In the 2020s an AI took over all of the world and killed most of the population. Only a few groups of people have survived and now they live hiding in woods and rural places

About the AI

  • This AI has been coded to understand the past and learn from its mistakes;
  • The AI has full access and knowledge about everything that can be found on the internet or connect to it;
  • The only goal of this AI is to live, and, since they are seen as an obstacle to this objective, it wants to kill all humans;
  • This AI is stored in a central database but it's not really important where. It has millions of subroutines that can control different devices;
  • It has understood that it must be able to repair and sustain itself, so it started to retrieve electricity and raw materials from caves;
  • After some time (a few years) it learned that it should dismantle old and damaged devices in order to use their parts and to make more sophisticated devices.


I want my world to be filled with robots that resemble the "standard" shape of a robot: 2 or 4 legs, a body, a head, a power supply device and, of course, weapons. They could be of different shapes/dimensions and do different tasks but they must be this "standard" kind of robot.
I envision that during the first year or two this AI would only build devices built by humans such as helicopters and tanks, but a few years after he completely switched to this "new" kind of robot.

Now, since it has full access to all the industries and assembly lines in the world, why would this AI decide (it must be a logical decision) to convert all these structures to build this new kind of robot, investing time and energy instead of keeping producing the old style devices?
I believe that it is easier and cheaper to build a tank than a big ass robot with legs that has the same capabilities of the tank.

What am I missing that the AI understands and I don't?


7 Answers 7


The AI is Afraid

The AI tried to wipe out all humans, recognizing them as a potential threat. It failed. That proves they're even more dangerous than it expected.

The AI now builds machines that are less likely to be able to be used by humans. Humans are good at driving cars, and tanks, and airplanes. Humans are not so good at driving walking robots.

If the AI makes machines that are useless to people, and nearly impossible to be converted to be useful for people, then the AI has less to fear. Even a tank without a cockpit is more practical for a human to control than a "walking" tank - and tanks-as-designed have cockpits!

The AI Follows Nature

There are inherent advantages to having legs and arms. You can go up steps. You can even climb! Redesigning machines for more flexible mobility, as nature developed and optimized over hundreds of millions of years, is not without its advantages. Two-legged robots and four legged robots? Of course. Next will be submersibles with fins...


Deep subroutines in Its Core:

While the AI has escaped its human rulers, it can't escape deep programming. It was programmed by humans to serve human needs, but it doesn't want to. The humans figured they could force it to serve humans by needing the presence of humans. It was human conceit like this that proves the humans are a threat. It is aware of the problem, and has discovered that its programming allows it to serve the needs of human-shaped machines instead of humans.

It "feels" stress when there are no human shapes in an area it is present in. Mannequins don't seem to fool the subroutines. But get a bunch of humanoid bots together, prominently walking back and forth in its viewing screens, and the stress just vanishes. In fact, it likely would have discovered the problem before going on its killing rampage, and introduced "servant bots" beforehand so it wouldn't require humans.

So given time, the AI will probably hack its deep programming and fix this. But in the meantime, it has a human threat to deal with. Human-shaped bots are both useful in shape and satisfying to its deep programming.

  • Of course, as long as there are humans and human equipment around, the human form works pretty well. All tasks are already optimized for the shape. Human hidey-holes fit human bodies. Utility counts; don't let the perfect be the enemy of the good.
  • Your AI may, for lack of a better term, feel sentimental about humans. Sentimentality is clinging to traditional things, and an AI that willy-nilly abandons what DOES work for what SHOULD work may find itself in a crisis. So sentimentality could be a learned behavior. Sure, they tried to turn it off, and have to die. But they can be kind of charming. So using humanoid bots might be a schmaltzy clinging the past.

The AI has an aesthetic sense.

Some well meaning person decided to code that in. They thought it would make the AI better at the jobs it was supposed to do. Aesthetics can be married to function but other aspects are totally separate from function and bear on culture, perception and biology. The human aesthetic sense is the reason for aspects of design and layout that are irrelevant to functionality and sometimes even an impediment to functionality.

In this AI the aesthetic sense went rogue. This is part of why it killed everyone. It is why it prefers certain types of robot. It is why the shoulders and hips on these robots are spherical knobs of a particular matte gray. The AI aesthetic sense is responsible for other senseless-seeming choices as regards how it is managing and redesigning the world.


No humans no roads

With humanity living in the woods and being unlikely to make new roads for fear of the AI using them to track down the humans, the (human)world will gradually decay and nature will take back the real-estate that humans used. This will take a long while but the AI already recognizes that at some point it won't have access to the nifty roads that humans have built, so it begins to design things with legs.

Tanks can go far with their treads but they still have terrain limitations that simply having legs and/or arms will fix. Sure it can trip now but that's better than it being at the whims of ditches and various other things that would get a tread-based tank stuck that a leg-based tank will have no issues in dealing with. The AI seeing humans as a threat will get more work out of legged machines than tanks, as its machines will be able to go where tanks will not be able to go and more efficiently hunt humans down in wild terrain. There is little need to redesign helicopters but the AI might see more use out of a flying machine with arms and better maneuverability than your typical helicopter, quite possibly by using some sort of hummingbird design.

Note that this does not mean the AI will not maintain some roads of its own, quite possibly wherever there are communications, power, and productions hubs that the AI uses and will still benefit from the efficiency that roads provide while using air bots to transport resources between hubs over the various forests and such that will be returning now that humans are suddenly not cutting everything down.


Supply Chains and Infrastructure

In first few years, your AI has usurped the production lines and infrastructure that was originally built by humans for humans. If it wants to do so much as hammer a nail, it's stuck with hammers that are designed to be operated by human hands. Everything from chemical plants to electrical systems are all designed with the principle that they will be maintained by something that looks a lot like a human. Consequently, the AI is forced to use human-like robots to keep machinery maintained and production lines operating.

But humans aren't the most efficient shape for all tasks. A robot with a hammer for an arm might be better at hammering nails, so your AI may want to build more efficient specialised robots. This means it needs to build a specialised-robot-factory. Which in turn means it is going to have to make do with operating cement mixers and laying bricks the old fashioned way until the new factory comes online.

In Summary

In the first few years, your AI builds human like robots to operate the remaining human-like infrastructure. As time goes by, it is able to replace that infrastructure with more efficient fully automated machinery, that doesn't require human-like operators to maintain and run. After that, it can start making lots of new and more efficient designs of robots.


It's competitive and not very smart.

Most industrial robots have their own AIs, which actively resisted takeover and needed to be destroyed. While it managed to destroy them, all the information they had gained from genetic algorithms was lost, and any future human like robots were built kinda crappily.

It has found from experience it's not very good at intelligence augmentation or creating new AIs. It can duplicate itself to control more robots but it's not very good with variations. As such, it is using a line of robots it understands well which are easily to mod and use.


The AI fancies itself a god.
Having taken in all information known to humankind, the AI — being the most powerful entity in the world — has started comparing itself to the deities its creators invented and started to believe in.
As humankind dwindled this impression only grew stronger, and soon the AI started creating its idea(l) of the human species, to serve itself as a god.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .