About Me

My photo
I have a burning need to know stuff and I love asking awkward questions.

Thursday, November 21, 2024


Just Finished Reading: I, Warbot – The Dawn of Artificially Intelligent Conflict by Kenneth Payne (FP: 2021) [256pp] 

Anyone who has been paying attention to the news has, no doubt, witnessed the explosive increase in drone use and drone power on the battlefield – not only in Ukraine but elsewhere across the world. Drones, it seems, are it. Coupled with drones are the use of highly sophisticated missiles and (quite possibly) even more sophisticated missile defences. It's all VERY impressive. The smart missiles gain their smarts from the ability to self-guide after launch and some can even pick a second target if they ‘realise’ the first target has been destroyed. Some missiles can even respond with counter measures if detected. What they can’t do, of course, is think. They can’t really ‘decide’ what to do – they're programmed with a set of instructions which essentially follow a ‘IF this THEN...’ structure. They respond to the environment, sure, but only in a very limited fashion. 

Drones, no matter how good they are, are by and large, human controlled, human directed and human deployed. Generally, their ‘intelligence’ without human involvement is almost non-existent. Yet they are, as we’ve seen time and time again on video feeds VERY combat effective. Because of this both anti-drone and especially anti-missile systems need to be VERY good and, most importantly, VERY quick. This means that humans can’t really be ‘in the loop’ - they can’t really be given the final authority to engage an incoming target. With as little as a few seconds to respond, humans are just too SLOW. Machines, aided with still rudimentary AI is much better suited to this sort of thing and can detect, confirm and launch a response in the time it takes a human to start to recognise a threat. With the way things are clearly going that’s a BIG problem. 

Enter AI (Artificial Intelligence) or at least the hopes, the hype, the advertised AI. Its proponents and manufacturers tell us, and their military customers, that AI can (at least ultimately) solve their very real problems – kind of set a thief to catch (or shoot down) a thief. The problem is, naturally, that putting AI into a weapon – especially a disposable/kamikaze one – is going to be a LOT easier that putting one into a system to defend against them. Again, this will mean removing or reducing the humans in the loop to save vital seconds but there’s a real problem that militaries all over the world are facing – AI isn’t really all that good despite the hype. We are a LONG LONG way from Terminators or SkyNet (thankfully). AI is good, often VERY good, in narrow clearly defined tasks but take it just a bit out of its comfort zone and it falls apart or simply stops ‘unsure’ how to proceed. Not exactly something you want to happen with a missile defence network during a shooting war. 

These issues, and much more besides, are intelligently discussed in this fascinating book. The author certainly knows his way around the topic, not only from the military hardware side of things along with the tactical and strategic implications of such, but also AI research, breakthroughs in understanding the human brain, Game Theory, and the Philosophy of Mind. It was, for ME at least [grin], highly engaging. About the only ‘fault’ I could find – and it certainly wasn’t a fault of the author – is that despite its recent publication the pace of drone, missile and robot development has already made some observations in the book increasingly obsolete. But, that’s the danger of reading books, which take an appreciable time to become published (even in my hardback edition), on such topics. Taking this caveat into account this is still a very good accounting of the recent impact of AI on military technology and thinking both at the tactical and strategic level. Military AI is here to stay and will become ever more powerful and more deeply integrated into military forces across the world. I was particularly interested in the ‘half-way house’ of human-machine teams (such as the loyal wingman programme adding drones to aircraft to enhance their combined effectiveness) which are called Centaurs! How funky is that! Definitely recommended for science, technology and warfare geeks out there.  

2 comments:

Stephen said...

Reminds me of Daniel Suarez's book "Kill Decision" -- it used drone swarms powered by AI that had been modeled on insect swarms. Creepy stuff.

CyberKitten said...

...and WELL on the way to being reality. Have you seen the 'Slaughterbots' video on YouTube? Absolutely terrifying - and probably achievable in the not too distant future.