Destroying Russian Tanks Is Just the Start for U.S. AI Drone Autopilot | Hacker News Reader

Destroying Russian Tanks Is Just the Start for U.S. AI Drone Autopilot

forbes.com

18 pointsjdmark10 comments

The next innovation will be networking (jam-resistant radio relays and/or light) and patrolling, which will optimize coverage, report hot spots, and fill in gaps from attrition.

The end result will be an air traffic control system that creates a minefield of the air—an area-denial death zone.

The real innovation with drones, though, is that there is no more "fog of war."

Drones provide a perfect information environment.

In 2019 I came up with the theory that one of the primary pruposes of the Starlink system may be to control autonomous drone swarms globally.

The development of perfect AI is tons of work due to the huge amount of possible different optimizations. Consequently, to the casual onlooker the early systems will look crude and later systems dangerously advanced.

The second primary engineering challenge is charging grids and logistics / manufacturing. NATO and carriers already provide lots of opportunities. I would however predict that some new, easy to deploy at global scale, form of charging grids will be in use within 20 years.

The advantage such droneswarm would provide over other nuclear nations cannot be overstated.

A lot of drone autopilots will use commercial, off-the-shelf (COTS) chips and software. Imagine an exploit for a common example program or for a library in a common development environment.

A state actor could build a database of ways to profile and exploit those chips and software.

Imagine hacking a drone and ordering it to return to home and drop its payload there.

We may see firewalls on drones soon, or watchdog timers that trigger a second microcontroller to take iver when the primary is hacked.

Metasploit already supports software defined radios (SDRs.)

Theres rapid innovation in this space ever since the war started. Its going to have interesting side effects once drones like these will start being used for policing as well. Combined with ubiquitous surveillance and LLM summarized real time profiles of people its going to be a killer app for stopping any crime and dissent before it even happens

Or we can have bomb loaded FPV coming out of nowhere and ramming into politician while being guided by face recognition software so all that EW stuff is useless.

They’re calling the AI enabled killer drone skynode not skynet so we’re safe people. Carry on.

Targeting command personnel jogging in the park comes next, I’d imagine :-/ Not hard to predict where it goes after.

Can we at least setup some kill switches, so a random person won’t be able to target schools?

The US has never and still does not care one iota about bombing schools. The answer to this question must therefore be "no, such a kill switch cannot be added". These are war crime perpetrating machines. Adding in a mechanism to prevent war crimes is entirely against the point.

This is not true, the US does have a strong preference to avoiding bombing schools.

A human is always in the loop too. From what we've seen with other AI systems is that humans will start rubber stamping targets left and right to reduce their work burden.

Reply on news.ycombinator.com