sam Altman put it,
“I think AI will probably, like most likely, sort of lead to the end the world, but in the meantime there will be great companies created with serious machine learning.”
Or as Elon Musk put it,
“I sort of came to the realization that it’s happening whether I do it or not. So you can either be a spectator or a participant. I’d rather be a participant.”
I think Elon’s quote is importantly different from Sam’s, in that it’s more about being a side character than it is about craving the money / power associated with building AGI. Everyone wants to be a part of the AGI story. If the world is watching Pantheon play out in slow motion, it’s only natural to want to be part of the show. When I really self-reflect, there’s something kinda sick about being the villain with wild dreams of an AGI future, who’s steadfastly resolved to build the machine god, as opposed to a random extra that you only see for a single scene.
This “kinda sick narrative” feels hugely applicable to Anthropic’s story. They split off from OpenAI because they thought it was too reckless. Instead, they want to get to develop AGI before OpenAI, then use their AGI to solve alignment and save the world from the bad guys. This is a pretty epic role in the show; much better than the coffee shop role AND the evil villain.
Who I strive to be
If I’m ever in a situation where I must choose between (a) helping AI go well, or (b) helping AI go well in a slightly worse way, but a way which positions myself to get more influence / fame / a better role in Pantheon, I hope I choose the former.
I want to be like a really good janitor of a public school. I want to be satisfied with supporting the systems that are making a difference, being the quiet janitor who keeps the school clean, takes care of repairs, and works with a quiet dignity, making the school a functional, clean place where students can learn. I think this was captured very well in the movie Perfect Days.
A less egregious instance of this
If you really care about being part of the action, you might do things that slightly harm AI safety overall, in order to steer the ship in your favor. It manifests itself as something like: “me and my in-group’s particular AI safety memeplex must be the dominant cultural force.” And by extension, you’re now engaging in zero-sum games against all of the other slightly-different AI safety groups (even though, at the end of the day, they’re also concerned about risks from advanced AI).
For the sake of making the question interesting you can’t use one to get the other: would you rather have influence, money, or fame?
- If you choose influence, maybe you’re a close friend of an important political figure, and the person they go to if they need advice on AI. You don’t have all that much money, and no one knows you as an “important AI thinker,” but you actually have a pretty huge causal effect on AI policy.
- If you choose money, maybe you just write silent grants to AI safety people, hanging around in the dark but supporting areas you think need help. Once again, this doesn’t give you much influence or fame, but you get to choose what gets support.
- If you choose fame, let’s say you’re Mark Ruffalo but not rich. Or maybe you have Roon levels of influence over AI twitter.
