Air and sea combat tend to place a big premium on being faster. If you can cruise strategic distances or sprint tactical distances significantly faster than your opponent, all sorts of choices open up for you. (Notably: you choose whether to engage at all, and you get all sorts of opportunities to concentrate force.) And for physical reasons that aren't going away just because of cheap chips and bandwidth, it's difficult for little vehicles to keep up with a conventional-sized aircraft, ship, or sub. Then unless the smaller vehicles are lots cheaper (in some appropriately scaled way, perhaps "cheaper per ton") to build and operate than the big ones, it's not obvious the smaller vehicles naturally have a direct advantage in slugging things out.

However, little drones (and other small-scale autonomous devices with fast computers and high-bandwidth connections) could still have a large effect on high-intensity air or naval combat even without scenarios of a modest number of supersmart drones duking it out with a conventionally-sized aircraft or ship or sub. One scenario that seems likely to me is many supersmart drones and balloons and buoys with staring passive sensors cooperating with a modest number of devices that resemble modern radar/sonar platforms and with a small number of devices that resemble 1980-era missiles or torpedoes. (In this scenario, the small vehicles do have a clear cheapness advantage in a suitably-scaled way: roughly "[much!] cheaper per kilometer of effective sensor aperture size".) Another scenario is that a war of attrition involving progressively smaller, slower, and stealthier devices might make it a new kind of PITA to secure the immobile facilities (e.g., roads and ports and airfields) that current big-vehicle systems rely on.

Of course, it's somewhat moot trying to analyze high-intensity war as a problem of engineering on the battlefield as long as we don't understand the strategic dynamics of high-intensity war in the presence of incredibly cost-effective indiscriminate weapons like hydrogen bombs. It seems to be at least as much an intractable game theory problem as an engineering problem. (And as David Friedman said, "when I am picking problems to work on, ones that stumped John von Neumann go at the bottom of the stack.")

Drones are much cheaper in a more important respect - human lives. When wars can be fought with machines, leaders don't need to pay as much attention to democratic will when choosing who to attack.

We're well on our way from turning war from a matter of human costs into a matter of monetary cost.

Whether this is a good or bad thing only time will tell.

> Drones are much cheaper in a more important respect - human lives.

Typically drones are used against targets that are human, I don't see the drones causing less loss of life.

I don't think leaders think of their enemies as human. I don't think they could sleep at night if they did.

s/human/American/

To go one step farther, with autonomous drones leaders don't even need to convince any human soldiers, much less the public.