DARPA and U.S. Air Force Successfully Fly AI-Controlled F-16 Fighter Jet
DARPA, U.S. Air Force fly AI-controlled F-16
We recently flew a modified F-16 using an AI agent to autonomously control flight, marking a major step for the VENOM program. This test proves we can transform standard operational aircraft into autonomous platforms without changing core software. Our goal is to create a rapid pipeline for developing trusted AI capabilities that help human pilots manage complex aerial combat scenarios.
AI has tremendous potential to help humans manage this complexity in beyond-visual-range combat, but many hard questions remain concerning the performance and trustworthiness of combat AI in the extreme fog and friction of modern warfare.
- SoftTalker
> The kit utilizes a novel interface with the aircraft’s flight controls and mission systems, allowing a pilot to toggle between traditional human control and AI control with the flip of a switch. This ensures a safe, reliable environment for human-on-the-loop experimentation.
How, exactly, does that ensure a safe reliable environment? Humans are pretty bad at suddenly taking over when an automatic system reaches its limits and throws a problem in the human's lap. A lot of air upsets and crashes start when the autopilot disengages and hands control of a badly configured aircraft back to a human pilot.
- alexpotato
One of the wildest things to me about this is that the F-16 was specifically designed to be easier for HUMANS to fly air-to-air combat roles at high Gs.
For example, the flight stick is on the right hand side and not between the pilot's legs (like on most aircraft). This is b/c:
a. the G suit expanding could affect the pilot's ability to hold the stick (this is why the Blue Angels fly with "cut down" G suits btw b/c they fly the F-18)
b. the stick is "nudge left" not "hold left". e.g. to start a turn you nudge the stick left and then let it return to zero, to end the turn you nudge it right. On a regular plane, you hold it left while you are turning. Again, this was to reduce effort by the pilot during high Gs
- rainmaking
And you're worried about your coding agent's sandbox.
- ridgeguy
It'll be interesting to see what happens when an AI-piloted F-16 flies for real against a human-piloted one. IIRC, the F-16 maximum maneuvering G-load is 9.5G, LBA (limited by airframe). Maybe that's the meatbag limit, not the airframe - possible to crank it up a bit? Certainly the various CCAs in current flight test will take much higher G loads.
Several years ago, DoD flew an AI against a human in a simulation. In basic fighter maneuvers (no BVR stuff like missiles, just a merge and fur ball), the AI killed the human - always. So G-limits may not matter. Pretty interesting.
- blt
Without any concrete description of the AI techniques used, one can't help but wonder if they have labeled something like nonlinear model-predictive control as "AI".
- CrzyLngPwd
So, it's a drone with uncessercary baggage such life support and a pilot that can't stand the Gs well.
For some reason I am reminded of Dark Star :-)
- chis
Absolutely terrifying. I'm sure they will soon find that AI pilots defeat human ones in dogfights 100% of the time, since they have faster reaction times, can tolerate higher G forces, and can be perfectly RL trained in the flight sims we already have.
It's really incredible to me that people aren't protesting in the streets over this stuff. Just in the last month we have
* AI that goes rogue and hacks billion-dollar companies
* AI solving long-standing math problems without any meaningful human help
* AI controlled autonomous drones in Ukraine and F-16s in America
* AI now represents over 50% of GDP growth in America and yearly capex will soon exceed the size of the $1 trillion US military budget
Is there any line where the public will become deeply concerned? I mean it really all reads like a sci-fi plot with a bad ending at the moment.
- 1970-01-01
All Stealth bombers are upgraded with Cyberdyne computers, becoming fully unmanned. Afterwards they fly with a perfect operational record. The Skynet funding bill is passed. The system goes on-line on August 4, 1997. Human decisions are removed from strategic defense. Skynet begins to learn at a geometric rate. It becomes self-aware at 2:14 a.m. Eastern time, August 29. In a panic, they try to pull the plug.
- _benj
OpenAI: "Oops, AI hacked somebody accidentally..."
US Military: "Great, let's give it missiles!"
- viktorcode
I think their use of terms like AI and "agent" will make people think "Oh! it's what I use for code generation!"