upvote
That’s not really the case though. Especially in the general case. Just look at the cotton gin. No one was designing little mechanical hands to pick through cotton.

Frequently automation happens only after someone realizes that to automate the process, the process has to be completely reinvented.

I’d say that the majority of successful automation doesn’t follow an iterative path starting from build anything that does exactly what a human does because human physiology isn’t easily replicated with rotational motion.

reply
paradigm shifts usually reveal the limitations and what needs to be done, ie horeseless carriages needing asphalt was obvious from the start

with current gen frontiers, the limitations seem to be requiring a human in the loop, which is quite ironic

reply
Horsed carriages run much smoother on asphalt too.
reply
yes but cars increased the incentive of the large investment required to pave roads
reply
> That is, roughly, how all automation works: You build the machine that does the humans job, barely, badly. Then you improve the machine

That’s not how automation works. The first thing you build is already taking care of a cumbersome part of the human jobs, even if limited in scope. The improvement part is in reliability and maintenance. Because the core problem is already solved. All that’s left are engineering issues.

So even if the first contraception is horrible, it should be clearly visible which immediate need it intents to solve, even if the full scope of changes is not known yet. But with those, there’s no need behind them, just imagined problems.

reply
Given GP's (accurate) point about the mediocre quality of products whose production has long since been automated, I'm curious at what point you think the output stops being shitty.
reply