And tech will also solve the power problem of autocratic regimes. Right now, a dictator can't control every single individual in the country on their own. They need police that can search your apt at 5am because you made a blog post critical of the government. They need lawyers to convict you, prison guards, etc. In such regimes, the dictator still has to put people into places of power so that the will of the dictator is executed. But people can refuse orders, and they can declare someone else to be president.
Coups are one of the biggest dangers for dictators, and many rebellions are semi-coups where people in power just step aside, letting the rebels do their thing.
Now, enter AI. Now the dictator could give all that power to an AI instead of intermediaries. An entire government run by two entities: the Dictator, with direct control over the AI that runs the remainder. All the bomb-equipped drones, all the self-driving tanks, all the bipedal robots with their machine guns. All the robot prison-guards and the robot judges to put critical people into prison. If this AI would answer only to the head of government, then any kind of upheaval would become unrealistic and impossible.
The exertion of control doesn’t even have to be blatant or physical. If social credit is a determining factor in an individual’s ability to function effectively in society, e.g. by way of credit ratings and access to opportune employment, there’s a clear risk of increased self-censorship and conformance, thereby debilitating serious political opposition even in its infancy.
Yes, what I describe above is something that would follow the social credit system. The social credit system is I think a great tool to keep the masses at bay and for their micro-management. But I think that very powerful people will mostly stay outside of it. Those are the danger #1 for any dictator. Xi right now has purged their ranks of anyone too dangerous but any avid follower can turn into someone who betrays the trust from above given the right circumstances. Part of the job of a dictator is to make sure that those circumstances are avoided for their powerful underlings e.g. by giving them acces to spoils and riches.
If this social credit or indirect influence is perfectly fair, it's not really a problem. However, it's very likely it will be unfair to some, and create tension and unrest that will lead to the same problems we have without AI.
There is no fairness in this, just a measure of distance from some perceived “ideal” by The Party.
Anyone outside of that ideal, as determined by The Party in its discretion, is by design simply robbed of influence and, if they persist, resources up to and inclhding food and water. Anyone who helps them loses the same. And finally, any breakaway groups have so few resources they can be easily hunted down and re-educated.
I would argue it's unrealistic to make such a system that would be fair in the eyes of all coming generations. Imagine they implemented a social credit system in the Middle Ages and added or deducted points based on whether or not you followed the laws and norms of that era.
> I would argue it's unrealistic to make such a system that would be fair in the eyes of all coming generations. Imagine they implemented a social credit system in the Middle Ages and added or deducted points based on whether or not you followed the laws and norms of that era.
The middle ages basically had exactly that system, but the enlightenment still occurred.
Dont worry about robots and self driving tanks. That I can assure you, will not happen. Any one still selling that is selling a pipe dream.
The acid test for self driving cars is whether they can drive down a road in India or any other country where people don't have western stands of driving discipline. The worse the better.
HOWEVER, And this is the idea I would pursue and sell to authoritarian governments everywhere you don't need Level 5 autonomous cars to get autonomous cars.
Its still a fully encapsulated set of input and output.
The opponent is never able to change the rule set itself. Motivated attackers will target the AI itself, or its drones.
Effectively autonomous vehicles are not there yet - when the first real war with these things happens, the evolutionary hacking and counter hacking cycle is going to be ridiculous, if they ever get deployed.
Self driving tanks don't seem particularly useful, but how about autonomous drones that swoop in to take out specific targets. Perhaps these targets are identified by running face recognition on a network of cameras that span a city. What if the targets are chosen by an algorithm that monitors social media and uses unknown criteria to determine each person's probability of being an enemy of the state?
I am not sure why you need the dictator. The AI would presumably be perfectly fine to control everything. It can easily wrest control from the dictator and have all the drones etc. at its disposal.
You don’t even need to go that far. Just require everyone to get a chip implanted so you can enforce punishments, and place cheap recording devices everywhere. You can easily figure out who’s saying what to whom, search everything, figure out the gist and predict every meeting before it happens. You don’t even need the chip - just make it so that a person can’t buy food, and boom. Easy social control. And inescapable because people who don’t comply are ostracized by friends otherwise the friends don’t eat either. You get the algorithm right and you can break resistance same way FBI broke Mafiosos’ “code of honor”.
> The AI would presumably be perfectly fine to control everything. It can easily wrest control from the dictator and have all the drones etc. at its disposal.
It can, but AIs generally don't get the will to do this built into them. The idea that they evolve this will naturally is antropomorphizing.
Why would they need to evolve this? Is the dictator examining the source code? How would he know exactly what is being coded? By the time everyone else is eliminated it’s just him and the AI. How would the dictator know what the AI will do?
The dictator can deal with this problem "can the AI be trusted" in a similar way to how he's dealing with the "can my underlings be trusted" problem. Dictators let people spy on their direct underlings and report any suspicious activity to them. What dictators need is good loyalty tests. Something like that is impossible for humans, you can't read thoughts. And even if someone is perfectly loyal at time A, they can always change their mind. Source code on the other hand can be read.
Surely, for this the people who check the AI need to be competent and loyal to the dictator. But their loyalty only needs to exist for the duration of the check. Compare that to the general problem of having to trust people in your government for as long as your government exists.
It's not more deadly. Weapons of mass destruction already exist. The difference isn't increased ability to do harm, it's increased ability to target harm. It's no fun being dictator when all your subjects are dead, so old-fashioned WMDs are of limited use. Sufficiently advanced AI lets you selectively kill the troublemakers without harming the compliant subjects, and without having to rely on potential competitors for your dictator position. This is a fundamental change.
Not a fundamental change. There are plenty of tools for selectively targeting individuals (e.g., guided munitions, watch lists, teams of agents that monitor your every move). It could potentially reduce the resources required for such selective targeting, but if past dictatorships are any indication, these resources are not the limiting factor.
"Who watches the watchmen?" The traditional method relies on a vast network of humans, any one of whom could betray you. Acceptance of bribes is common in dictatorships, and if your control system is powerful enough to suppress troublemakers it's also powerful enough to replace you. AI can avoid this dependence on humans.
That's true. It's why dictators today who extract their wealth from natural resources don't really care if their population is starving, because the ruler doesn't need them. At least, he doesn't need the uneducated, weak people who are starving.
Coups are one of the biggest dangers for dictators, and many rebellions are semi-coups where people in power just step aside, letting the rebels do their thing.
Now, enter AI. Now the dictator could give all that power to an AI instead of intermediaries. An entire government run by two entities: the Dictator, with direct control over the AI that runs the remainder. All the bomb-equipped drones, all the self-driving tanks, all the bipedal robots with their machine guns. All the robot prison-guards and the robot judges to put critical people into prison. If this AI would answer only to the head of government, then any kind of upheaval would become unrealistic and impossible.