The future of warfare - robot or nobot?

[Viking]

If it were that easy to make artificial intelligences who are more capable and more intelligent than we are side with us, then we would not need more of them to defend ourselves against the ones trying to kill us...

Or they didn't put in safeguards, or safeguards that were adequate.

Which was my point in the first place, that an AI can make its own decisions and learn about/come up with its own concepts. What you are describing are not AIs but merely machines as we have them now. An AI in a computer game so far is not a real AI, it is more like a series of scripts that pretends to be clever.

An intelligent AI isn't magic; it's a machinery based on logic. What you need to do is to put in blocks and limitations in its foundation that prevent it from reaching conclusions you don't want it to reach. The exact implementation of these things would vary from AI to AI.

Given that we have designed them, we can test their software over and over and discover the most probable causes for a rogue AI. Several AIs going rogue at the same time from a little-probable cause is of course improbable, so unless they can really efficiently convert other AI units to their cause (which we could put extra blocks in to avoid), we'd be dealing with rare and isolated causes of AIs going rogue, much like is the case for humans within our own societies.

[Husar]

You seem to think that software development or AI development for that matter is some kind of exact science, that the way it develops can always be predicted and that the complexity can be handled and the software be free of weakpoints. Yet you don't even seem to trust armor made of metal sheets to be free of weakspots. But surely a complex multi-million LoC AI software can have no loopholes, weakspots or other issues?

As Montmorency pointed out, we are also just machines, and millions of years of development have not made us flawless or been able to remove our issues, malfunctions and other problems of the complexity in our algorithms. And since we are so flawed, how could we be expected to create flawless machines?

[Greyblades]

Eh, that's not a given. The molten metal would have a hugher air resistance and the parts that melt first would be shaven off the rest of the metal, making your metal slug smaller and smaller until not much is left. I would also think for a stream of molten metal to be that dangerous, it has to be even faster than a normal tank gun round and very concentrated as in a shaped charge (HEAT round).
The muzzle velocity (and therefore not the terminal velocity on impact) of a modern tank projectile is around 6,000km/h whereas according to Wikipedia, the stream of a HEAT round travels through the metal with up to 25 times the speed of sound or >30,000km/h, which is at least around 5 times faster than your molten projectile would be even if it did not disintegrate during the melting process due to air resistance. And don't underestimate air resistance, one countermeasure to HEAT rounds is spaced armor because just a little bit of air between two metal plates will seriously disrupt the stream of molten metal, making it a lot less effective.

Hrm, you have a point, though I would point out that a lazer capable of heating a sabot shell to the point where it is rendered harmless while the shell is going at supersoinic speeds would require both the space and power only found on naval craft.

]As for the size, I guess we are close:

Eh that's a drone not a self thinking robot, which is what I originally thought you were expectng the mecha-mosquitos to be. When I try to imagine these mini bots working I imagine a bunch of drones commanded by a central carrier robot the size of, say, a car, and it becomes a lot more feasable.

[Husar]

Hrm, you have a point, though I would point out that a lazer capable of heating a sabot shell to the point where it is rendered harmless while the shell is going at supersoinic speeds would require both the space and power only found on naval craft.

Let me bold the relevant part:

As for a laser stopping those, of course it can, depends on the power of the laser.

Eh that's a drone not a self thinking robot, which is what I originally thought you were expectng the mecha-mosquitos to be. When I try to imagine these mini bots working I imagine a bunch of drones commanded by a central carrier robot the size of, say, a car, and it becomes a lot more feasable.

Well, yes, it's a drone for now. At the moment we also have not developed the AI that we are talking about, otherwise the pledge to stop making it would be a bit late. Whether the AI will later fit into such a tiny robot remains to be seen, IBM wants to propose neural chips for smartphones and you have my condolences if your smartphone is the size of a car.
Besides, they could also employ/give a new meaning to swarm intelligence since they'd obviously be networked.

[Greyblades]

Let me bold the relevant part:

[Gilrandir]

As Montmorency pointed out, we are also just machines, and millions of years of development have not made us flawless or been able to remove our issues, malfunctions and other problems of the complexity in our algorithms. And since we are so flawed, how could we be expected to create flawless machines?

We have our emotions in the way. Machines don't.

[Papewaio]

The chances of creating an AI that would be harmless is very high.

Just look at how few zero day exploits exist or patches software require.

The next time you think you could have a tame AI read up about the Xbox red ring of death or the Windows blue screen of death.

If we could create perfect software there would not be IT helpdesks fixing issues.

[Viking]

You seem to think that software development or AI development for that matter is some kind of exact science, that the way it develops can always be predicted and that the complexity can be handled and the software be free of weakpoints.

Uh, I just said the opposite.

we can test their software over and over and discover the most probable causes for a rogue AI

we'd be dealing with [...] AIs going rogue

[Husar]

Uh, I just said the opposite.

That you say it does not make it correct or in any way realistic.

[Viking]

^ That post fails the Turing test.

[Husar]

^ That post fails the Turing test.

Is that a personal attack?

How long do you think it would take to create a software that has no bugs?
And why would removing only the "most probable causes" be enough?
How much do you think it would cost and who would be willing to pay for that?
Wht if the unfinished, buggy AI already gets leaked and used by others?

[Viking]

Is that a personal attack?

It's to say that the post makes perfect sense without context, but in the actual context, it doesn't make much sense.

I say we cannot create a perfect AI, then you say me saying it does not make it true, even though you've been saying the same thing just above.

How long do you think it would take to create a software that has no bugs?

That's almost certainly impossible, as I've already hinted to.

And why would removing only the "most probable causes" be enough?

Enough according to what? If you can remove the most probable causes, then the time that passes between rogue incidents should be long; possibly much longer than corresponding times for rogue humans (i.e. spree shooters etc.)

How much do you think it would cost and who would be willing to pay for that?

You might as well ask me to predict the future.

Wht if the unfinished, buggy AI already gets leaked and used by others?

Then we'd have to destroy or quarantine the robots that implement it.

[Husar]

It's to say that the post makes perfect sense without context, but in the actual context, it doesn't make much sense.

I say we cannot create a perfect AI, then you say me saying it does not make it true, even though you've been saying the same thing just above.

I see what you mean, maybe I was not clear enough, I meant testing it long enough to find even the most probable causes and "just preventing them" is not easy just because you my think so. In the end it may not even be an AI anymore because you completely restrict its ability to think for itself, IF you can find the will t build in enough restrictions. It would probably be easier to just program each function and not let the machine think for itself.

That's almost certainly impossible, as I've already hinted to.

Well, see above.

Enough according to what? If you can remove the most probable causes, then the time that passes between rogue incidents should be long; possibly much longer than corresponding times for rogue humans (i.e. spree shooters etc.)

How do you remove the most probable causes in something that can develop in almost any direction and can think for itself? In a network that can shift functionality from one area to another and that develops based on inputs long after you have put restrictions in place? As I said above, it may be possible but at some point it is not an AI anymore since you basically took away all of its abilities that made you call it an AI in the first place.

You might as well ask me to predict the future.

The idea was merely that it would cost a whole lot since an AI is supposed to be really complex and able to develop in a lot of ways. At least what I would call an autonomous AI. To predict the possible errors, you have to predict all the possible developments and then you have to either put certain restrictions in place for a bazillion of possibilities, if the system even allows for this (the conditions could become really complex if e.g. the billion neurons all have to be in a certain state, yet you may not even know which state some of the neurons are in, or not before the error has already occurred). Some of it may also depend on whether you have a neural chip or a virtual neural network in a more traditional computer. If the neurons are saved in a memory, you may be able to check the state of the system more easily than if you just have a chip that generates an output based on inputs but does not give you access to intermediate steps, where errors might occur.

In other words, I can see that this would potentially be a really, really complex issue to solve, which may also be why creating such systems is usually not done in a garage so far. Also fixing an error in one place could cause errors elsewhere, meaning you may have to test everything again once you fix one potential problem......

Then we'd have to destroy or quarantine the robots that implement it.

So we invade the Middle East again, this time fighting against autonomous killer robots?

[Viking]

How do you remove the most probable causes in something that can develop in almost any direction and can think for itself?

As you would do in regular software development: testing. You would have to try to include all possible input to the software and see how it responds. Certain conditions would lead to hostile behaviour against humans, and then you patch the software so that it doesn't happen and re-run the test from scratch.

It's impossible for me to describe this in convincing detail; that would be a task for a subfield of computer science.

As I said above, it may be possible but at some point it is not an AI anymore since you basically took away all of its abilities that made you call it an AI in the first place.

An intelligence does not cease being an intelligence just because it refused to consider certain scenarios. Many or most humans would fail this criterion.

it would cost a whole lot

That's industrial and technological development in a nutshell.

[Husar]

As you would do in regular software development: testing. You would have to try to include all possible input to the software and see how it responds.

All possible input is earth and everything that can happen on it, good luck with your testing.

It's impossible for me to describe this in convincing detail; that would be a task for a subfield of computer science.v

I know how software testing works, but we are talking about an AI here, not MS Word, there is a difference.

An intelligence does not cease being an intelligence just because it refused to consider certain scenarios. Many or most humans would fail this criterion.

The AI I'm talking about has a neural network, a neural network learns from experience. It might be quite hard to know where to block something because you may not even know where exactly that decision comes from. If your friend keeps poking you wih a stick, which of his brain cells would you affect and how in order to make him stop? Preferably without affecting any of his other functions.

[Papewaio]

You cannot test for every scenario as that would have to predict every interaction with every human and every environmental situation (the entire ecosystem).

An AI of ant level might be containable. However I doubt an ant farm of said AIs would be.

[Gilrandir]

If your friend keeps poking you wih a stick, which of his brain cells would you affect and how in order to make him stop? Preferably without affecting any of his other functions.

No need to affect the brain. Take another stick and poke him back.

[Viking]

The word try was crucial when I spoke about testing. Even relatively simple programs can have so many possible states that you can't expect to be able to test them all. By intelligently choosing the tests, they could become very decent. Intelligent software could help structuring such tests.

Implementation of the AI itself would also act as crucial testing. The first robots to implement them could have limited lifespans hardcoded into them. First 1 hour, then 1 day, then a week, then a year, then 5 years and so on as the operators grow more confident. Indeed, x years could be the upper age limit for such robots required by law, after which their software should reset itself automatically (before this, a copy should have been downloaded for study ) - and then perhaps the robots and their memories should be destroyed manually by humans.

Another external software could study the intentions of the AI before it was able execute anything - i.e. all orders for physical movement, communication or other interaction with the external world would have to pass through this software. If it sees anything suspicious, it could shut down the robot and raise alarm. Such software could also be very intelligent.

[Husar]

The word try was crucial when I spoke about testing. Even relatively simple programs can have so many possible states that you can't expect to be able to test them all. By intelligently choosing the tests, they could become very decent. Intelligent software could help structuring such tests.

Implementation of the AI itself would also act as crucial testing. The first robots to implement them could have limited lifespans hardcoded into them. First 1 hour, then 1 day, then a week, then a year, then 5 years and so on as the operators grow more confident. Indeed, x years could be the upper age limit for such robots required by law, after which their software should reset itself automatically (before this, a copy should have been downloaded for study ) - and then perhaps the robots and their memories should be destroyed manually by humans.

Another external software could study the intentions of the AI before it was able execute anything - i.e. all orders for physical movement, communication or other interaction with the external world would have to pass through this software. If it sees anything suspicious, it could shut down the robot and raise alarm. Such software could also be very intelligent.

So basically you have a lot of coulds and know better than the thousands of experts on the subject who seem to think that all your coulds are still too dangerous for us to even go there?
How about we implement such AIs with machines that aren't per default equipped with weapons first and see how that works out?
Which I think is the indirect message as none of the experts seem to have promised to stop AI development completely.

[Viking]

Making predictions of similar quality to 'experts' on something that does not really exist yet is not very hard.

To my knowledge the current armed robots are not designed in such a way (nor advanced enough) that they can go rogue in a spectacular manner; like a Korean one that is immobile and mounted; so there is no immediate danger of robots killing us all.

There's also a difference between putting requirements on development and banning these things outright.

[Husar]

Making predictions of similar quality to 'experts' on something that does not really exist yet is not very hard.

To my knowledge the current armed robots are not designed in such a way (nor advanced enough) that they can go rogue in a spectacular manner; like a Korean one that is immobile and mounted; so there is no immediate danger of robots killing us all.

There's also a difference between putting requirements on development and banning these things outright.

Current robots do not use an actual AI, what they use is more like scripts that allow them to do certain things. If anything not covered by their script comes up, they do not know what to do, do not react at all or reply with a default. An AI would/should be able to react even to situations it was not prepared for. The idea to make future robots completely self-sufficient using AI so that they become as flexible as actual soldiers seems hardly far-fetched. And to just dismiss people who know a lot more about the field than you obviously do is also not the greatest idea IMO. You even seem to question that they are experts....did you even look at who signed it?

[Viking]

And to just dismiss people who know a lot more about the field than you obviously do is also not the greatest idea IMO. You even seem to question that they are experts....did you even look at who signed it?

No-one has been dismissed.

If you wanted to know about cars, who would you rather ask - an automobile expert from 1897 or a random guy you found on the street today?

[Myth]

I would much preffer it if we started a second space race.

[Husar]

No-one has been dismissed.

If you wanted to know about cars, who would you rather ask - an automobile expert from 1897 or a random guy you found on the street today?

I do not see how that relates to the topic.

If you wanted to know about Intel processors in 2020, who would you rather ask - the Intel engineer who is working on their design today or a random guy on the internet who claims the intel designer has no idea what he's talking about?

[Viking]

I do not see how that relates to the topic.

Basically a variant of what I already wrote: how can you be an expert on the future of something that at worst does not exist; or at best is in its infancy.

[Husar]

Basically a variant of what I already wrote: how can you be an expert on the future of something that at worst does not exist; or at best is in its infancy.

There are people who have studied artificial intelligence since the early 80ies or so. And they do apparently think that we are close enough to actually having AIs that it is time to warn of them. I also tend to know some very basic things about how computers can "think" and these basics give me the impression that they do indeed have a point. If even I can see it, these experts probably know a lot more about it. Your comparison was off because it assumed a 100 years+ gap and infancy of the subject. But as I have said, AI has been studied since the early 80ies and companies are developing neural chips already, there was even a robot that could participate in jeopardy, I would not say it is in its infancy anymore.

And I find it especially weird given that you have not demonstrated a lot of knowledge on the subject so far, that you proclaim other people have no idea. In fact I suspect that you are trolling.

[Viking]

there was even a robot that could participate in jeopardy, I would not say it is in its infancy anymore.

That jeopardy robot is probably going to kill us all.

I never said that there is 'no point' in that line of argumentation. The co-signed statement in the OP does not mention robots going rogue and killing us all, so I am not sure in which context you are referring to experts now.

[Husar]

That jeopardy robot is probably going to kill us all.

I never said that there is 'no point' in that line of argumentation. The co-signed statement in the OP does not mention robots going rogue and killing us all, so I am not sure in which context you are referring to experts now.

Indeed, it only says we should not develop AI for them because that is likely to be dangerous and may kill the wrong people.
I did get a bit sidetracked with the rogue thing because the entire argument sort of got sidetracked in that direction in case you didn't notice.

[Papewaio]

The murdering robots won't kill off humanity, it will be the robots that are programmed to love us...

Those ones will kill off many future generations of geeks... Whose potential parents can't reproduce with their robot girlfriends...

[Kadagar_AV]

Drones will NOT be very functional against a modern enemy, with the hardwiring they have today... We can just swap them like flies.

With that said, technology moves on...

I don't remember the exact quote from Starship Troopers, but in my own words:

Maybe one day we will have evolved bots advanced enough to go down a hole, separate civilian from enemy, and rescue the former and kill the latter...

But until that day has come, the military need guys like me.