Conversation

This point leads down a very interesting bunnytrail. Circa 2003-06 when I was doing autonomous systems research, I came across a factoid: an F-16 needs a staff of ~30 to keep it in the air, one of whom happens to be in the air. A Global Hawk needed ~50 at the time.
Quote Tweet
One thing that’s endlessly fascinating to me: despite its place in the cultural imagination, it turns out that about the least autonomous thing ever is a spaceship. They require huge networks of infrastructure, maintenance, and care. twitter.com/julia_bergeron…
2
61
The main motive of UAV research back then was to get humans out of dull, dirty, dangerous roles, DDD. Not to eliminate them. But you can overstate this point. As systems mature, the human support levels can slowly drop. But you still need tons of other automated infrastructure
2
18
In the short term, automation drives up the need for human labor. In the long term it drives it down. That’s why gdp and labor share of agriculture plummeted to single digits over a century. But autonomy always seems to go up, in terms of integration-dependency.
1
16
Primitive slash-and-burn agriculture us actually a lot more autonomous. As in local and not dependent on a broader infrastructure in the rest of the world. Modernization and automation are often accompanied by autonomy-lowering switch to cash crops and global markets.
1
9
And before you know it you need tractors and other stuff, and have tight technological, financial, raw materials/input dependencies. And all your agency gets programmed by incentives designed to accompany those dependency links so you lose agency-autonomybtoo.
1
9
Replying to
Examples: — Operational autonomy is what “driverless cars” are about — Some space missions have limited survivability autonomy but mostly they’re just built ultra conservatively and robustly — Agency autonomy is Stone Age — Intelligence is improving
1
5
I do think true autonomy in the sense many people mistakenly think we’re getting closer to, is possible. Like Wall-E I’d argue was an autonomous robot, and a realistic vision. In a vast junkyard there may be enough spare parts for a sufficiently smart robot to become autonomous.
1
2
Then of course there’s ideas about self-replicating robots and “clanking replicators” that could be legit autonomous in a natural environment. There’s research that shows such systems would need a minimum size (Von Neumann figured that out and bounds have been computed)
1
2
But the interesting thing is that a realistic “autonomous” system will likely get there simply by reproducing something comparable to all of today’s infrastructure.
1
3
The interesting question is whether humans are needed at all, even if it turns out any truly autonomous artificial technological system will be civilization-complete and planet-sized. Ie anything that looks like a Mars rover will have an whole-earth like tech capability behind it
2
3