https://youtu.be/HrhuivuZb74
Killer robots are in the news, with tech celebrities including Elon Musk recently calling for a ban. So are robots in war good or bad? First things first, an actual ban probably isn’t happening. Because it’s kind of foolish to ban weapons when your enemies won’t obey. You’re handicapping yourself. Countries that have to follow weapons bans like the US, Europe, and Japan, probably won’t push for bans that will be ignored by China, Iran, or North Korea. 
Okay, a ban’s unlikely. But it’s important to ask whether killer robots are a good thing or a bad thing.
So the main effect of killer robots is swapping blood for money. That is, you take your humans out of combat and replace them with something that, ultimately, just costs money.
This has a couple counterintuitive effects. Getting into it, then, we have two fundamental scenarios: first, what if only one side has robots, and then what if both sides have robots.
Starting with the one-sided scenario, in short we get cheap war. Politically cheap. These days the main deterrent for a country like the US is bodybags. It’s much easier, politically, to sell a war to Americans if you’re not suggesting “boots on the ground” — an actual invasion with soldiers.
You can impose sanctions, you can fund guerillas, you can even launch air wars far easier. Ever since Vietnam, the US government is very aware that the American public loses its appetite once 18 year olds start coming home in bodybags.
We’ve seen this again and again, where the US public is eager for even devastating air wars, from Clinton in Iraq or Serbia to Obama in Libya or Syra. Yet ground wars can be politically disasterous, such as George Bush’s misadventures in Iraq and Afghanistan.
So war becomes politically cheaper if you’re doing it with robots. Now, what makes this interesting is that it affects countries differently. Namely, blood is more expensive for rich countries and, especially, for rich democracies. North Korea or even China, for example, can probably take casualties with less political fallout than the US, Europe, or Japan. War gets cheap for any country using robots, but that effect is most concentrated among rich and democratic countries. Making them potentially more aggressive.
Of course, playing out the game theory, if rich democracies are more willing to go to war, it’s possible that other countries avoid antagonizing them. Iran’s “death to America,” North Korea’s hellfire fantasies, these may be reined in if those countries fear an America whose population enjoys robot war. The pariahs of the world might shift from “rivers of fire” to a kind of passive-aggressive attitude like France in NATO.
Now, what about the other guy? What happens to countries who don’t have robot armies? Well, we can assume that robots are more cost-effective — they kill better for the dollar — which is why they’re used in the first place. So we can assume robot armies at a given budget are more powerful, hence harder for the other guy to fight. Even so, we don’t know if more enemy soldiers die since short wars may kill fewer enemy soldiers than long wars.
What about enemy civilians? On a tactical level, robots should eventually be more precise, reducing civilian casualties. On the other hand, robots have two problems: they aren’t moral and they can make mistakes, especially in the beginning when robots aren’t very sophisticated. Taking the moral issue first, a robot set to shoot any trespasser, for example, can murder kindergarteners in a way a human soldier cannot. Indeed, a human soldier ordered to do this may disobey, sabotage or defect, or become psychologically useless as a war tool — essentially, the humanity of common soldiers works against gross human rights violations in war, and robots take that away. Of course, you could also see the opposite, where the robot programmers could actually be more compassionate than regular soldiers.
As for robot sophistication, every new technology has kinks to work out. In general rich and democratic countries pay a political price for harming civilians, and even bad guys don’t usually waste RPG’s on goat herders. These create an incentive to improve robots and, after all, robots can be improved faster and cheaper than humans. So, granting that robots are immoral, the accuracy incentives alone push towards fewer civilian casualties for a given war.
What about when both sides have robots? This essentially doubles the fun: it converts war into a kind of auction, where both sides place their bets using money, not blood. Now both sides get cheaper war — even North Korea can replace expensive soldiers with cheap drones — but the power imbalance is the same as one-sided war. That is, war remains disproportionately cheaper for rich democracies who don’t have to pay in blood. And, speaking of blood, a two-sided robot war obviously decreases military casualties on both sides, and should again reduce civilian casualties to the extent both sides’ robots become more accurate than human soldiers.
Now, whether robots are only available to rich countries or being used by both sides, either way you get that bodybag effect, with rich democracies facing lower political cost of war. To the extent this increases their appetite for war, potentially to the point that poor dictatorships are almost enslaved by rich democracies.
Even if rich democracies are run by better people than poor dictatorships, rich countries are also self-interested. They may strip poor countries for their profit or ideological whim: the West, for example, could force Venezuela to democratize, or Saudi Arabia to hand over oil assets or grant women and homosexuals equal rights. Again, how you feel about this probably depends on how you feel about the US or Europe dictating how other countries are run.
A final point: shifting military R&D to robots will probably massively speed-up progress in robots for non-military use. Given non-military robots are far more economically useful than non-military tanks or non-military F-16’s, we get a very nice quality-of-life bonus from robot R&D alone.
Bottom line: killer robots replace blood with money. This probably means fewer deaths on all sides, including civilians, unless it makes rich democracies hyper-aggressive. Meanwhile, killer robots dramatically change the balance of power so rich countries get to boss around poor countries. To the extent rich democracies are run by better people, this may not be a bad thing at all. And, finally, we get an economic dividend if massive military R&D is steered into robot technologies that can actually makes our lives better.