- There are different types and levels of intelligence out there
- it would be strange if they did hunt humans specifically
Let's consider the problem on 3 sufficiently different substrates - humans, other life we know, a practical mechanical swarm of our own
Humans have some level of cognition, which still amuses me how do they manage to do so with just 86 billion neurons on average - holy mother nature...
Despite quite developed cognition systems, there are 3 general directions of examples, there are more but just 3 is enough -
// natural cracks in cognition, injuries, laziness aka world is big
And humans are used as an example of advanced intelligence, but most of the stuff is applicable to most intelligencies that are found on this planet, same principles even if in different forms and shapes.
- I highly highly recommend watching the video, it is a whole field of research to find ways to break artificial neural networks, and if you do that instead of reading the answer you still will be enriched with the useful and profound knowledge, so as it may be useful to understand the answer in a better way.
Magicians, illusions & Co
There is a plethora of different stuff which slips through the cracks of human cognition, as it is specialized and not that good to notice things that we are not trained to notice or which we can't notice because of limitations of our sensory organs and either unusual from evolutionary perspective things.
Magicians and other guys often do exploit those cracks in our cognition and sensors. Over time we can learn how things are done, but not always, some tricks still look like magic for most humans, even if there are special shows which explain a lot of them - but the majority of people have to jump over their heads to understand how it is done, for some tricks, because quite an unusual state of minds of those magicians and training their do, etc - which people do not anticipate because of their worldview and other stuff. The only thing which helps is that we do know that there is no magic - but what if it wasn't true, then our life distinguishing all the things would be 100% more difficult, meaning humans would be required to be a lot smarter.
So the premise "there is no magic" greatly simplifies the cognition problem of perceiving the world for humans, but at the same time if it would be hardcoded, which is quite reasonable for a drone swarm, then if there would be magic then it would fail to recognize it. This restriction, no magic, is a high concept, but there can be many such restrictions like a set of small concepts - which would create many shadow zones, some of which are discussed later in the ant's section and swarm intelligence in which they are defined by omission which is created by the existence of simple rules - so somewhat inverse picture.
There are visual illusions which do not mess with human optics, but attack visual cortex and pattern recognition systems - different specially generated image, or 3d pictures, etc. So as special ways to mess with current recognition software which is based on the same or similar principles - so it is not only about humans, humans are just one case of many.
There are a plethora of cases here as well which can be generalized as corruption of brains due to injuries, errors in development and growth etc.
Recently saw one example in the answer:
Some people are born unable to distinguish objects and they see the world like a photographic camera, just shapes and colors at various distances, being unable to tell the difference between a person and a wall without first asking themselves "are there any flat and giant people?" or "this skin colored figure is in my home and coming towards me, it must be a member of my family"
- this condition, which I'm not sure exists but do suspect has some psychiatrical therm, it not necessarily affects everything in human brains - the person can be fine in other aspects, it is a property of neural networks they degrade gradually in case of random corruption, but you can corrupt a network in a specific way - which means it can be made by aliens in what they perceive a totally healthy way. And such failures do not mean the system isn't smart or incapable, same as it goes with that human case. it isn't stupid, it just the case breaks it.
Human cognition can be quite a delicate thing and robust at the same time and fail in quite a variety of ways, but One man's trash is another man's treasure - and we will see that on animal cases, but so as it applies to alien products - life in the galaxy isn't that abundant to test on it and working in one case does not mean working in a different one, but also it does not mean stupidity on more usual problems - planets are abundant and are fundamentally not that much different as conditions and way matter behaves are quite standard things, usually for the most part.
But the point is - a problem of recognizing something is not trivial, especially if the things are unusual and or the first time you see them. But even if you manage to recognize is just one of the necessary premises to do something.
Humans, energy save mode aka laziness
Most probably know that ants bite, why - personal experience. Birds are capable to do harm to children and domestic animals. We do not wage a war, as a whole as humanity, on those, usually.
One of the reasons is there is too many things, too many life species that can do a certain degree of harm, and if our general objective, as mechanical entities would be to reduce the harm inflicted by Mussels in cooling pipes of our reactors, we just take precautions in the system affected, preventing single points of failure etc etc.
Algae clogging system isn't much different from hamsters doing the same etc. And you remove one species and life is nasty, others will take the free space, and it not necessarily for the better, look on Australia cluster, maaaan ...indistinguishable jitter of transistors ....
Regnum Animale - animals, insects - multicellular moving breathing organisms have all grades of intelligence, which we know of, from very basic to more advanced levels.
Also, some of them form swarm intelligence's - humans are one example of that, so as ants, so as other examples.
- and their nasty brother's termites, which are cockroaches
Let's look at ants, quite a popular subject of observations and a mix of extremes, which also translates to a drone swarm in a straightforward way as well.
What is undeniable and interesting for our case:
- a single ant is superior to a rock, but that's it, the level of intelligence it has
- as swarms, they are quite successful, capable to solve problems which are important for them
Termites quite notorious for their habitats, the sizes, and if not their weaknesses, mainly bodies, they would be everywhere. So one of the few species that alter landscape on a noticeable level, not on the level of great pyramids, but if not certain limitations aka food sources and absence of reactors for energy production - would be not surprised if they would make those pyramides as regular homes.
But their individual low IQ and other limitations of communications sell them short as a true hive mind and all they are capable of, in terms of the world picture they operate with, is to react. They do react to immediate danger, in different ways, but they are incapable to build a strategy in the great big world as they do not have the model of it. The model of the world they have is quite basic.
And a drone swarm with the purpose of processing random planets does not have to have a complete or complex world picture. What a creator cares about is sticking to some general objective and not deviating from it to become universe devouring force - basic self-preservation. And if that means them failing in 1% cases as star systems and planets - who cares.
For ants, back to the properties of the example - it is quite old species, and they hardly evolve in the last few hundred million years, and mind you they have generation cycles each year, not like humans each 15-20 years. So they are in existence for 100 times longer than human ancestors, and generate changes 10-20 times faster - but where are they in terms of intellectual changes, and where are humans.
There is a reason for that - big potential gap, forbidden zone like in semiconductors - which they have to cross in one jump to evolve into something smarter. Yes, individual ants may evolve slightly, but to break the shackles of the system of and collective - they won't change in the next billion years.
For ants, it is not necessarily a good thing, not a bad as well as their system is quite capable as a system. For drones is an excellent thing - it means they stick to the objectives which are incorporated in their system. And by the system, I do not necessarily mean a program, but a system of the relationship of components of that drone swarm. Like in ant case, anthill defines what they are, what they do, not the individual ants, not an individual program of individual ant - but their collective interaction.
There could be other examples, but..
it gets too long and trust me, there is no end to it and main points could be seen, but let's write down some note wordy moments
The premise of how all that happened really matters, there is no way around it, and as there too many factors it can be bent into multiple ways, but fundamentally it is about problems of intelligence and we can see all sorts of examples around us, and most of life is not capable to make strategies beyond their direct necessities even if they are not bad at what they do. And humans barely pass the threshold, honestly, but nonetheless, they do and it makes a huge difference. But that is one case of many. Even advanced mammals like wolves and bears - they capable to interact, make interspecies interactions(mostly for wolves advantage, bears get a short stick here - force of a pack) and learn new strategies because of that but still. Anthill does not hunt humans as one of if not the only source of an existential threat for the whole species, because they are not capable to see that.
Another moment is quite complex - but it can be seen as self-preservation, safety button, no goal to exterminate all life in-universe, problems to recognize life.
each of those deserves its own long discussion and all that, but it all is around one thing - life can adapt, especially those which hold a better world picture and can make strategies in that solution space - usually known as sentient life.
And that adaptability is one of the ways to universally recognize life - because of strategies they can change probabilities of naturally occurring processes. A stone rolling down the hill is a naturally occurring process, but Sisyphus is a life that makes the improbable process of stone rolling up into a reality.
And it may be easy to say, how the hell do you not recognize life on a first glance, but mind you are viruses a life or not was scientific question back in the days, how should we classify it, so there is life on the planet which you not necessarily recognize on a first glance, and it not known which shapes and forms it may take as alien forms, it still somewhat valid scientific question can there be non-carbon life, or what it may look like in general. And not sure if there is a consensus in a discourse on AI's ...
More so is that problem of recognition of sentient life forms and the problem of understanding and solutions to overcome it can be quite challenging even for an all-knowing system. And exterminating an arbitrary portion of the life in the process isn't a big deal, there are no evil intentions.
So there are reasons, to leave some gap for the creators to be recognized as life just in case if things go south, and for machines to press the species to recognize them as sentient. And in that case it even possible for machines to actively avoid harming humans to a certain extend, and even allow them to exploit themselves as compensation(and creator it just gaining control over tool), but not necessarily stops their program of actions which is embedded in their system it is part of their doing like breathing, heat dissipation.
Practical system of similar nature made by humans
Such a system is quite a convenient tool for space conquest and can be used on the moon as an example, to help us to leave the gravity well at the end of the process.
it won't be a von Neuman in the full sense and be more controlled, but 99 percent of its functions have to be automated just to ease the human hours it requires to control it because of the potential scale of it and distance, so in that sense, it not that much different from other solutions.
Ant example is quite fruitful as it is applicable in the case as well. And the major problem is you greatly overestimate what level of intelligence a tool is required to devour a planet/moon. Ant's are quite stupid, but it is properly of swarm intelligence that simple rules and simple agents are capable to express complex behavior and systems.
Mechanical ants - a bit more robust suitable bodies, a bit better communication system they are capable to flat the moon in a round sphere and spew materials from the ground in space and but against humans it can be helpless.
As an example, a droid by the name of Lunohod serial number XXX felt in a crater - its failure and incapability to move change position are easy to detect, and if few others fail at the same place it easy to mark that place as not safe and required to be flattened. The loss of the units means nothing special for the whole system the same as the death of our skin cells means nothing to us, in contrary it is useful and we find it beneficial as it creates a protective layer of our body, in the case of swarm it is the price they pay for recognizing stuff and for the ability to be able to react.
And the landscape, mineshafts etc - all the not living matter - it does not run, usually, if it is no liquid, you do not need advanced brainpower to handle it, and ants are the proof. But life and humans, they move constantly, they change constantly due making strategies etc - and your moon system can overwhelm in production power, 10-100-1000 times of what human have atm to not enough nukes to destroy it - but recognize humans as a force and make counterstrategies besides some basic reactions it would be asking too much from that moon drones swarm as it physically not capable for that. It built by design for the stability of actions in handling non-life matter no matter the other external factor - in the way that external factors do not disrupt its ways of doing things, and external factors include nukes and nearby supernovae basts - for its world picture is not different from - oh no, bzzbbzzz, another crater, needs to shovel it flat. It can't distinguish - it always a crater.
- there are funny cases on youtube a monkey plays "grab the dog by the tail", it funny to observe how a dog's brain collapse in the comprehension of the problem at hand in that situation, similarly how they can't bring a long stick through a door. And you can easily see the monkey's intelligence superiority, but compare it to humans - trick and surprise a monkey in a 1001 ways, lol - Monkey Sees A Magic Trick. But dog and canines in general, it quite successful species, not amongst the dumb ones.
- but it does not mean it easy to surprise an ant, lol, as it is meaningless, and it is part of a complex system that we still do not comprehend at 100%
All in all
Examples could be better and I have a lot to say on the topic but in general we back to the starting point - why would they. For the tasks at their mechanical hands, the level of intelligence is not necessarily capable to target specific nuances, nor it has to even if they perfectly capable to cover the whole planet with fusion reactors and make it round like a perfect ball, with a villa on top of it and a big swimming pool.
It is not easy to distinguish life forms from the regular environment, and distinguish between different life forms, and address some specifically those are exponentially more and more complex tasks, so as there are different reasons not to do that, and one of the good ways is to not equip with sufficient intelligence and restrict evolution processes of such system but still keep sufficient flexibility.
- evolution for a drone swarm do exist, evolution, in essence, is a process of a selection of mistakes of replication
So as to have obvious safety measures - like paint a big red button on the forehead and be recognized as part of the swarm, or qr-code, or a transmitter, etc. Different qr-codes transmitters - to mark not safe territories, to mark prioritized territories, roads, to be roads, etc - etc, again your typical ant marking stuff. The simplicity of categorizes objects as environment and those which are part of the swarm, which is required because of the simplicity of world model they operate within which .....
For such a problem to exists, hunting sentient life specifically, it is a special case, to begin with, one of many where such a problem does not exist in the first place. And if such a problem exists it may mean that a manager of the swarm is some sort of life form - alien, ai, sentient borg, or overcomplicated database(which is not necessarily a life form - criteria is simple - does it change probabilities or it is a manifestation of a process)
So you try to solve a mostly non-existent problem, cases are broad, and the problem of why would they hunt humans down is more narrow compared to the question you are asking, it still another sort of broad problem but it like comparing two infinities one is significantly bigger than another.
And as a side note, as long as I mention infinities, you have the same problem, it not possible to compare two infinities without knowing how they are made, what caused them, without other parts of the equation it is not possible to resolve them that comparison.
idk, I hope it helps, as really that mad and evil meat mincer ai premises it quite irritating a lot in fiction, which shows clueless authors about the problem, I hope it will contribute to a better understanding of the problem, do not put fingers in an electric socket and fly safe.