@duerig What is new is what computers can do today vs what they could do then, how quickly, and with how little energy and heat dissipation. It is far more feasible to integrate computing smarts in otherwise mechanical processes today than in 1990.
@duerig Taking you completely seriously, yes. Automation is a never-ending treadmill. My only question is where are the inflection points at which the implications of automation necessitate reexamining our fundamental economic beliefs.
@duerig I don't disagree with you. But I think that every industry, every situation, now *wants* to apply automation. The task is figuring it out. And so we may see a wave of solutions emerging in another 10 to 20 years.
@duerig But we have crossed the Rubicon not only of possibility, but of practicality in many cases. And continued development in sensors, actuators and algorithms will reduce the inter-step precision requirements.
@duerig Yes, it's hard. Especially in cases that involve multi-step processes. The fewer the number of sequential steps that need to be automated, the easier automation is to potentially achieve. And if there's a recognition/processing step in between…
@duerig Anecdotally, the parking garages I frequent regionally tend to have somewhere between a 1:1 and 1:4 ratio of manned to fully automated stations. In the higher-ratio cases, the human exists primarily to address anomalies/aberrations.
@duerig Why would each new task begin without automation baked in, today? Even as we invent new tasks, we may not create enough new work for humans to offset the erosion on the "bottom." That's really all I'm arguing.
@duerig Advances in sensor technology, databases like maps—these things are allowing automation to move out of the computer and off the assembly line, and into the everyday physical realm with more variety.
@duerig Indeed. The trend line of automation is to consume any task that can be deterministically reduced to a finite number of steps. Advances in computing have allow us to make the step determination more variable, and will continue to.
@duerig I didn't say replicate, which is fool's gold. I said *approximate*. If you have the advantage of well-defined task boundaries, we can do a pretty damn good of *approximating* cognition within an exchange or process.