In the US, the extremes you are fantasising would be unconstitutional "cruel and unusual punishments", and I can't see it flying in any civilised country, whether or not its constitution says anything. I'm sure governments have more direct ways of shutting down dangerous activities, e.g. directly capping AI development. Which people are arguing for already.
It's certainly unconstitutional, but not necessarily for that reason. I'd point instead at the prohibition of bills of attainder, rather than cruel & unusual punishments.
Here's the core of the proposal:
It would be possible to pass a law that such-and-such a corporation has, due to its past bad actions, been designated as a ULC.
Such a law would closely match the notion of a bill of attainder, "an act of a legislature declaring a person (or group of people), guilty of some crime, and providing for a punishment, often without a trial."
In the American system of law, imposing penalties for past bad actions is the role of a trial court, not the legislature.
I'm going to be quick about the background to this post, as I assume that here if anywhere everybody knows about it. The background is as follows: one worries about AI superintelligence, recursive self-improvement, yadda yadda; one thinks that perhaps we're not as far from the foom as one might wish we were; one worries that the foom might lead to doom; one thinks that it would be nice if everyone would pause development of AI until we are a little more clear about things; but it appears that nobody will pause AI development because there is too much money to be made -- because, among other things, everyone is thinking to themselves, "If I don't develop it, someone else will."
I first had this thought about ULC's a few months ago, watching this youtube video with Tristan Harris. And then yesterday came this twitter thread by Jacob Coxon, which Gary Marcus discussed this morning on his substack. But there are of course a hundred different flavors of the idea out there.
Tristan Harris seems to think that the only solution to the problem is some massive collective action. Jacob Coxon doesn't seem to have a solution other than to resign his job and call for others to think hard about what they're doing. Gary Marcus is calling for a boycott of AI companies. Other people have talked about firebombing data centers and so on. The ULC is just a different thought, a potential way of thinking about how to address the same problem.
Also, I want to be clear that I'm not a doomer of any stripe! The ULC is for those who are more doomerific than myself.
OK, so that's the background.
.............................................................
What is a ULC?
A ULC is exactly like an LLC (a "Limited Liability Corporation") but with two extra letters: "Un." In other words, an ULC would be an Unlimited Liability Corporation. A ULC would expose its CEO, CFO, other decision makers, ... and perhaps others? ... to Unlimited Liability.
What would be comprised under Unlimited Liability? Hard to say! One could start with: the loss of all assets of all kinds anywhere in the world. Perhaps one might go further, depending on just how much "U" one wants to put into the ULC: perhaps loss of right to own property in the future? loss of right to enter into contracts? perhaps even loss of right to fair trial? right to life? right not to be tortured? I don't know! The concept of the ULC is slippery, because "unlimited" is hard to parse.
Who would want to invest in a ULC? Well, nobody! Who would want to be associated with a ULC? Nobody!
So the idea is simple. ULCs do not currently exist. But they could exist.
It would be possible to pass a law that such-and-such a corporation has, due to its past bad actions, been designated as a ULC. Anyone who wants to sue the company can go after not only the assets of the company but also the assets of the CEO, CFO, and others (details to be specified in the law). Anyone who can prove in a court of law that the corporation's actions have led directly to the deaths of people close to them can hold the same set of individuals directly responsible for those deaths.
Designation as a ULC would of course destroy the corporation. It would as well destroy the lives of those closely associated with it.
Designation as a ULC would be a last resort, presumably. It would be the sort of designation that would only be applied to the worst of the worst of the worst.
But this is what the AI x-riskers are talking about!
Let us say that OpenAI releases an AI into the world that ends up killing a million or fifty million or a billion human beings. In such a world, OpenAI would perhaps be considered for designation with ULC status. If and when ULC status were applied to Open AI, the company's CEO, its CFO, perhaps other top employees -- perhaps all of its employees?, perhaps lawyers who have represented it??, perhaps shareholders??? -- would be subject to Unlimited Liability. And who knows what that might entail. The details would be worked out by the angry survivors of the chaos that the company unleashed. It might get ugly.
.............................................................
The point as of today is not to pass any laws about ULC. The point is not even to agitate that people start thinking about passing laws about ULC. The point is to get people to realize that it is possible that in the future things like ULC laws might exist.
In the current structure of capitalism, it is easy to form LLCs and C-Corps, and forming those sorts of corporations shields founders and employees from certain consequences. But that structure is not written in stone. It can be changed. People can get angry, and when people get angry they can want to take revenge, and when they want to take revenge is it hard to say just how far things will go. The ULC is just one possible end result.
Perhaps, just perhaps, if the idea of the ULC gets a bit of traction, AI executives who are doing their cost-benefit analyses as to whether or not it's a good idea to create the next-gen AI system ... will think a little bit differently. Perhaps they will say to themselves: "If I don't do this, someone else will -- and thank God they will do it instead of me, because at least then I won't be the one to lose everything I have ever owned, have my arms and legs eaten off by rabid dogs, and spend the rest of my life rolling around the floor of a dungeon begging for the right to eat rat feces." (Or whatever image comes to mind when you start thinking about what Unlimited Liability might mean -- what retribution a hundred million furious human beings might demand for the loss of their loved ones.) Or maybe the AI executives will just think there's a tiny chance of an outcome like that, and maybe that will be enough to change their actions.
To be clear, I think a world in which ULC's existed would be a much worse world than our current world! I'm mostly a fan of capitalism and corporations and so on. I don't trust the masses not to use a ULC law to destroy perfectly ordinary companies who happen to fuck up. But a world in which ULCs existed would be better than a world subsequent to the mass death and destruction of the sorts that AI x-riskers worry about.
More importantly: a world in which ULCs do not actually exist, and yet ULCs are imagined by elite actors as a possible future if they act badly enough -- that is, in my view, a better world than the current world.
In the current world, elites seem to believe that they cannot truly be held account for their worst actions so long as those actions are conducted behind the shield of a corporate entity. Maybe they shouldn't be so sure.
If and when the shit truly hits the fan, maybe a corporate structure won't protect you. Maybe at the end of the day you're just a primate living among other primates, and if they get angry enough they're going to find a way to literally tear you limb from limb, just like they did to your most heinous ancestors on the savannah ten thousand years ago. Maybe you want to take five minutes to think seriously about that.
.............................................................
Again, I'm not a doomer! At all! I don't think that foom-and-doom is going to happen, or can happen. Also I'm a peaceful person; I hate violence. Please don't take any of this as a call for violence. Please don't believe that I'm making a threat of any kind.
I do think that the elite AI doomer discourse is suffused by two ideas: (1) that everything in the future is going to be awful; and (2) that nobody powerful is going to pay any price for making things that way. I think this combination is extremely unlikely!
The ULC is really just an attempt to translate the language of the jungle -- the language of the primitive primates that you and I truly are, deep in our hearts and souls and bodies -- into a language that AI executives, corporate lawyers and accountants can comprehend. Maybe you can come up with a better one.
Thanks for reading.