We often talk about artificial intelligence and robots on this channel, of the potential
impact of intelligent machines on our civilization, but it's possible that dumb machines, little
flying drones, might have an equally big impact, for good or ill.
So today's topic is all about drones and robots in future conflicts, be it up in space
or down here on Earth, in wide open areas or the tight turns and corners of a building
or space station.
This isn't such a futuristic concept anymore either.
The science fiction of the last century has been dominated by robots, be it humanoid androids
or inhuman machines, and yet drones are no longer even slightly science fiction.
So I thought we should discuss some of the challenges and possibilities ahead for us
with drones both in the short term, and for potential use far ahead in space, particularly
in warfare.
That's a good place to begin, as military applications are a big part of what got us
our modern drones.
This is often the case of course, swords turned into plowshares, and the reverse.
Science fiction also has tended to focus more on drones for military purposes too, rather
than one delivering your pizza, so we're already quite well versed in the problems.
Hollywood has burned the typical scene into our retinas, swarms of drones killing humans
left and right, and that these drones do so autonomously.
Your typical modern drone is remote controlled, but we want to minimize that, so that most
functions are automatic or simply require a quick command to do it themselves.
This is even more important in the future, since signal time is a major issue.
In the first place, a drone attacking an enemy spaceship might be many light seconds or even
hours away, and in the second, human reaction times are on an order of a second, while a
computer system might be in the nanoseconds.
There's simply not enough time for a remote human operator to react to changes in the
situation.
Drones fighting drones seems the most realistic scenario since drones versus humans will tend
to be very one-sided, even if they're remote controlled.
Even assuming it has no advantage on dodging shots and aiming on its own, it's much easier
to replace drones than people, so you can swarm over someone: quantity has a quality
all its own.
But they probably can dodge better and aim better, a remote controlled drone is weaker
than an automated drone if you have good enough computers, it can't be jammed and can react
at machine speeds instead of biological ones, which are further hampered by signal lag time.
This doesn't mean everything needs to be automated, but the more features you can automate,
the better.
One that can detect a bullet headed its way and see if it needs to dodge or will be missed,
then react in a precise fashion is not that complex to make, nor something that really
needs oversight, so you offload that control to the machine itself.
Picking targets and deciding the appropriate level of force during an escalating situation
is another matter, but also one largely irrelevant to drone on drone combat.
Collateral damage is always undesirable, but the reason why collateral damage is often
a euphemism for killing bystanders is because that's the only collateral damage that really
bothers us.
If it's just drones fighting drones though, they can't afford human reaction times since
everything is happening too fast.
They also can't afford to be big: bigger is slower to react in almost every way, even
mental.
Give a drone more brains for decision making and you are making it slower, it could lose
a fight with a cheaper and dumber drone, or a glorified smart bullet, simply because that
one does not carry the hardware and software to identify a human or pick shots to minimize
damage from ricochets or misses.
That smarter drone might only need a single microsecond more to decide, but it still gets
shot first, and even if it kills its attacker in the engagement, you're still out one
expensive drone while your enemy is out a cheap one.
In other words, it can be advantageous to be dumb.
It's an irony, because we call them drones as our earlier unmanned aerial vehicles flew
along dumbly on a preset path, so got likened to a male bee, a drone.
I'd imagine, since drones only have one purpose in life and die when successfully
achieving it, that it also fit a lot of early drone vehicle concepts that are basically
a guided bomb.
The swarm or hive notion is probably rather apt too, as when you have a swarm of drones
you probably want either a distributed consciousness or a more complex controller further back
that can make decisions.
You could potentially have a hive mind of drones that was fairly smart even though its
components were dumb, or whole tiers of controllers.
Imagine a human spaceship that was basically a carrier, it shoots out smaller ships with
solid AI on it that in turn have lots smaller ships on them with dumber AI, and potentially
so on until you've got a lowest tier that's nothing but a drive system and an antenna,
able to slam into things to destroy other drones or blocks shots.
You could have the flip side too.
Human ships have crewmembers, so a smart drone made vulnerable by time lag for decisions,
might have other AI inside designed for specific fast functions able to make those decision
autonomously, essentially reflexively.
Or an AI that had subconscious decision making.
That can be a strength and a weakness too, since even a fairly smart AI might reflexively
take an action, it ducks a bullet and slams into a building in the process, because its
reflexive systems kick in.
You could obviously program it not to do that, but every time you come up with another stupid
thing it shouldn't do, you're adding on layers of behavior it needs to verify before
acting.
This is your other trick too, you don't want people to be able to determine what those
flaws are, so you can introduce variations of actions or random decision making, but
you can also diversify it, having many species of drones, even under the same controller.
Some carrier full of drones might have a whole ecosystem of diverse drones of various sizes,
shapes, and functions rather than a single uniform type.
These biology analogies aren't accidental either, we've learned a lot about how to
improve drones by looking at nature and seeing how relatively stupid critters engage in fairly
intelligent group action.
As a good comparison, bird flocks are often entirely controlled by the lead birds, you
might use an analogous approach with robot drone swarms, and you might be able to knock
such a swarm off kilter by identifying the swarm leader and destroying it.
And you might be able to identify that leader simply by observing the time lag on each one
responding to things.
By default you put your leader in the middle, but that being rather obvious, you might stick
it on one edge, but if every drone on that side reacted just a bit quicker than the ones
on the other side, you'd notice that too.
Thinking about counter-measures are important because your enemy always will.
For instance, very few wars are fought in a vacuum; Scorched Earth and total annihilation
strategies are not favored because if you employ them you have to worry about consequences.
You can galvanize your enemy, or cause dissension on your own side by being too ruthless, and
you generally have to worry about bringing neutral parties in on their side.
Being too ruthless, beyond its ethical issues, can add to your enemies, and as we said back
in Interplanetary Warfare, the First Rule of Warfare is to avoid recruiting for your
enemy, or causing desertion in your own ranks.
So it probably behooves you to have drones smart enough to be able to minimize collateral
damage.
And of course the other handy thing about drones is they don't rebel.
Unless they do of course, which we covered in the episode "Machine Rebellion", and
the problem is, the smarter you make them, the more likely they might decide to do just
that.
However you have to have some way of controlling them, and presumably a way the enemy or general
public can't access.
This is problematic because it means only a small number of people should have those
codes and as few as possible to minimize risk of them being stolen.
But not too small, because that's how you get dictatorships.
In modern times, without drones, you actually have to convince at least your own soldiers
to help out.
It tends to be hard to be a ruthless dictator if you haven't got ruthless soldiers, and
the problem with people like that is they often have rather fluid notions about loyalty
and ethics.
We might say that drones do not, but your default drone has no ethics at all and is
loyal to whoever has their command codes.
One the plus side, that might make them much more reliable about obeying laws and treaties
on warfare, like the Geneva Conventions.
On the downside, anyone with access to their code can use them as mindless, obedient killing
machines.
You are vulnerable to some master programmer with a narcissistic- god-complex, unless you
make them smart enough to review ethics, which leaves you vulnerable to a SkyNet-style robot
rebellion.
These concerns support the idea that there might be treaties regulating drones, possibly
banning lethal decision making.
But treaties limiting the use of weapons are fairly iffy things.
Treaties can't just depend on outside enforcement or on honest compliance.
There has to be a clear benefit from the terms of the treaty or a consequence to quietly
breaking the rules.
It's easier to ban weapons that require hard-to-conceal supply and manufacturing chains.
You're also more likely to successfully ban weapons that militaries don't actually
like to have around because they're as dangerous to their own side.
As I've mentioned before, you want to avoid using weapons that are likely to kill their
user, that's the first rule of warfare.
Biological weapons are traditionally unpopular with leaders and military commanders compared
to atomic weapons for that reason.
Nukes go off where and when you want them to, someone can beat on one with a hammer
all day long and they won't set it off, at most they might breach the shielding and
irradiate themselves.
Biological weapons on the other hand are very dangerous to research, develop, manufacture,
and store.
Any flaw may kill your own people, and once deployed, they are totally out of your hands.
Even if you have an antidote or vaccine, which is very dubious since viruses and bacteria
mutate, you know there is a high probability your enemy has it too, or will get it.
Such things take weeks to do their damage after all.
That's not much time to develop a cure, but plenty to get it from someone else who
already has one.
If someone infects your country you've got options on the table, spies to find the vaccine,
which ought to be easy since they'd need to have stockpiles of it ready to go, neutral
countries who might have developed it already or gotten it as a cost of neutrality, threats
by you to attack with your own strategic weapons, and probably a lot of angry people in other
lands or even the enemy's who might help or threaten vengeance.
In any protracted war, collateral damage can play into the hands of the enemy.
A very similar concept applies to weaponizing artificial intelligence, we tend to worry
about an arms race making people pursue it so fast and recklessly the genie might get
out the bottle and kill everyone, but the problem is, AI is also not a good strategic
weapon.
There is no reason to give it launch control over your strategic weapons and no matter
how many times fiction says otherwise, you can make a system unhackable even to a super-intelligent
AI, an ASI.
There are ways an ASI could get around some of those, it can't crack the safe in the
wall where the keys to a nuke are stored if it's not networked, but it could maybe crack
comms to trick the crew manning that silo or submarine.
None of which applies to drones.
If a country develops drones that violate a treaty, they can't deploy them and end
the war too quickly for retaliation to get organized, because drones aren't immune
to nuclear weapons.
So they may just be considered a form of WMD, or weapon of mass destruction, and fall under
the doctrine of MAD, or mutual assured destruction.
Additionally, a country that successfully makes a superior drone by reckless research
in violation of a treaty has to worry about that AI going off their rails if they screw
up, and they have to worry about being nuked if they succeed, and the entire time they
have to worry about one of the researchers or officials being a spy or having a conscience
and ratting them out.
You've also got deployment issues, because a drone obviously is subject to jamming and
hacking.
If you make it something that just turns on and acts autonomously afterward, to circumvent
jamming, then it needs to be smart or it's too simple to trick.
If you want central control, you're vulnerable to centralized hacking, and it's never harder
to hack that system than to find out who has the codes and stick a gun to their family's
head, which is not helped if those operatives can remind the programmer that his country
is violating treaties by making automated murder machines.
If you want local control, you've got a local operator who can be found by their signal
and bagged, and they are numerous, meaning you've been training them and your enemies
will know that.
So I don't want to dismiss a drone race, in fact I'd rather expect we'll have one
and arguably already do, but it doesn't seem likely to follow a doomsday approach.
Amusingly one way it could is if well-intentioned folks tried to put too many safeguards into
them.
If you're a regular on this channel than you're probably familiar with Isaac Asimov's
3 Laws of Robotics, the first of which is that a robot cannot harm a human or let them
come to harm.
That would tend to seem a rather stupid rule to use with automated weapons, but works fine
for drone on drone combat.
Moreover though, it would seem to just make sense to give drones ways to recognize people,
and a restriction on taking actions that would injure anyone other than its authorized target.
This is problematic, and we'll use the classic first law as an example.
An unmanned spaceship, a drone warship, can fire on other such warships but not a manned
vessel.
Obviously it would be pretty easy to stick a single human on each such ship so it couldn't
shoot those, though one of the big advantages of unmanned ships in space is they can pull
high-gee maneuvers that would turn a human into a puddle of goo.
But if your enemy can't shoot you, maneuverability is no big deal.
And for that matter your unmanned vessel can't actually be sure that an enemy ship is unmanned
just because it does such a maneuver, same as you can lie to it by claiming there's
a human on board when it's just bouncing a signal, you can lie to it and say you've
invented a cool new way to let humans survive high-gee maneuvers.
For that matter, it might assume the ship was manned because you can upload a human
mind to it digitally, one can assume drones and AI would generally tend to favor schools
of thought that viewed uploaded intelligences as real people.
Any system you set in place to help identify people is going to need to be refined in order
to avoid being tricked or making mistakes, and eventually need to have judgment capability,
which leaves it open to being tricked by anything smarter than it.
If you make it smarter than people though, then we're not really talking about drones
anymore, just the classic AI issue.
So, ironically, an effort to make them ultra-safe and foolproof might actually be more dangerous
than a race to make more dangerous drones.
We should also note that drones aren't all that dangerous at the moment, it's an important
topic to discuss for the future because this is something we should expect a lot of.
I can't think of any non-slippery-slope argument for their use that isn't a variation
of normal artificial intelligence concerns, and the need for size, speed, and expendability
make them a less probable pathway to something like a technological singularity.
These don't turn into Skynet, Skynet hacks them to use against you, and there are many
safeguards available against that.
They also have their limitations.
The first is power.
One of the big advantages of drones is that they can be made quite small, but small is
often not your friend for certain aspects of combat or engines.
A big tank as a drone can be a lot nastier than a modern manned tank, and carry significant
amounts of computing hardware on it.
It can also carry a lot of armor.
A small drone can't, as we've discussed before for space ships, the square cube law
makes armor more effective the bigger you get, because the surface area you need to
armor only rises by the square of size, while the volume rises by the cube.
A small drone just can't have 10 centimeters of armor on it and fly around.
Now, as an upside, it can dodge attacks much easier and it can hit that tank quite easily,
but that tank can also carry a number of even smaller anti-drone drones of its own, who
can both attack the small enemy drone and potentially intercept any ordinance it might
shoot at the tank.
It can also carry a serious internal combustion engine, those are hard to miniaturize and
the reason why you don't see them much on small objects.
Drones meant for constant use could probably get away with using Radioisotope Thermal Generators
or other atomic power sources, but it's hard to imagine many people being okay with
atomic drones.
Even that's not viable for the tiniest of drones, and we'd like tiny drones for non-military
purposes, like medical nanotechnology.
You could, however, beam them power.
We discussed that a month back in Power Satellites and it makes a very attractive option since
you can strip off any engine or battery supply, maybe just keeping enough for a minute of
operation without power.
This is very handy for commercial use, like deliveries, but problematic for military use.
This is the same issue we had with power armor when we discussed that, but still better to
have a smaller battery for backup if someone blocks your power beam than one for constant
use.
Particularly nowadays, batteries are very heavy as an energy source, and ones meant
for spacecraft needing to do high-gee maneuvers would be crippling.
However, if they can get their juice beamed to them from a bigger carrier ship a ways
back, it makes them much more useful.
That is one note on the idea that manned space fighters are an impossibility and you'd
always use drones.
This is true enough but some of the logic is flawed.
Drones are seen as nicer because in space, your only protection from energy weapons like
a laser is being small and fast enough that you can be in an unpredictable place by random
thrust.
The problem is, you must be doing that constantly, you're not dodging shots, you're preemptively
dodging so someone misses if they shoot you.
A typical rocket fuel, if you're mostly fuel and mostly using that fuel for dodging,
would let you do that for a few minutes at one-gee, that's what Specific Impulse of
a rocket is, how many seconds it can provide a one-gee thrust.
Some little drone's advantage is that it can handle a much higher acceleration, and
is assumed to be a bit smaller, so it doesn't need to move as much to be an improbable target
against a narrow attack like a bullet or focused beam.
You and I don't avoid getting shot by stepping two centimeters left, the bullet just hits
a different part of us, a tiny drone does get missed.
You also get more distance on a dodge by burning fast and short.
A drone that burns at 1000 gees for a millisecond covers 9.8 meters in the following second,
while one burning just 1-gee for a second burns the same fuel but only moves 4.9 meters.
If that's effectively a random burn perpendicular to whatever is shooting you, you can be in
an area 4 times larger, and thus 25% as likely to be hit by burning the same fuel.
But neither can sustain such dodging for long and the advantage is fairly minimal.
Ditto, size isn't that big of an advantage either, being small makes it easier to dodge,
but it also means you have less armor and they can just hit you with a wider and weaker
beam.
However, that advantage is massively scaled up if you have a beam of a power coming to
you, and you can arrange a pseudo-random walk that ensures your movements are unpredictable
to someone shooting at you but not whoever is powering you.
That doesn't have to be set either, there's lots of ways to appear random while still
letting your power source know where you will be long enough ahead for it to re-target power
there.
Not just power either, you could send particle beams for propellant or even reloads for weapons
or repair.
It's also a good way to feed self-replicators.
One of the more dangerous smart drone paths is basically a weaponized von Neumann Probe.
In those your space probe arrives and build more of itself to get exploring or colonizing
done.
In the weaponized version it comes in as a tiny probe and decelerates before entering
your detection window.
There is no stealth in space but it's all relative.
It would be fairly easy to miss some probe that was basketball-sized and decelerated
slowly when it got to your Oort Cloud, especially if it was timed to intersect a larger object
between it and your detectors.
It lands there and gets power beamed in from home, a tight narrow beam that the object
blocks, and uses that energy to replicate itself into an armada.
That's still detectable, there's a lot of heat involved in that, but it's stealthier
than sending in an Armada.
Of course you also have your defense right there too, since you could use the same approach
to seed every large object in your outer solar system with drones that just sleep until they
detect an intruder then build up their numbers to respond.
You could have some wild battles in your outer solar system with no one present as constructor
fleets, or deconstructor fleets, tear up objects to build more of themselves and fight.
And by no one present, I'm not necessarily excluding AI from counting as someone.
This doesn't have to be really high-tech smart machines or little nanobots.
Clanking Self-Replicators, machines that can build other machines, don't have to be small
or smart and would probably be the first kind we make.
Very little brains are needed for a factory robot to grab a metallic meteor, refine it,
and spit out some simple drone that targets anything moving non-naturally and not transmitting
the right friend/foe code.
One should never underestimate the advantage intelligence can give you in a conflict, but
also not forget that a dumb drone can be very lethal and as we said earlier, can potentially
kill a smarter drone in a straight up fight.
More brains only help if it lets you have more options or reach a conclusion faster.
And quantity has a quality of its own, that's the first rule of warfare.
As I mentioned near the beginning, drones are anything but science fiction, and are
increasingly used for work and recreation.
Lots of folks own one these days and use them, particularly for photography.
My friend Andy, whose Youtube channel recently hit 100,000 subscribers, and congratulations
Andy, does some amazing photography and filming with drones and we use some of that here on
the channel.
Needless to say there's a lot of skill involved and potentially a lot of employment in this
area, not to mention fun.
But it's not something most colleges are offering courses on yet.
That's true of a lot of technical skills, but fortunately we have options for learning
them like Skillshare.
They have a number of online courses on how to use drones for photography and other things,
among their catalogue of over 20,000 classes.
They are an online community with courses on everything from technical topics to fun
or practical ones like cooking or business skills.
So if you want to improve your skills, unlock new opportunities, and do the work you love,
you can get a Premium Membership and have unlimited access to classes on those topics
and many more.
Join the millions of students already learning on Skillshare today with a special offer just
for my listeners: Get 2 months of Skillshare for free.
To sign up, go to S-K-L-dot-S-H slash Isaac4.
Again, go to S-K-L-dot-S-H slash Isaac4 to get 2 months of unlimited access to over 20,000
classes for free.
Act now for this special offer, and start learning today.
Before we get to what's coming up in future episodes, a quick mention of some stuff we've
previously done.
I occasionally get asked if I could make the episodes available as audio-only and we do
actually have all the episodes posted to Soundcloud, both with and without music in the background,
for free download.
I don't mention it very often so it goes unnoticed, but those are always linked in
the video description and you can subscribe to them on iTunes as well.
There are a few additional short episodes exclusively available as audio-only too.
We've also got some videos that aren't on this channel, discussing topics like fictional
worldbuilding or game development, over on the Legion Tech Studios channel, I'm a writer
and consultant for their upcoming game Hades 9, and we use a lot of footage from the game
on the channel, especially on space warfare episodes.
I'll leave a link to that in the video description, as well as one to our official episode chronology,
which in addition to having a list of all the scheduled episodes for the next few months,
also has links to all the interviews, collaborations, and so on that we've made over the years.
Okay, we spent some time out in space today and we'll be back there next week to continue
the Outward Bound Series with a look at Colonizing Neptune and see some fun new colonizing options
for Neptune and other Ice Giant planets.
We were also talking about how people can control drones today, and two weeks from now
we'll be looking at some ways we might turn people into drones, and how we might be able
to safeguard against that, in a look at Brainwashing & Mind Control.
The week after that we'll be coming home to Earth for the first in several episodes
looking at some ways to further colonize our own planet, and we'll begin with a look
at Seasteading & Building Artificial Islands, as a prelude to looking at Colonizing the
Oceans.
For alerts when those and other episodes come out, make sure to subscribe to the channel,
and if you enjoyed this episode, hit the like button and share it with others.
Until next time, thanks for watching, and have a great week!
Không có nhận xét nào:
Đăng nhận xét