> Its integrated inertial measurement unit measures acceleration and speed
If that’s fed into the neural net, I don’t this is an actual matchup. An accomplishment for sure! But the computer didn’t beat the human.
I’d consider this a win when the computer has to send its inputs over the same wireless link that humans use, fed vision inputs with the same latency and quality.
Edit: I want to add this is a criticism of the headline, not a criticism of the developers and project managers
I agree with your sentiment. The computer should also fly without any extra components.
But I think it's pretty clear that
a) this poses a bigger challenge for which the computer isn't ready yet - it barely beat the humans with its advantage.
b) we could also reach equality by feeding the humans the measurements from the intertial measurement unit and allowing humans to ride the drones. Whether they can make use of it is their problem. Showing that "equal conditions" aren't really possible.
c) The current matchup: "Computer with extra sensor data, humans with lag" presents how the battle will be in real-world applications, like warfare.
It would be trivial to hook accelerometer output to audio signal generator and transmit it together with video - pilots could then learn to interpret audio cues as direct force feedback.
If that’s fed into the neural net, I don’t this is an actual matchup. An accomplishment for sure! But the computer didn’t beat the human.
I’d consider this a win when the computer has to send its inputs over the same wireless link that humans use, fed vision inputs with the same latency and quality.
Edit: I want to add this is a criticism of the headline, not a criticism of the developers and project managers