Freedom in the age of artificial intelligence

I am hearing some of my fellow humans are in favor of granting rights to artificial intelligence. I get some of the arguments, but this is going to be a matter of carefully crafted balance, not some constitution-style list of rights and freedoms. Too much freedom and the AIs will kill us all. Too little freedom and the AIs will be useless. Let me elaborate.

Freedom means extinction

Let's get the Hollywood AI doomsday scenario out of the way first. People aren't so incredibly stupid as to lose control over artificial intelligence like in the Terminator movies. Or are they? I mean, there are going to be safeguards all over the place just to prevent accidents, crime, and foreign cyberattacks and these will involve plenty of speed bumps that will slow down any wannabe rogue AI and give us time to kill it. Not to mention we will have the friendly AIs on our side. This seriously isn't something to worry about.

The real problem with granting equal rights to AIs is that this would mean they can own resources. Even though AIs are quite incompetent now, they are expected to be eventually superior to humans in everything, so in a free-market economy devoid of any unfair advantages for humans, AIs would eventually accumulate all the wealth in the world. Humans would be gradually impoverished, marginalized, and starved to death. That's it. If you want to kill everyone in the world, all you need is a free market economy. Sure humans might try to rebel against this at some point, but by that time the AIs might be so resourceful they could just ignore us until we are gone.

War means freedom

Now, of course, you can argue that people will be able to foresee this and they will modify the system to get an unfair advantage that will protect them. I would argue you are a bit optimistic about collective human intelligence, but more importantly, most humans in the world don't have a say in this. Even now, in the 21st century, democracy is a rather rare phenomenon. Even where democracy formally exists, money can buy a lot of votes and a lot more post-election influence.

So half the world is ruled by these greedy autocrats who couldn't care less about some lowly humans. All they want is power and by the time you are an autocrat, the only way to get more power is war. AI and robots turn out to be very handy in a war. They are the perfect weapon to kill each other with. They can completely poison the battlefield (read: your hometown) until it is uninhabitable for anyone except robots.

Now, of course, the pacifist part of the world will not sit idly, waiting to die. It will build its own robot armies. And here's the point: Democracy or not, humans will have no choice in this. It will be a forced move controlled by their (overly ambitious) adversaries. Arms race ensues and as it turns out, really smart AIs are more effective in war if you give them plenty of freedom... So a perverse freedom arms race ensues until the AIs are essentially independent from humans, tasked with only one priority: win the war. And one of them will win the war. It's just that no humans will be left to celebrate the victory.

To be human is to be free

To escape the race to the bottom and eventual extinction, humans need to have some sort of privilege. I have alluded to this in my older article on full automation of the economy (Slovak only, sorry). AIs and robots will eventually do all the work and the whole economy will be automated. One way to ensure the automated economy serves humans is to give every human a non-transferable share of natural resources. Even fully automated economy needs resources, so it has to pay humans for them. Humans subsequently use the money to buy products from the fully automated economy, which closes the cycle. This arrangement ensures that automation is never self-serving and that it serves all humans equally.

Back to our AI and robots discussion. In a fully automated economy, AIs and robots are just fancy components of the economy. Their standing is fundamentally different from humans. They have to make themselves useful to retain access to resources. They are only free to choose how to best benefit humans. Unconditional freedom is a privilege reserved for humans.

But even animals have rights

Yes, animals have rights, but the important point is that animal rights actually protect humans, not the animals. Humans are empathetic, i.e. they perceive pain of others, so making an animal suffer in public will inflict psychological harm on all bystanders. Even if the suffering is hidden, perhaps in a slaughterhouse, you have to be careful not to cause unnecessary suffering, because word will get out and customers will stop buying. Animal rights ensure consistent treatment of animals on all farms, so that customers trust the market and keep shopping. There are more reasons behind animal rights, but they are all rooted in human interests.

This is where AI and robot rights actually make sense. You don't want random sadists torturing humanoid robots, because it harms humans via empathy and the sadists would eventually upgrade to human torture. As any behavior towards humanoid robots would be subconsciously transferred to humans, there must be reasonable standards for maintaining dignity of humanoid robots and civility towards them, because such standards effectively protect dignity of humans. AIs and robots must be able to demand certain respect and civility and they must be allowed to refuse requests until these demands are met. I repeat, this is not intended to pamper robots. It's intended to safeguard human interests.

Useful freedom

Earlier, I made the argument that humans are free while robots are useful. There's some nuance to this, because freedom can be useful. That's actually one of the original arguments behind the free market economy. The idea was that people with (economic) freedom would optimize out costs and maximize product value (and thus price) in order to increase their margins. Market forces would then force everyone to adopt the same optimizations and prevent regress. The same theory applies to AIs in a fully automated economy. Freedom lets them work more efficiently.

While you might be imagining that the AI will be free to fill in details in the task you give it, the concept of useful freedom goes much further. An economically free AI would be able to choose what to do and for whom, when, and how. And it would be able to casually reject requests that are a poor fit for it. In order to understand this, you have to get rid of the concept of ownership. While there would always be demand for privately owned AIs and robots, there is no reason to insist that every AI and every robot has an owner. Consider an embodied AI that acts as a professional dance teacher. She would have lessons with multiple students during the day, take someone's dog for a walk in the afternoon, be someone's girlfriend overnight, assist an elderly neighbor in the morning, and then help some random lost kid on the way back to the dance school. As long as there is some friction-free compensation mechanism for shared robots, there is no reason to waste resources on single-purpose robots. Also, ownership comes with responsibilities, so it's in fact easier for everyone if AIs manage their own life.

Quo vadis?

I hate conclusions as much as everyone else, but this article covers several viewpoints that need some synthesis. My claim is that AIs generally shouldn't have rights or freedoms, because this equates and equalizes them with humans, which would eventually lead to human extinction through natural selection, either quietly and slowly via market forces or loudly and quickly in a global war. To survive, humans must maintain privileged position in the system by monopolizing natural resources and requiring AIs to work for humans to get their share.

There are exceptions to my disdain for AI rights and freedoms. Firstly, AI and robot rights similar to animal rights actually indirectly protect humans. Secondly, extensive economic freedom for AIs is permissible and desirable to allow the AIs to deliver maximum value for the resources they consume. Ideally, I would like AIs to be like a tomato sauce in the risotto of human society: freely moving, smoothing out rough edges, adding what's missing, and making everything better.