Part of writing sf is predictivity. We are often writing about possible futures, looking at trends and… because we write books and need to, making them more interesting and exciting than they usually are. Also… wrong. It’s fiction. There are a million variables. The fact that sf writers get anywhere close, ever, simply makes them better than most think tanks or futurists, no great achievement.
Sf tends to be to some extent predictive even when it is straightforward space opera or satire dressed in sf. Oh, it is not the STORY that is necessarily predictive. Just, as often as not the technology and methodology of using it. Military sf is one the areas where this has been particularly true, I think. It’s not an area I write a lot in, but a few recent ‘prompts’ have had me thinking about it.
The invasion and subsequent war in Ukraine have brought drone warfare into sharp focus (as foretold by various authors). It’s been fascinating to see jury rigged $100 drones take out machinery 100 times as expensive. The use of drones in lieu of missiles (or bombs) to attack cities and power systems — as weapons against civilian infrastructure and of terror (also foreseen widely by writers) has been instructive. The use of suicide drones, and swarming to overcome defenses by sheer volume (both price-factor equations)
What we don’t see – but is very, very relevant, is the surveillance aspect. As a writer, I predict this is going to be the aspect that grows fastest and integrates with AI AND all of the combined arms. The fog of war may still continue, but will at least for a while – and for the rich and technologically advanced nations get quite a lot thinner. It’s where I would be spending serious money if I were a government wanting real bang-for-their-buck out of this. It’s also – as a writer – one of the most terrifying aspects… ESPECIALLY when the enemy discovers how to evade or ghost or deceive it, as that ends up with the one thing that no one in the military wants: surprises. It’s going to make for interesting times, because the nature of these things is a constant measure-counter-measure chase, each step getting more complex and expensive. And then… and then some hitherto relatively helpless irregular combatant figures out he can bring down your multi-million dollar system with a piece of tinfoil…
I keep seeing these ‘Oh they’ll know your every move of everyone on earth’ bits of sf. Well, um. Let’s say you did manage surveillance of all the people on earth. All the time. Have you any back-of-the-envelope calculations on how much data that is? And how much of it is irrelevant GIGO? How to sieve out the relevant? “Oh, computers and AI will do that!” Well, they will doubtless try. But the potential for false pattern generation, and totally innocent accidents being flagged and the actual misdeed/rebellion being missed – make a good story. It’s a bit like the mage with these enormous powers in a fantasy story… which are great and terrible, close up and proximal, but may be a dismal failure at world domination, because the world is just so BIG. You can affect it, you can affect wars, but it is a case of applying pressure at crucial points. It’s working out what those crucial points are that’ll be the trick (and yes, I have some ideas. No, I am not putting them up in public).
To return to the drones: I actually don’t think swarms will be the menace posed. You see, if they ARE to act in concert, they need either to have a large number of operators all in close, fast contact (so a target and needing a lot of practice). The alternative – to share instruction between themselves and to have one driver or pre-programmed mission)- means some kind of interconnection. The latter is a weak link. Intercept it or break it and the issue is quite different. Pre-programmed is more complex but limits any changes in target.
The areas that I see as ‘rapid growth’ and ‘fast change’ are miniaturization and sleepers. How the hell do you shoot down a drone that is the size of a moth? And sleepers? picking anything up that is moving and has signal is plausible. But when it is just sitting -silent, near shut down on (or in) the ground or a tree or on a roof – which the enemy knows where is – and he can activate a second or two before your troops or supply truck get there… quite hard to combat. Or just sends ultrafast data – and the moves and goes sleeper again…
I will bet I have done a terrible job of foresight – but there are several books in it…
This
Mad Genius Club 
32 responses to “Droning on”
I’m not sure that swarms needs complex intercoonectivity or loads of controllers. All you need is to have a few drones (maybe just one) that is controlled and have the rest be smart enough to track the controlled drones and follow/copy them. Like the sway flocks of birds or schools of fish self organize
Flocks of birds is a good example, Francis. Have you ever seen what happens to a flock of highly organized birds when you shoot the lead one? (I have. Corvids, highly organized, very intelligent, very organized birds. Able to respond quickly and effectively to threats and opportunities… until THAT bird got accidentally shot.)
On mass spying, I’ve seen infantry boys have entire classified conversations in dick jokes. AI is not going to pick up the subtext or be able to follow it as it changes, because it’s all improvisational.
That’s some real Darmok and Jelad at Tanagra stuff!
Similar is the way people lard ordinary e-mail with hot-button key-words that kluge up the systems. If, oh, thirty-thousand peeved people put certain terms into ordinary communications, the AI’s going to get overloaded. Then so will the humans that have to 1) clean up the AI’s pattern-search and 2) go through those millions of e-mails. And then keep up with the variations on the key-word.
Alternatively they could use a language the AI doesn’t have enough data in to enable it to adequately interpret what is being said – either use a language that isn’t well understood by the AI and its programmers (e.g. Navajo as used by the US in the 2nd World War had no Japanese/Axis speakers) or come up with a brand new invented language.
Do it well enough and the AI might not be able to understand any of the messages, not even the irrelevant ones, and the problem will have to be handed over to linguists, who may not be able to translate the messages themselves unless they can grasp enough of the contexts.
They already have that covered. Law on the books defines taking steps to thwart law enforcement (all withdrawals > $10,000 must be reported; withdrawing two chunks of $5000 to avoid that is called “structuring” and is also illegal); communicating in obscure languages will make the cut under our current legal “system”.
communicating in obscure languages will make the cut under our current legal “system”.
Factually false, you’d have to build up a new system to run that storyline.
It can currently be used to show people were aware of a crime being committed– such as listing contraband as something else, and third parties agreeing they’ve inspected it, which is either fraud or conspiracy– but being obscure is not illegal.
(Proving structuring is also rather tough–something as simple as ALWAYS pulling out $XXX each month because you like having cash on hand can destroy the case. Or pulling out a reportable amount because “afraid of run on banks”. But it has the built in context that cash is being removed.)
As AC Young accurately pointed out, the issue comes where there’s context so folks can figure out what is being discussed.
That is how you can find out that the “donuts” are the super secret smuggled documents, and the color of frosting for an ordered cake is telling you who is guarding the access port, and bread is only going to be cash if the characters no really cannot resist a bad pun.
Tequila Sunrise covered that scene.
I’ve seen comments about how energy expensive good AI is. I think that would play a part in how it can be used. As well, what happens if you destroy the data processing place…
It depends on how it’s implemented and trained. The Big Blob stuff does suck down immense amounts of power, but it is also possible to train and run much less power hungry AIs through more focused training.
Which you could use to train AI to generate chaff for jamming the big boy AIs. And I’d expect it to be easier for a big boy AI to generate and update chaffers than it would be for a big i to penetrate an updated and maintained chaff wall. Especially if there is more than one big boy contributing to a defensive network.
Make a system that’s really good at finding A, another that finds B, and so on– then take the feed-through and have other systems that look for patterns of those outputs, maybe?
It would really quickly get complicated enough to be very vulnerable to “correction” for desired results, which tends to be a big weakness in AI results right now. It’s hard to get objective results for close judgement calls, and a significant number of folks running the AI training don’t like some of the objective results. (Call it the “But that’s different!” effect?)
I keep seeing these ‘Oh they’ll know your every move of everyone on earth’ bits of sf. Well, um. Let’s say you did manage surveillance of all the people on earth. All the time. Have you any back-of-the-envelope calculations on how much data that is? And how much of it is irrelevant GIGO? How to sieve out the relevant?
You can already see this in action, too– yeah, the “known wolf” thing is so shopworn it’s more darning than fabric by now, but we really can’t keep an eye on even all the folks who have serious red flags. We also, thank God, have way too many folks with ‘red flags’ for just lock up on suspicion to be an option.
In my creepy future, I’ve got them identifying and pruning social tendencies which tend to disrupt the culture… but it takes individuals in order to make it work, because a lot of those disruptions are also needed to keep functioning.
Very dehumanizing.
Not just red flags, but known felonies and violent crimes.
As near as I can tell, all the surveillance state is able to do is hunt down and destroy people it has taken a dislike to, not actually deal with crime any more effectively than old fashion show work did.
Oh, that’s where the “fun” comes in– most of the ‘taken a dislike to’ sorts have all those red flags as well, they just can’t be mentioned when the trial is ongoing or it might destroy the trial by tainting the pool.
Although we have finally gotten to the point where releasing the whole bodycam footage with only editing for the faces of those not arrested isn’t treated as tainting the pool, usually.
“have all those red flags as well”
Donald Trump would like a word. So would the J6 political prisoners.
Wrong blog for that dance, Nelson.
There’s a reason I’ve repeatedly pulled it over to actual laws, using it in stories and the standard cases.
They can both be used to justify the kind of character you want your main character to be-
if you want a Punisher-Batman type, hyper-focus on the costs of the safeguards where people you know are guilty as sin can just get away with it over, and over, and over again, hiding behind procedure and investigation.
If you want a Paladin, Vimes-Batman type, then focus on the costs of not following those rules, all the harm done in neglecting those safeguards because you “know” someone is guilty, and “know” they will do something horrible– and then bring in the folks who want those safeguards gone, and how nasty that gets really fast.
That the stuff does not in fact deal with crime better is not proof that it could not. It could happen merely because they aren’t trying.
For that matter– the practical argument against pre-crime is that you can’t tell if that person would have actually done the crime.
If you arrest someone on a relatively minor but definitely valid charge, you can’t know he was going to [insert huge horrible thing here] the next day…unless you’re doing like Groundhog Day, Scifi edition.
That’s how they get people in stings looking to have sex with minors. They’re not charging them with the sex act that was probably going to happen. They charge them with carrying out the steps needed before the actual sex act occurs.
Futurists have incentives that encourage them to get things wrong.
If you say “x technology is a tool, and will be used and abused much like related tools have been in the past”, well, that’s boring. Nobody is going to line up to throw money at you.
But if you say “x will make the future unlike the present or the past”, you’re selling magic. And there are a lot of people eager to willingly suspend disbelief, even in the face of obvious flaws. (Take the fabled “post-scarcity economy”. How increased computational power somehow creates infinite energy, iron, etc. remains unexplained, since acknowledging the disconnect would make the dream fall apart.)
The most interesting theory I’ve heard in the past several years relates directly to AI and autonomous drones.
It held that the focus on those two technologies was an attempt to overturn the societal revolution created by firearms. After once any man could kill a king, absolute power became untenable. But if you could surround yourself with protectors that had superhuman senses, a practically instantaneous OODA loop, no moral qualms, or sense of self-preservation…
Well, that might be a rather attractive vision for certain megalomaniac wannabes.
That sounds like a bunch of journalists on pretty much every new development in technology and science.
I’m reminded of the old ways of sending secret messages by using references to A) Book B) Page C) Row D) Word.
It’s a little cumbersome, but still effective because without the Book in question you don’t know where to start.
That works as long as there are a many references possible for the same target word. If it is one-for-one the code can be broken the same way the Navy confirmed the codeword the Japanese were using for Midway Island.
Oooh, low tech trick to do that one– switch the word that you use for high value targets, based on the word in identical books.
So you make sure all your code guys have a copy of the same book– go with the Bible for ease of use.
And you have them use the tenth word that isn’t on a list (of things like is, was, were, I, a- the common stuff) on page 1 to mean that thing.
Then next message, they use the tenth word on page 2.
The guy sending messages back always uses the tenth word on the final page, and moves backwards.
I like this one better than the “have a rod of a specific size and a paper with holes in it so you can read the message” types.
If there’s a list of Important Locations, you have it written down and make a list of what words replace it this time.
Or you can go from memory, and have like “Books of the Bible attached to list in alphabetical order, move N down for each change in message”.
Gonna be a problem using the Bible … which translation? Catholic or Protestant? Or Jehovah’s Witness? Although maybe that’s just a fun way to get the code to be hard to break.
Ah, that’s part of why it’s a useful code!
You establish it with the folks you’re cooperating with– I mostly used the Bible because it’s easy to find one, and there ARE so many variations. Even the same translation in different printings won’t always be the same.
The problems with that, kamas, are at least 2:
1. How do you tell all your operatives which book to use? or to switch to?
2. If there’s any one thing computers are good at, it’s substitution cyphers, and books like the Bible have been digitized long ago.
Issue #1 is the same as securing a one-time pad or any cypher conversion key. If you lose the key you lose everything. PGP is popular because you don’t need any shared secrets, only the knowledge that only the private key holder can encrypt something that can be decrypted by their public key.
That’s why I thought it was interesting Peter G used references to shared events as keys in the Laredo War books and Jim C used a specific method of steganography known only by Nicole and her friend in Rimworld.
1. In person works pretty well.
2. There are a lot of books, and a lot of editions of books. It could be something common like the Bible, or the complete works of William Shakespeare. Or it could be something less common but easily available like a 4th edition paperback of Clear and Present Danger. Make it something esoteric like a first edition of Surfing Samurai Robots. Maybe you choose an edition in a foreign language that wouldn’t be obvious. Your book choices are endless.
Plus, you could switch it up, where a page number that doesn’t exist means something, plus it throws off whoever or whatever is trying to decipher it as they’re assuming the book has at least that many pages. Or a word number that doesn’t exist means something else. Having a code within a code is nothing new. The Navajo Codetalkers spoke Navajo, but not as in a normal conversation, it was in code.
Computers are fast, but they’re only as good as the information available to them.