Having AI kill humans on their own without any supervision is an extremely controversial move. It will result in friendly fire sooner or later, and the resulting controversy is pretty much guaranteed to terminate the program.

Laser comms are neat, but they suffer from things like clouds or smokescreens. And once you figure out the wavelength, a hostile laser comms sat can trivially jam it.

It's not that controversial in full blown war zones.

Air defense systems already have to operate autonomously for instance, since not every threat will be detected early enough to allow for human review.

Plus, friendly fire incidents also happen with humans, we don't punish those unless they're obviously negligent, just work on ways to not repeat that mistake, so how is AI any different?

The determination of negligence would just fall upon whoever was responsible for configuring the drone for that mission.

So, for example, if the drone was used to target a certain kind of tank which is in use on both sides and it was known ahead of time that the drone can't differentiate between them, it'd be negligent use, while a mission programmed with full understanding of the drone's capabilities results in friendly fire due to a previously unknown error, it'd be no different from any other current smart weapon messing up.

"resulting controversy is pretty much guaranteed to terminate the program."

I'm not aware of any controversy that stopped weapon progress. We have biological weapon research, atomic weapons and the US used chemical agents in Vietnam.

The US stopped tactical nuclear weapons because it thought they were not useful, not because of the protests in Europe.

SDI was cancelled because laser weapons were not feasable in the 80s, not because of the controversy.

There are more mines used in Ukraine than ever before - despite "bans". There is cluster ammunition delivered despite a huge controversy.

"clouds or smokescreens."

Sure smokescreens can only jam low flying drones? If you smoke a target the drone waits for you running out of smoke rounds.

Pretty sure "controversy" killed biological weapons and most use of nuclear (thankfully).

They're even trying to get rid of certain cluster munitions.

We have enough nuclear weapons to kill off humanity. They have not been used because of fear of retaliation, the number has been reduced because of costs. But the US with it's largest conventional military would benefit from a removal of all nuclear weapons.

We'll see if Russia uses a tactical nuke when Ukraine enters Crimea.

Biological weapons have been killed because they are hard to be used operationally. You don't want to kill civilians, this creates more resistance (shown in WW2 and Iraq for example), also if you conquer a country you need people to run it. Biological weapons are not that useful.

> We have enough nuclear weapons to kill off humanity.

Humanity yes, mankind I'm not so sure. Earth is huge. There are a lot of humans. Some will survive winter. Some will survive mutations. Some will continue to reproduce.

Biological weapons also have the distinct problem of control: you can release them, but you can't stop them afterward. So, you need to be able to keep your own people safe from the release, which means expensive vaccinations across your population. And how do you test that your vaccination works? Or that it'll still work after your bioweapon mutates?

Bioweapons are useless unless your goal is to kill off everyone.

Modern chemical weapons are a better example of a thing we've banned.

But they were banned because of the consequences on their extreme use on the battlefield, not because were controversial before use.

> The US stopped tactical nuclear weapons because it thought they were not useful, not because of the protests in Europe.

This didn't happen... most of America's tactical nuclear weapons were withdrawn when the Cold War ended, but America still has several hundred of them and has continued to upgrade them with new guidance systems.

See the B61, specifically the Mod 3, 4, 10 and 12 varients. These are tactical nuclear weapons with 'dial-a-yield' down to 0.3 kilotons. The mod 12 was developed during the 2010s and is in production now.

Weapons that automatically kill people are prevalent. It may seem trite, but that's exactly what most mines do. Some anti-ship mines are smart enough to distinguish friend from foe using acoustic classification and would be programmed to only kill enemy ships and ignore friendly ships.

Whether or not that constitutes "AI" (pointless discussion since nobody can decide what "AI" truly means even in conversations about state-of-the-art LLMs), it's a weapon autonomously monitoring its environment, classifying targets and killing them without a human in the loop.

Whether the resulting controversy kills the program probably depends on whether there's a conflict between great powers in a total war, or involving pariah states who don't care about making a few new martyrs.

How can you jam a laser from a satellite to a drone, if it is one way? With the sensor on top of the drone, e.G. a Predator? Wouldn't you need something flying higher than the drone?

you can directly target the satellites. Currently in development in Russia, China, and the US.

I thought we were talking about jamming.

But if we talk about killing, we'll get thousands of micro satellites, good luck shooting them down.

So like cluster munitions then.