MICRO DRONES KILLER ARMS ROBOTS - AUTONOMOUS ARTIFICIAL INTELLIGENCE - WARNING !!

Disclaimer: I also work at Google I think that the problem is more specific than working with the military or developing weapons, and it's important to disentangle the more broad concerns with military work versus the concerns about this specific project occurring at Google. I personally would not like to work on developing weapons but it's definitely much harder to defend a position such as "building ships for the U.S Navy is evil". The problem with Project Maven is plugging A.I. into drones. Sure, you can argue that A.I. today is limited and the technology isn't there to make them autonomous. However, perception/object detection is a huge component of what would be needed as part of an intelligence that can run a drone autonomously. Then you arrive at basically 2 questions 1. Is Google's participation Project Maven making it easier for the Department of Defense to develop fully automated drones, or will they always leave a human in the loop? 2. Is enabling the Department of Defense to develop automated drones ethical? #1 Yes, Project Maven enables automated drones. Sure, the Department of Defense could get automated perception technology from somewhere else, but that is in fact what the parent post is proposing Google should let the Department of Defense do. Also, A.I. advances very quickly and if the project continues there may be other ways in which Google is enabling full automation of drones, for example reinforcement-learning based control. I personally do not trust th

I'd envision "kamikaze" drones being the most feasible thing to be effective first. Basically much smarter missiles. A vision of a swarm of something.. Reminds me of this foreboding demonstration of micro drone kamikaze assassinations: this video

Wow so crowdsourcing the military industrial complex's artificial intelligence KILLBOTS? This almost looks like it's satire, but it's completely real and for sale, just $499! Reminds me of this: this video

An earlier dramatization video: this video It comes across to me as a bit over-the-top and alarmist, but, if I consider what I know of technology and (less) history, such events and dynamics seem plausible.

Something like this terrifies me- this video Tl;dw, mini weaponized AI drones get loose and mass havoc ensues. I'm sure the feasibility behind these isn't close but still, it's an interesting design fiction.

Here's the video you're referencing. And yes, it is chilling in its utter believability. this video

Nice picture, but does this give anybody else an uneasy feeling? :-) "Micro Drones Killer Arms Robots" this video "Hacking Microdrones for Lethal Gain" link "Attack on the drones: security vulnerabilities of unmanned aerial vehicles" link "U.S. Army’s New Drone Swarm May Be A Weapon Of Mass Destruction" link "New Chinese weapon launches swarm of suicide drones annihilating targets" link

If they're completely autonomous they won't need radio or line of sight communications. Load them with explosives or potent chemical weapons and it could get really scary. This YouTube short film terrifies me: this video

You won't get any reliable numbers during wartime. I heard Ukrainian government recently mention they ordered a million. Hundreds of small companies (around 10 employees each) in Ukraine are making and assembling drone's right now (according to journalists who filmed them in Ukraine last month). I counted at least two dozen companies in the Netherlands, probably a few hundred all over Europe. Besides the frame that holds the 4 or six rotors together all other parts are off the shelf, mostly ordered from Shenzen. You can CNC mill a carbon fiber or aluminum frame from a single plate on an x-y flatbed cnc. I researched all this last night because I need $20K-$200k investment right now to start my own startup making unjammable autonomous kamikaze drones with 20K payload on top of 20km optical fiber spools unreeling. I think it is possible to make them lighter without fibers if you put a few thousand microprocessors in with a wafer scale integration for $500. Of course you could just make cheap drones that kill every warm body in sight[4] indiscriminately. To disable a tank you need these fiber tethered drones with bigger payloads[3][5][6]. You cab see two fiber strands still in the air in a few frames of the video. What you need is a Paul MacCready style development cycle for drones in the field. You need to be able to make 10 crashes per hour, fix the drone by replacing with off-the shelf components and try again. The programming should be live-coding like Squeak Smalltalk, no r