Can a Combat-Grade AI feasibly exist with early mecha?

[Omega]

Hello, everyone... I have a peculiar question.

I am working on a mech-based RPG Maker Project called 'Mecha Cross War' which is supposed to be a generational SRPG. But the purpose of this topic is related to the first generation, where mecha are still in their early stages.

You see, in my game, mecha have been around for just two years, and they share the battlefield with conventional vehicles, which are on their way out. However, in those two years, one mech has consistently proven itself superior. This machine is called 'Anima'. It's design is meant to be intentionally simple, armed with just head vulcans, missile pods in the shoulders, two heat knives, and an assault rifle. However, the main thing that's supposed to allow Anima to stay on top, is it's defining feature; It has a combat AI that allows it to react with such speed and precision, that even the best of pilots are afraid of running into it.

But the question is; Can such a feature reasonably exist among mecha such as the Zaku I, Leo, and Fanton?

[Geoxile]

I don't think there is any real disparity between mechs and AI that would make a combat-AI impossible. As long as it's consistent with the rest of the technology in the universe I don't see why mechs can't have combat AI.

[Ferrus_Manus]

The HRL of Gundam 00 built mobile armours with AI-control, but I don't know if they implemented the technology into their Mobile Suits.

[Brave Fencer Kirby]

They're two very different kinds of technology -- building mecha requires advanced materials science, while programming a smart AI requires advanced computer science. It'd be more convincing if your setting already had AI in use in other venues. For example, there's no reason why you could have AI-controlled tanks before you had AI-controlled mecha. If you're going to say that your mecha and your AI were developed at the same time, then you have to come up with a reason why the two innovations are linked -- ie, why they couldn't have mecha without AI, or have AI without mecha.

[Dendrobium Stamen]

If you're going to say that your mecha and your AI were developed at the same time, then you have to come up with a reason why the two innovations are linked -- ie, why they couldn't have mecha without AI, or have AI without mecha.

On that note, you could quite happily justify the former with the argument that operating a mecha would be too complex for a single human pilot, without AI support. It'd be like the early Cosmic Era setting, only without Coordinators with the multi-tasking capability to operate an MS in combat; the AI would fill the gap, in this instance, much like the Natural OS did for Atlantic Federation pilots.

One could even reverse it to justify the latter argument; AI was created more-or-less by default while crafting an OS capable of handling the complexities of mecha operation.

Either way, if it's an original universe, there's nothing stopping you from creating a world where AI and first-generation mecha exist side-by-side, as long as there's sufficient world-building around both to justify their respective existences, of course!

[Wingnut]

I should note that in the story there will be mobile doll grade AI mechs later on so it's a bit of a fine line as to how much "power" the AI can have early on. Something more real world like programmable drones with limited improvisational ability might fit as a starting point.

[Kuruni]

One note, Layzner use drone mech in interesting way. The Skull Gunner's AI isn't so advance. They're mostly used in simple seek-and-destroy mission, take no prisoner. If they want to use it alongside manned SPT, then they have to be remote controlled to avoid friendly fire. Its predecessor Terminal Police use alternate method by scan target first to identify friend and foe, which actually leave it open for a moment.

[Omega]

Hm... thank you all for the advice, I'm really grateful to those who posted in this thread.

[Outlaw]

Little off topic, but what version of RPG Maker are you working on?

P.S. I googled your project and I'm pretty sure I've found you guys. If so, it's lookin' pretty good so far, I hope we get the chance to play it.

[Omega]

I'm building it in RPG Maker VX Ace, Outlaw. Once I have a playable demo going, I plan to put a topic up with that. XD

[MythSearcher]

They're two very different kinds of technology -- building mecha requires advanced materials science, while programming a smart AI requires advanced computer science. It'd be more convincing if your setting already had AI in use in other venues. For example, there's no reason why you could have AI-controlled tanks before you had AI-controlled mecha. If you're going to say that your mecha and your AI were developed at the same time, then you have to come up with a reason why the two innovations are linked -- ie, why they couldn't have mecha without AI, or have AI without mecha.

In our world, building mecha would require much more electrical and computer engineering advancement than material science.
A seven degree-of-freedom robotic arm already need pages of small prints of equations written correctly to get things to move right if you are doing things the top-down method, and the bottom-up method(letting the robot test out everything by itself) requires a very long period before it can do the simpliest motions knowing what it can do.
A full mecha that is combat ready is basically development hell no matter which method you are taking, on the other hand, quite a lot of materials can already be used on it.
(Sad to say, my major, mechanical engineering, is too developed and basically not really helpful in today's robotics engineering, one simply need a computer program to cover the mechanical parts which the other engineers are not that familiar with but would still have a lot of basic knowledge since its just applied Physics.)

The problem of a high AI in real life is that we cannot get a fully logcial program to think on its own. Thinking and creativity do not come from logic, doing the most logical thing on the battle field usually leads to being fooled. And there is no way you can convince the great people(who are funding your research) that your robots will do different things on the battle field in the same situation, even if it means preventing the logic pattern being decrypted. Because that means not all of your robots are doing the most efficient and effective action, and that seems like they are not doing the best thing, and military operations require most efficient and effective actions.

[Kuruni]

The problem of a high AI in real life is that we cannot get a fully logcial program to think on its own.

Hmm...we're actually pretty close, really.

The 5 Most Terrifying Robot Advances in Recent History from Cracked.com

[Outlaw]

I'm building it in RPG Maker VX Ace, Outlaw. Once I have a playable demo going, I plan to put a topic up with that. XD

Cool, I recognized some of the default assets and figured that was the case.

[MythSearcher]

Hmm...we're actually pretty close, really.

The 5 Most Terrifying Robot Advances in Recent History from Cracked.com

Making computers that look like having a logical mimicy of some known situations are nowhere near getting a high AI.
What they are doing are emulations, not actual success.

The so called "Schizophrenic Computer" does not even know the words it is telling or been told, it is just randomly giving you answers according to the things you tell it.

The robot that "Lies" basically is just choosing a logical answer to fool other robots that it knows what their logic are. And to be honest, it still can't get out of a simple logical pit fall of all AI logics nowaday, it cannot adapt to things that are designed beyond its knowledge frame. It can fool other robots in a "preset" field, not in "randomly created" ones.

The "Ruthless" robots are pretty much the same. They are told to do a logical selection of things, and they ended up following that logic, what's strange about that? And not to tell, they are not just fooling each other, but being fooled and not stopping to follow the lights. Which fulling match what I am saying up there, any AIs using logic will usually be fooled. If they can't "learn" to know understand deception or at least become doubtful, they can't be put on the battefield with humans, which has all the rights to think outside the box and be fully adaptive.

The "imagineative" computer, I must say, is just creating a pretty common merged image of multiple cats. Technically it is just taking similar pictures of the same kind(with the logic written for the google image search function) and merging them, which is simply a 90's tech. Saying it looked for cats is also just ignoring the fact that an ad company did a statistical research and found out 70% of the most popular youtube videos are about cats. If you randomly pick out videos from the top list without search keywords, you are highly likely ended up with seeing a lot of cat videos, that's not picking, but simply a probability issue.

The prediction computer, I must say, is even more simple logic. What it is doing is just using an uncanny amount of data(notice I did not use the word information) that human beings cannot even hope to read through, and find the common value in such data. The meaning of this? if you have a robot on the field doing this with not enough data, it is not useful and the prediction will be completely useless and inaccurate. If it has a vast data base, it is still going to fail to predict an uncommon action(like doing something strange to fool it) because most of its data are telling it that you should be doing something else, and the little data it just newly acquired is minimal compared to the large database.

[Strike Zero]

if you have a robot on the field doing this with not enough data, it is not useful and the prediction will be completely useless and inaccurate. If it has a vast data base, it is still going to fail to predict an uncommon action(like doing something strange to fool it) because most of its data are telling it that you should be doing something else, and the little data it just newly acquired is minimal compared to the large database.

I am reminded of the automated assistants the police carried with them in Demolition Man.

"Approach, and repeat ultimatum in an even firmer tone of voice. Add the words, 'or else!'"

[Kuruni]

Well, may be not close, but it's start anyway.

Plus, when I think about it, combat AI should be "limited" to process data and carry out its mission anyway. My impression is that , AI with free will and personality will be as unreliable as we are (ultimate AI won't act by our order, instead it will build a lesser AI to does it, then it will sit down and run an application that's equivalent to marijuana).

[MythSearcher]

I am reminded of the automated assistants the police carried with them in Demolition Man.

"Approach, and repeat ultimatum in an even firmer tone of voice. Add the words, 'or else!'"

Well, if you only want a robot for a short mission and all it does is "kill everything that moves/in sight within a specific time frame" for a matter of minutes or hours at most, you don't need an AI.
Aim, shoot, make sure kill, that is all your logic needs to do.

The problem comes when you need it to do more.
From a logical deduction point of view:
1) If robot can communicate with HQ:
1A) If you can crack the communication encryption -> tell it to stop firing and capture it intact.
1B) Else -> turn on ECM, stop all its communication -> go to 2)
1B ex) If you can't mess with its communication, you picked the wrong opponent, your enemy's tech is much more advance than yours.(messing with communication for a short period requires a lower tech level than keeping the communication channel well)
2) If robot cannot learn:
2A) Find something cheap yet the robot will determine it to be hostile and will open fire, like a balloon(filled with hot air to mimic heat pattern), foam board or asteroid(if in space). -> throw enough of these into the robot's range, waste its bullets, when its spent, capture it intact.
2B) If robot does not attack cheap things: go to 3)
3) Create a bigger version of the things it determined not hostile and will not fire. Test it and confirm that. -> disguise as it and get close enough to your own range, plant bomb or anything destructive enough, destroy it in one shot or at least damage it enough so it stops functioning, capture the remains.
4) Else(i.e. robot can learn)
4A) Find something cheap yet the robot will determine it to be hostile and will open fire. -> keep doing it until the robot stops firing, so it learns that it is not hostile, hide inside one and get close enough, you know the drill -> destroy.
4B) Robot somehow open fire again, keep moving different things in its range until it runs out of ammo. -> capture it intact.
4C) Robot retreats before using up ammo -> you won this battlefront, move your defense/offense line forward.
5) Once getting above results, you've learnt the logic pattern of all the same model robots, repeat and capture/destroy/force retreat all robots, you won the battle.

Well, may be not close, but it's start anyway.

Plus, when I think about it, combat AI should be "limited" to process data and carry out its mission anyway. My impression is that , AI with free will and personality will be as unreliable as we are (ultimate AI won't act by our order, instead it will build a lesser AI to does it, then it will sit down and run an application that's equivalent to marijuana).

The problem is that we don't even have the logical theory of doing a free will AI that can perform well in battle, and it is very unlikely that we can get one smart enough to do so since it will be far too unstable and economy and efficiency reasons will bar them from getting on the field.

Being able to get oneself out of a logical trap requires a certain level of uniqueness, yet this means the robots are not doing the most effective choice on the field, and is very likely to be destroyed by more simple logic robots that are only used in short period of time and really easy to develop.
On the other hand, you can't have these logical robots running around for too long, they can be fooled pretty easily by humans.
So you ended up having no one funding the research since logically, you can't meet the demands of a military high AI.