Ex-Google CEO urges US to ignore calls to ban autonomous weapons

And as far as I know, autonomous weapons don't inherently involve scorched earth scenarios. Perhaps in the far future the worst war we have will be each side attacking the other's autonomous weapons and then calling a ceasefire because humans are too costly to be in the frontlines.

Think of it through the lens of an iterated prisoner's dilemma: if I suddenly develop deathbots and you don't have them, it changes the payoffs for defection/cooperation which could cause me to change strategies and kill you.

I would recommend Axelrod's book "The Evolution of Cooperation" for ideas surrounding this.

With the above information, what would you do, if you were a state actor of an advanced country? Surely, (1) you either want to develop these yourself, or (2) ensure that nobody else develops them. (2) can be achieved by international treaties, but the question is whether it can be enforced. If state actors don't trust option (2), then the only option left is (1).

Because the consequence being the end of the world rather than just a large part of it is what keeps the peace.

At the very least weapons are controllable; they have a kill switch. Things like SARS-Cov-2 do not have a kill switch.

I'm against taking the human out of the loop and leaving the decision and responsibility to kill to a machine. I can think of many situations where a drone is a necessary technology to win a war but of far fewer where a human confirmation cannot be afforded.

Drones are assumed to be remote controlled, therefore retain a human decision maker at all times. Robots would make such decisions themselves.

While drones come with their own set of problems (usual suspects are transparency, deployment regulations, sovereignty, decision making without skin in the game, etc.) they are distinct from the set of problems robots / "autonomous" weapons introduce. (Like distinction, proportionality, transparency & responsibility)

>too much investment in immature tech

Or research into paradigms shifting tech negates moat of old advantages. Russians good at antiair because US would not want to invent tech that would erode US air supremacy. Kind of how UK lost naval gap due to dreadnaught reset. I can imagine AI air planes undermining US air / naval superiority balance. Lots of obsolete hardware and training if a fleet of drones become smart enough to penetrate a carrier group. Perhaps it may even be sensible not to pursue / start arms race if it means undermining decades of expensive acquisitions and expertise. Or at least have enough security to ensure adversaries don't acquire kryptonite of your own making.