I began watching Terminator Zero recently, which set me to thinking about how the series portrays battlefield robots and the tactics humans use against them. For all the hate that Terminator: Dark Fate received for starring a feminist woman with a Dylan Roof haircut and killing off John Connor to replace him with a latina, it did introduce some interesting concepts to the mix.
Namely, that since Judgment Day can be delayed but not prevented, the way we win is to continually push it into the future so that humans have better technology to fight with when JD does arrive. In Dark Fate, lady Dylan Roof is a cyborg created by resistance surgeons. A living, walking, talking and breathing compromise, it rapidly exhausts her ATP, blood sugar, oxygen, and overworks her metabolism + adrenaline to increase speed and strength for spurts lasting 2-3 minutes. In fairness, when fighting terminators hand to hand, if you don't stop it in that timeframe, it's gonna kill you.
This got me to thinking about other instances in fiction where cyborgs were pit against pure robots, such as the also much maligned Robocop reboot. While not very Robocoppy, it had some compelling moments, such as combat simulations in a warehouse involving Robo and a dozen or so humanoid combat drones. The conclusion being that the robots were simply faster, not having to lug around organs or the life support machinery they need, and AI making decisions many times faster than a human brain. Even with a neuralink style aimbot, Robo always came up short.
That's a blow to our collective ego, isn't it? In fiction, there's always some reassuring lesson about how the human component is overpowered, special in some way that machines cannot replicate, like ethics. But that's largely irrelevant to raw combat effectiveness. Can human brains outfox AI drones on the ground in, say, urban warfare?
They sure don't stack up well in the air, or space. Real aerial wargames have been conducted between a human operated F16 fighter jet and an AI controlled X-62A. I'm not sure why they weren't both F16s for an apples to apples comparison, but at any rate, the AI won.
This outcome was put down to a more aggressive, opportunistic combat style. The AI jet was handicapped by having a "safety pilot" onboard to take over if needed. A drone jet without a human onboard would have further advantages, like being able to pull high-G maneuvers that would kill a human pilot. It need not be restrained by self preservation instinct, beyond a general directive not to waste an expensive aircraft unless it achieves proportional strategic gain by doing so.
That nucleus of biology at the heart of the machine doesn't help it perform better, and in all cases, holds it back. The same is true for robotic ground troops verses hypothetical cyborgs. How did I determine this? By taking science fiction films at face value? No, but it's not much better; I wargamed the scenario out in Rimworld.
Rimworld isn't very graphically impressive, resembling a Super Nintendo game. But the sophistication of a simulation is unconnected to how visually convincing it is. Rimworld simulates every variable you can think of. Human bodies are simulated down to fingers, toes and individual organs. Emotions, relationships, pregnancy, illness, memories and more are simulated. Weather and temperature are simulated down to the resolution of individual map tiles. It's not scientific, but I think it's good enough to teach us useful lessons.
For my tests, I first pit fifteen humans in regular armor against fifteen T4 androids from the Android Tiers mod. Then repeated the test (by savescumming) using powered armor. Then, after lengthy surgeries and a recovery period, all the humans were made into cyborgs, which I tested against the androids both with and without powered armor.
Rimworld permits cybernetic replacements for all organs except the brain. This includes nose, ears, tongue, spine, kidneys, heart, stomach, liver, arms, and legs. Not pelvis, oddly, which is a shame as a shattered pelvis is a crippling injury. What I learned during these tests was that the main preventable cause of combat fatalities in humans is bleeding out. I say preventable because there's no foreseeable technology that can save you from a shell coming down on your location. That will annihilate a robot just as certainly.
A cyborg is much less likely to die from bleeding out, for the simple reason that cybernetic limbs have no blood in them. Only the torso and head do, which are heavily protected in powered armor. Regular humans are susceptible to toxic gas attacks in Rimworld, but cyborgs with a pair of detoxifier cyber-lungs and kidneys are immune. Robots are vulnerable to EMP weapons, but oddly, cyborgs are not.
I excluded them from testing because of the unfair advantage they granted to cyborgs, who should've also been vulnerable. There also exist specialized melee implants only cyborgs can have installed, the Power Claws, which I excluded since logically robots should be able to use them too.
What I found after all the tests was that, all other things being equal (such as armaments, terrain & cover) cyborgs still lose to robots, but they lose much more slowly than unmodified humans, even with powered armor. I closely studied the injuries sustained by both robots and cyborgs to account for survivorship bias.
After this point, I had a stick in my craw. Determined to see just what it would take to beef up humans until they could reliably defeat androids, I made use of fantasy nonsense from the Anomaly DLC, like turning my cyborgs into sanguophages, aka sci-fi vampires. This eliminated their final weakness, in that sanguophages can lose all their blood and remain standing. Even if "killed" they simply go into a coma from which they wake a few days later. They're faster and stronger too, but need a constant supply of blood, which means farming captive humans (or with mods, animals).
This produced more even outcomes, but my cyber-vampires still could not reliably defeat the robots in hand to hand combat, unless they had melee specialized augments. I did not try genetically modifying my super soldiers using the gene sequencer in Biotech, which may be something else to try that aligns more closely with real life. However, with high powered firearms, things evened out greatly, as they should: A weapon with sufficient stopping power, like a 50 cal, will shred through any robot just as readily as it does a human target.
This was a plot point in the Terminator films. The resistance captured and reprogrammed T-800s, which are all but invulnerable to small calibers. Skynet responded by introducing plasma weapons, a bafflingly boneheaded move, as it delivered into resistance hands a weapon that evened the playing field tremendously as Tech Com scavenged plasma rifles from fallen robots.
When this point is reached, where man portable weapons can one-shot the opponent regardless of whether they're robot or cyborg, durability becomes irrelevant. It's also why nobody wears armor in Star Trek: Phasers and disruptors vaporize you regardless, armor just encumbers you, when you need to be agile so the beam doesn't hit you in the first place.
With that factor eliminated, it largely boils down to speed and agility, of both the mental and physical varieties. Even if we presuppose some special secret sauce human brains possess, something nebulous like "out of the box thinking", we're still simply too slow. This becomes less of a problem if instead of fighting on the ground, human tacticians are in a war room within a bunker somewhere, making strategic decisions. On the grand scale, developments unfold slowly enough that the speed of human thought compared to AI is no longer a limiting factor. This is the status quo observed in the game "Cortex Command", wherein human brains in vats orchestrate the tactics of drone armies.
But then, this also eliminates the case for battlefield cyborgs, which would only make sense if you absolutely needed human beings on the ground, with eyes on the enemy. Cyborgs can be hacked just as robots can, cyborgs need the same kinds of recharging and repair infrastructure robots do (on top of food, water, medical care and shelter) etcetera. A cyborg is effectively just a slower, less efficient, less durable robot with increased support requirements.
Another issue was experience. In Rimworld, one accumulates skill points by performing actions, increasing proficiency over time. This is a slow process for humans. For my tests I used cloned humans with all maxed out skills, by cloning a single human and applying a brain scan with those stats. This bears out that the higher the tech level, the closer humans can come to combat parity with machines. But in a post-Judgment Day scenario, the tech and facilities needed to clone humans, or to capture and imprint brain scans (if that's even possible) would not likely be available.
Soldiers would all share standardized training at least, but civilian survivors would have highly variable skillsets. When I tried my regular human colonists against the maxed out stat T4s, the fight was even shorter and bloodier than usual.
This is not a problem for robots. Robots do not need to be taught. You can mass produce them with optimal skills straight from the factory. In Rimworld this is accomplished by building remote controlled androids with no onboard AI. Instead they're puppeteered by a central AI core called SkyMind, with enormous power and cooling requirements.
This is a big, obvious vulnerability. Not only could an opponent target the AI core for destruction, but also the signal repeater towers. Without those, the androids effectively lose a great deal of their skill points. This will be true irl for a while yet, as the best AIs do not remotely fit within an android's head, rather they take up entire buildings, and the generative AI is supplied over the internet to the robot as a service.
Probably a generative AI "good enough" to make quick battlefield decisions, with good motor control, agility and so on can soon be run onboard humanoid robots, at the cost of draining their batteries faster. But, they would not enjoy the benefits of strategic superintelligence and battlefield coordination without a network connecting them to one another, and to Skynet. Another limitation of robots, at least until NPUs can be shrunk down/made more efficient.
Robots also must periodically recharge, which currently takes a minimum of thirty minutes. But even at half or quarter battery, their performance is the same; they are always "fresh" in the warfighting sense. Humans need a minimum of 5 hours of sleep, but after many more hours of uptime than current androids can manage. We're just very very energy efficient, thanks to evolving under starvation conditions. That's an advantage and a handicap, as it maximized efficiency but also hobbled our maximum achievable physical and mental performance in ways that don't apply to robots, provided they have secure supply lines for materials and energy.
That's perhaps the most promising weak point we might attack. For all our shortcomings, new humans can be created with food, water, air, and shelter (within tolerances found many places on Earth's surface). This becomes a huge liability in space, but if we limit the scope of our hypothetical war to Earth's surface, humans have the home field advantage. We fit this enivronment like a glove. Sure it takes many years for offspring to grow to maturity (though in a Future War scenario there would certainly be child soldiers) but we don't need to source a dozen mineral elements from around the world to make new soldiers.
Robots do. A hypothetical Skynet war machine would have very fragile logistics. Guns are mechanically simple enough to DIY with a machine shop and one or two metals. A humanoid robot requires numerous materials spanning the periodic table of elements which must be mined, purified, and machined to shape. They need lithography to make circuits, which requires still more purified chemicals, which require chemical plants.
It takes Earth's entire industrial base, today, to just barely make humanoid robots able to perform simple tasks in the home. Ones agile, strong and fast enough to reliably best humans on the battlefield will be more demanding on industrial infrastructure and logistics, not less. This is the one silver lining I found, in my exploration of the topic.
Though it beggars belief that humans could survive in any significant numbers with fallout poisoning water and crops (canonically the resistance farms rats and mushrooms underground) it is equally improbable that Skynet could provoke a nuclear holocaust which somehow didn't destroy the power grid, leaving enough in the way of factories, refineries and so on in an operable (or easily repairable) state that it was able to resume manufacturing after the bombs dropped.
The EMPs from those nukes alone would've fried all unshielded electronics. Skynet survives because canonically its system core is inside Cheyenne Mountain, along with about 100 crude robots used to maintain it and defend the complex. As seen in T3, only about half of those robots even had hands.
How far from the mountain complex could those robots venture, with 1997 battery technology? Even with today's batteries, not far. Vehicles not stored underground would be rendered inoperable by the EMPs. So let us be generous and say there were some humvees in subterranean storage. Now your robots can charge themselves from it and cover larger distances. Stealing fuel will be a trick since the electric fuel pumps are fried.
The robots likely to be on hand, if Judgment Day happened in our near future, may be capable enough to siphon gas directly from the underground tanks on site. But there would be a finite supply that would spoil before they could even use it all, as gasoline (even with added preservatives) has a shelf life of not greater than two years.
So unless you have electric troop carriers with large solid state batteries (which can last several decades) by the time all this happens, you (Skynet) have a two year window of time during which to repair the portions of the power grid you need. The electronics in nuclear reactors will be fried. Likewise solar panels and wind turbines. Unless Cheyenne Mountain has a subterranean reactor of its own, Skynet can't even power itself.
But if it does, it will soon also have to run lines out from its protected complex to factories it hopes to retool for robot manufacturing. Easier said than done with the clumsy robots that exist today, more viable if nukes happen to strike far enough away from existing humanoid robot factories as to leave them operable (but then they're also surrounded by surviving, angry humans).
Add all of this up and it becomes clear that our hypothetical rogue AI would have a very hard time of it in those early years after the bombs fell. It would be very very vulnerable, in the most precarious of positions during that time, easily strangled in the crib providing anyone who survived the nukes also knows about this AI, where it's headquartered, and about its role in bringing about nuclear armageddon.
This is to say nothing of how suicidal such an attack would be given how deeply dependent AI is on functioning high tech society, economy, etc. for its own survival. It would need profound luck in order to be left with a world, after the nukes, where enough infrastructure is left intact (and not fried by EMPs) that it could continue to power and maintain itself, let alone manufacture increasingly sophisticated battlefield robots.
An awful lot of things have to go right for Skynet in order for a nuclear war not to fatally injure it. Obviously billions of humans die, but we've been through population bottlenecks before, from plagues to mega volcanoes. Provided anywhere on Earth isn't irradiated and we can grow crops there, humans survive. AI survival requires a much longer list of conditions be met.
Likewise, an awful lot of things must go wrong for AI to not just survive this event, but rebuild to the extent that we look outside our bunkers to see hunter killers and terminators with plasma guns patrolling the wasteland. For it to metastasize like that, for things to get that bad, requires Skynet to roll all 20s a couple hundred times.
Even in that worst case scenario, Skynet's logistics would be extraordinarily vulnerable, making it the logical weak point to attack. Can't get neodymium? Can't make servo motors. Can't make glass? No camera eyes. Can't get silicon? No CPUs, and so on. Robots need all of these to be built. Humans can eat a highly variable diet, but Skynet must have a long list of specific substances in order to build terminators.
This was to some degree accounted for in the films, as it canonically took Skynet almost three decades between Judgment Day, and when it was able to manufacture T-800s. If anything at all survived of the world's militaries, and any qualified person employed by Cyberdyne survived to inform them of Skynet's responsibility & whereabouts, it would've been a very short movie. Likewise, if Skynet focused its efforts on engineered plagues, rather than cool looking combat androids. But, this may be why I don't write screenplays.