# How would AI human detection differ in a synthetic landscape?

If AI based robots and humans were the only heat producing macro organisms how would the AI detect and thus humans avoid detection?

The only things humans do in this setting is be mass produced and put in a fight pit as well as return to their cells to recover. Food and water is given to them via nutrients injections and all food, waste, reproducing, and cleaning are taken care of by machines.

The only areas they would have as background to distinguish from would be a metallic grey and the brown/green farming area.

In traditional facial detection it is often avoided with the application of non organic often colorful shapes as face paint, but this appeals to a cityscape with things such as traffic signs that would not exist in this setting

on an additional note, how would they be able to determine living from dead for clean up? and if magical regeneration is possible, would self mutilation be an effective strategy to avoid recognition or at least detection in terms of being singled out from the group?

I am dealing with hard robots here in the sense that they can't develop emotion or sentience or anything simply following code.

• .... uh, who's making all the humans fight? Why? If the robots, what purpose does this serve? Dec 17 '20 at 0:08
• The construct was set up for the entertainment of a civilization but they are long gone, the ruling nobility that over looked this project's degrading skeleton is still up there over looking, but is outside of the robots cleaning and stopping escape designated area. The robots still self repair and sustain because they are coded to do so regardless if there's a practical reason. Thus, your question is essentially highlighting the irony here that in truth I probably should have clarified, there is no practical reason anymore and that this simply is because the operation is self sustained. Dec 17 '20 at 0:30
• If they are the only heat producing organisms, then the robots would use thermal detection. Assuming all robots are connected on a global network they would broadcast their coordinates (detected by gps and lidar) to each other, allowing the robots to match the location of detected thermal signatures against all nearby robot locations. A heat signature that doesn't have matching coordinates must be a human.
– jla
Dec 17 '20 at 5:39
• I do like the concept of a GPS and I now fully intend to implement such a thing thank you! However under this model it does run the issue of firing at a hot piece of metal just because the sun shined on it, or a hot patch that's hot from firing at it. this would however open up the fun concept of confusing the AI with a fire but I digress. Dec 17 '20 at 8:24
• Side note: you need cameras to detect a lot of stuff (ie, fruit on trees). At that point training a detection model for "human" (note that you don't necessarily need to detect which human, just a human, although chipping your humans would help here) is relatively trivial. The real question then becomes: how good are your robots with dealing with something they haven't seen before? If I hide in a cardboard box, what do they do? Dec 17 '20 at 16:42

The AI might implant remote devices inside the humans. I mean, if they're breedings humans just to kill them for sport, it's not like they view the human body as inviolable.

Those devices could potentially do all kinds of things, but the applications that suggest themselves immediately are:

• location tracking of each human
• identity verification; not just "a" human, but HUMAN_967813451145
• remote incapacitate and/or kill
• monitoring of vitals

The only limits are size and cost (materials, energy, time). Sufficiently advanced tech could do just about anything, but they're not going to bother installing a Rolls Royce inside each of their cattle. The devices will be mass-produced, so they'll all be the same and they'll all be as cheap as possible (but no cheaper).

So:

• location tracking could be GPS, but wifi may be more practical
• incapacitate & kill are probably done by releasing some toxins, not by explosive
• vitals may be very basic: body temp as a proxy for health, motion as a proxy for consciousness or death -- think "evil FitBit," not "NASA astronaut"

The device is a key part of maintaining control over the slave population, and that means defending it is paramount. No rebellion can get off the ground as long as every slave is lo-jacked, and the cattle can't defeat the device if they don't know it exists. The need to prevent discovery forever probably dictates most of the design.

They need to be small enough to be implanted safely and without lasting impacts to health or mobility, and they need to be durable enough to not require frequent maintenance. None of that is for the benefit of the human: it's about convenience, cost, and tamper-resistance.

Ideally, the device is tiny, implanted at birth, and never needs to be serviced (or, not within the human's lifetime).

If every slave is fitted with a device at birth, there will be no problem detecting and tracking any human anywhere. Except perhaps some kind of naturally-occurring Faraday cage, which weakness could prove useful to storytelling.

• If placed in the chest cavity, would an audio sensor be enough for heartbeat and breath monitoring? (Good backups for the health and consciousness checks) Jan 17 at 11:41
• I'm not an expert but that sounds pretty plausible to me. It could also work like those Jawbone headsets, which apparently rely on vibrations conducted via the skeletal system. You can probably handwave as much as you'd like and say that the AI overhears everything spoken by any slave.
– Tom
Jan 17 at 14:40

The construct was set up for the entertainment of a civilization but they are long gone

Okay, but the robots don't know that.

The humans have a very simple disguise: they dress up as nobles. Historically, people have always wanted to "inspect" the merchandise, which means that nobles (or other people) would have visited the arena.

The robots see what they expect to - somebody in a noble's clothing.

• while that would work if they had a set sequence for detecting the nobility that wasn't a feature I intended to add, the reason the noble died up there was because once the construct was severed from the main section he was essentially trapped with the only way out being to go into the detection of the machines in an off limits area. there is also the logistic issue of obtaining such garments as they would only be in the off limits and beyond as well as at least a century old. (Don't be deterred though! I am loving these questions from you and I thank your enthusiasm.) Dec 17 '20 at 8:20
• edit: I guess its wrong to say at least a century old at the point of origin but where I intend to set it they would be, and even if it was applicable it still misses the point of the question somewhat being how would you identify as here it just adds the question of how you identify humans, and also humans from noble humans. Dec 17 '20 at 8:51
• @SIKMephisto - A certain minimum level of "competence" has to be assumed, or the robots wouldn't be able to do the job at all. Particularly if you have them doing anything like maintenance on other robots (who break down rather quickly). Note that "noble's clothing" could be almost anything - like a piece of fancy jewelry - and depending on environment even cloth can still be usable after that long. Dec 17 '20 at 16:39
• that still seems very tricky and slightly risky, all it takes is a second of misinterpreting for a bullet to be fired and how would they respond to say a necklace that would reveal nobility from a different angle? though I suppose they would need quality detection you are right there but clothing seems like a very buggy form of detection, perhaps a unique chip or something though. if we did want to go that route but it still would be difficult to obtain such items to go out of bounds without already being able to go out of bounds, Dec 17 '20 at 23:25