Interesting point. But it brings up that scarier part of AI. I believe AI will become better at humans at mental processes far before it will be better than human physical labor. For exactly the points you mention. This is already becoming true for many logistical industries. What happens if AI replaces all the high level decision making positions and humans are relegated to handling the low level "last mile" tasks? Where is the inherit quality that makes AI replace unskilled as opposed to skilled labor? Order pickers still exist but the employees that handled what to pick, what truck to put it in and in what order to do it were out of work quite a while ago.
I would generally agree, even for a fairly limited definition of AI. An aside: IIRC, what you're laying out is the "history" of the Dune universe (and I'm sure many others).
I think that computation <i>tends</i> to replace <i>relatively</i> unskilled labor, relative to the skill of whatever created the AI/algorithm/whatever. So right now we're in a situation where software people are automating jobs which are generally less skilled than their own. Which is a little scary when viewed from the haves/have-not perspective I laid out above, but is much scarier from the AI/meatbag perspective you're talking about. It's not so much a difference between skilled & unskilled labor as it is a question of where on the skill totem pole the algorithm's creator resides.
Regarding order pickers, I think that's just a question of economics. When lots of semi-skilled jobs have been automated away and you have thousands of people clamoring for any chance to be paid, it is frequently cheaper to have them do the work than to have a robot do it. Although Amazon did have robots pick orders this last holiday season, suggesting that perhaps economies of scale have finally caught up on that particular type of manual labor:
> Regarding order pickers, I think that's just a question of economics.
I think it will in effect be those tasks that are fully digital. Working in the physical world is a much more expensive and libelous undertaking. So I believe the tipping point for most jobs to be replaced by AI will be when little or no physical action is required on the part of the actor. When a task is 100% digital inputs and outputs AI will be able to use that activity as a training set and replace most of those jobs fairly quickly. That is once AI matures to a point where it can be set up quickly and easily.
