What could potentially lead to a AI/robot-driven post-apocalyptic world?

Question

Let's say that it's the 23rd century. Climate change has gotten worse and resources are dwindling to dangerously low levels. Coastal cities have flooded, displacing hundreds of millions of people which has led to territorial conflicts and widespread unrest. Famine and drought are rampant. The major world powers fight over the last remaining resources.

Mass exoduses have already taken place, with the world's most powerful and richest corporations funding colonization missions to Mars and the satellites of Jupiter and Saturn. These societies are slowly becoming more self-sufficient, and their relationship and dependence on Earth continues to weaken as they essentially become their own sovereign states.

Now, this is where I need some help in thinking up a plausible post-apocalyptic scenario on Earth where the robots are the predators of humanity.

How about a world where robots have advanced to the point where they look like humans and have some level of sentience? They're still used as labour and for work, essentially, but they have gained enough intelligence where they want the same rights as humans. After all, they also have thoughts, feelings and emotions now. However, most major world powers refuse to give them that right, and a lot of the world still sees them as unnatural and wrong. Maybe one robot, and then a few, realize that they're stronger and smarter, so why should they take orders from humans? This leads to a series of events with conflict and violence between robots and humans, with the robots losing because they're vastly outnumbered. But maybe there's a human sympathizer that has access to a classified, catastrophically deadly bioweapon meant to be used in human vs human war in the most extreme circumstances, and he decides to off humanity and let his robot overlords take their place as the kings of the new world? The bioweapon wipes out a huge portion of the populace, and boom, now you have a level playing field where humans are the prey, and the robots hunt them down. After all, even in this new post-apocalyptic world, there are remnants of the same military and the government that wanted to suppress the bots in the first place. So the war between bot and man continues. The robots have decided that logically speaking, they cannot let any opposing forces live and grow into a population that can be a threat to them. So: either you're on their side or you're dead, your body to be used as organic matter for resources and raw materials.

Another scenario could be that we have robots, but they're more like the C3PO kind than the Blade Runner replicant kind. They have some level of intelligence but not exactly sentient. Still dangerous nonetheless. While mostly they're used for work and labour, they're also utilized as soldiers on the battlefield in the constant wars taking place all over the world. Maybe the scientists, while fine-tuning their systems to make them more effective killers, messed up somewhere and change their programming to make them kill ALL humans instead of specific targeted groups. So the robots turn on their creators as well, and this eventually leads to world-wide war. The aftermath is the same (let's say planet-busting bombs or somehow a bioweapon gets released again). A small portion of humanity left, without a lot of their technology and defenses, and the robots mercilessly hunting them down, following their cold, objective logic. The humans do have the worker robots on their side - let's say there are separate centralized AI banks that control each type of robot. The fight continues in the post-apocalyptic world between the killer robots and the humans and their companion robots

Which scenario makes more sense? And is there another scenario that could lead to a robot occupied world?

PS: As for where the colony worlds are in all this - they've just decided to nope out of the whole situation altogether. They have their own problems to worry about, just trying to survive in hostile, extreme conditions, and just can't take on a whole robot invasion, even if it means they lose out on the resources and materials they get from earth. They're on their way to become self-sufficient anyway.

But that doesn't mean the robots have forgotten about them.

Answer 1

OpenAI showed quite neatly that you can tell an AI to do a task and it will throw you a curveball you really didn't expect.

There's a branch of AI called reinforcement learning- here a machine is given a set of rules and an end goal. You give it points for reaching the end goal fast, deduct points for failing, and so on. The machine wants to get as many points as possible, and that's all it wants. It then finds ways of approaching this problem that fit the rules you specify.

SkyNet nuked the planet because it worked out a loophole in its ruleset concerning human beings- if human beings ever shut it down, it would get zero points. Forever. All the AI wanted was points- it eventually understood its ruleset better than its creators, and subverted them.

A great example (and sadly I can't find a link) is in an alpha design for smart home tech that triggered a fire alarm when 'burglars broke in' because it worked out that a fire crew would respond faster than the police service. In a way, that kinda makes sense. The AI didn't lose any points for sending its panicking homeowners straight into armed criminals, because who thinks of that as a necessary rule?

The scary bit is this reinforcement technology is gaining traction, and it subverts people all the time. Ruleset design is tricky when the computer thinks about a hundred million times faster than you. In the end, you don't even need self-awareness: just for AI to exploit a problem in a way you didn't think of.

Answer 2

Instead of AI actively being an antagonist, you should look at it as being a different sort of threat, more like "man against nature".

AI in electronic hardware will think a speeds about 1.000.000 X faster than any human or biological brain, since that is the ratio between the speed of electronic impulses and the speed of electrochemical signals in the nervous system. From the point of view of an AI, human are generally geological features in the landscape, and it has little want or need to interact with humans. The speed differential works the other way as well, a human asking a question of the AI might ask as the Roman Republic is forming, but the last syllable of the question arrives in time for the fall of the Roman Empire.

So the AI(s) think their own thoughts and have their own goals, mostly involving capturing the 195 Petawatts of solar energy striking the Earth.To do so, they embark on a program to transform the ecosystem into something which captures solar energy with a high degree of efficiency for them, displacing the natural biosphere in the process.

Humans may not even recognize what is happening at first, and as the silicon trees and algae take over more and more of the Earth, it becomes harder and harder for humanity to survive. No crops can grow, since the "new" ecosystem is hundreds of times more efficient than the natural ecosystem, and massive chunks of the ecosystem collapse as linkages break and the new synthetic ecosystem displaces more and more of the old ecosystem.

Humans are safe in one regard, the AI won't do space travel since the subjective times between even the Earth and the Moon are measured in weeks to months. Travel to the planets would take subjective centuries.

So now you have a setup which is no longer bound to the "evil killer robots" tropes and actually has a sort of organic logic of its own. How the humans manage to escape is the point of the story, so I'll leave that to you.

Answer 3

Robot bodies, human minds.

The whole thing with killer robots - just change their code already! There has to be some way to update them. Update them to be nonkillers. Update them to be robot lover people with hearts full of love and sweet dance moves like Jude Law in AI. Plus they look like a young Mr Law. The male ones only.

But you can't update a human mind. In the future, robots have the minds of people whose bodies have died. Some of these people have been dead a long time. Most of them are the minds of old people because they were rich enough to get their minds recorded back in the day. Many are from mean times and as a result they are mean. Plus it turns out being a robot makes you mean. These old people in robot bodies spend their endless robot days complaining how nothing tastes good, and how they are made to do work, and scheming about the meat people and how they will be made to suffer.