Nick Carter proposes Robot Soldiers

Joshua Slocum

LE
Book Reviewer
Robots need specialist treatment, if we imbue them with AI they will start to learn from the soldiers
look what happens next
elektro-nurse-x640.jpg
 
Interesting essay on “Unherd” on this topic, which summarises the problems involved quite usefully:
 
Last edited:
This might be a movie, but the technology for this is all out there. People seem to be under the impression that drone-tech has only been around recently. Wrong. I was looking at research done at MIT and JPL around 25 years ago. The prime object of the research was to supply Govt agencies with the means to track, or hunt, man, or vehicle shaped objects. Even back then the better ones were using AI pattern recognition (face and shape) algorithms to differentiate between targets. I was doing my AI stuff back then at uni which is how I bumped into their research. I know that the algorithms have improved over the last 25 years, processor speeds have increased, and we all know video technology for image capture has definately improved. And, back then they were already working with drones that would comfortably sit on the flat of your hand - that sizing can also be squished down now with the advancements in battery and brushless motor tech.

 

Cold_Collation

LE
Book Reviewer
This might be a movie, but the technology for this is all out there. People seem to be under the impression that drone-tech has only been around recently. Wrong. I was looking at research done at MIT and JPL around 25 years ago. The prime object of the research was to supply Govt agencies with the means to track, or hunt, man, or vehicle shaped objects. Even back then the better ones were using AI pattern recognition (face and shape) algorithms to differentiate between targets. I was doing my AI stuff back then at uni which is how I bumped into their research. I know that the algorithms have improved over the last 25 years, processor speeds have increased, and we all know video technology for image capture has definately improved. And, back then they were already working with drones that would comfortably sit on the flat of your hand - that sizing can also be squished down now with the advancements in battery and brushless motor tech.


Jane’s was running stories about fractal recognition 30 years ago.
 
Jane’s was running stories about fractal recognition 30 years ago.

It was a right old beehive of activity in the 80’s as PC’s made access to computing easier. Probably the leading light in image recognition research in the UK was Igor Aleksander out of Imperial College, London. I think back to the kind of stuff that was being done by the megaminds then and it was like a baby laying there waving its arms and legs around. Nowadays, I would say, with the advances in capability that the baby is approaching being a teenager, running around and getting in trouble.

I was writing back propogation neural networks to data mine humungous data sets to predict consumer behavior - people are very predictable. The pattern/image recognition side was done using (Kohonen) self organising maps which was impressive to see in action, even back then. Nowadays though you can download better, faster software from the interweb.

It was the smart kids at MIT who really bought together all the aspects of technology to create drones. Pretty easy nowadays to have an explosive charge carrying drone and tell it that it should fly at, and detonate near, any human/vehicle* shaped object in a certain grid square.

Edit to add: With the use of image recognition/AI/smart systems you could theoretically (only theory cos it has not been seen to be done yet militarily) teach/program the drone to recognise prominent vehicle features, and/or particular movement patterns. Meaning that the drone would effectively know the difference between a Foxhound, or Mastiff, and the other sides GAZ. Along with, depending on imaging quality discriminate between the helmet, weapon and personal equipment of our sides infantry and the other sides.
 
Last edited:

Dread

LE
The funny bit is that the RAF Regiment don't man barriers!
You are of course correct.

Manning the barriers is too complex for the typical RAF regiment member and typically the barriers are too far from the nearest EFI/Pizza Hut/Sunglass shop.

If they did deploy robots to support the RAF Regt, would the robot have to be programmed to shoot its own armour (if fitted), or would it just use AI to learn from the example set by the apes around it when safely out of any combat?
 

exsniffer

War Hero
Given that, in multiple experiments, AI couldn't distinguish a child holding a baguette from an armed assailant, nor correctly identify an umbrella* then humans soldiers will be here for the foreseeable future. Automation and machine learning is perfect for mundane repetitive tasks where multiple data-points can be predicted and programmed to a correlating action - tasks requiring judgment will remain in the human domain.
Public Duties?
 
OK, I've looked and done a search and I can't see that anyone else has picked this up (if I'm mistaken, then Mods please edit).

Just seen this on the Grauniad website:

'Robot soldiers could make up quarter of British army by 2030s'​

Are we being quite serious here?


What we need are Mobile Infantry....
 

theoriginalphantom

MIA
Book Reviewer

HCL

LE
It was a right old beehive of activity in the 80’s as PC’s made access to computing easier. Probably the leading light in image recognition research in the UK was Igor Aleksander out of Imperial College, London. I think back to the kind of stuff that was being done by the megaminds then and it was like a baby laying there waving its arms and legs around. Nowadays, I would say, with the advances in capability that the baby is approaching being a teenager, running around and getting in trouble.

I was writing back propogation neural networks to data mine humungous data sets to predict consumer behavior - people are very predictable. The pattern/image recognition side was done using (Kohonen) self organising maps which was impressive to see in action, even back then. Nowadays though you can download better, faster software from the interweb.

It was the smart kids at MIT who really bought together all the aspects of technology to create drones. Pretty easy nowadays to have an explosive charge carrying drone and tell it that it should fly at, and detonate near, any human/vehicle* shaped object in a certain grid square.

Edit to add: With the use of image recognition/AI/smart systems you could theoretically (only theory cos it has not been seen to be done yet militarily) teach/program the drone to recognise prominent vehicle features, and/or particular movement patterns. Meaning that the drone would effectively know the difference between a Foxhound, or Mastiff, and the other sides GAZ. Along with, depending on imaging quality discriminate between the helmet, weapon and personal equipment of our sides infantry and the other sides.

Marks and Spencers used to Model 204 for predicting footfall and store layout in the late 80s, it was relational and had early "pattern" recog. It was creaky old tech even back then.
 

New Posts

Latest Threads

Top