Vrftjkry

Mining is a good guess, and so is construction.
Moving through a complex environment, dealing with unexpected circumstances, picking up and moving or using wide variety of tools and materials. All these tasks are very hard for robots. Robots can do well defined tasks like mining out a defined ore seam, or building walls on top of a foundation, but defining the seam and laying the foundation requires humans.

Also, repairing the robots, especially after they crash into something

Construction seems like the primary use case to me.

First, automata are generally not very good at improvising or thinking on their feet. They could follow a blueprint, but blueprints are rarely 100% accurate to the actual structure that ends up being built, because they rarely account for absolutely everything. Parts arrive that aren't compatible, the ground settles in a strange way, cables need to be run in the wrong order - these things happen, and humans are flexible enough to cope with them, but automata aren't. They would need either a human guide to make corrections on-site, or constantly-updated blueprints, both of which would defeat the speed and cost benefits of automation.

Second, automata aren't in a good position to judge how the building will be used. A human who's working on a construction site can visualize how the building is coming together and how it will be used. They can walk through to decide if a doorway is placed properly, or if there's adequate lighting, and make corrections if there are problems. Automata can't because they can't judge what humans will think of it when it's done.

Third, automata can be a security risk. If a prankster (or a criminal, or a hostile government's agent) makes a correction on a blueprint, a human can double-check it before acting on it. Automata don't necessarily have that capacity. They could also themselves be compromised, requiring additional routines for physical and electronic security that human workers would not be subject to. It's certainly possible to achieve secure automata, but it would be expensive, possibly more expensive than simply hiring living workers.

Where at least the following apply, somewhat following upon Bald Bear's answer:

Complex novel/unanticipated situations can be expected to occur, even rarely.

Tasks occuring under highly varying conditions, where compensation strategies are not easy to automate.

Any tasks not comprehensively & reliably addressed via script.

Short of advanced AI, robots are best used for rote tasks or remotely guided.

EDIT: As HMI sophistication evolves, especially including sensory input, the utility of remote bots increases significantly. This probably happens more readily than AI evolution, since the technological hurdles are far less difficult, less expensive to develop.

Demolition, artists, craftsmen.

Demolition and some construction, demolition is inherently unpredictable, everytime you move something you have to reevaluate the stability and even value of the surroundings and debris. By definition demolition is often very unpredictable, you don't know what is behind the walls of a building And increasingly building materials are recycled which you want a human for.

"those are nice logs we can reuse those, but those are rotten, wait those fittings are cast iron we can get some money for those! Careful that wall looks unstable!"

Construction will see some use, especially where aesthetics are important. A robot will not understand that a seam is ugly or that the planner forgot to account for some fixture being moved. However a large part of the process will be done by robots. Consider how many modern homes are built in factories as modular units.

Similarly exoskeletons will see wide use in anything in which artistry/aesthetics is important, You want a human placing your statues or doing landscaping, but a robot is probably fine for excavation.

Small scale businesses will see exoskeletons over robots, a mom and pop sawmill will have exoskeletons because they are doing mostly custom work often with odd shaped pieces, a large factory sawmill on the other hand will be all robots. Likewise a auto factory will be all robotinc, but a custom car shop will have an exoskeleton.

Mining on the other hand will be nothing but robots, sans a few planners/managers. It already has extensive uses of robots and for the jobs in mining for which humans are needed, adding an exoskeleton will not help with the job, they aren't doing any lifting. The exception again will be small scale custom stone cutting, like sculptors or custom cutters.

Robots are better unless humans are disposable

Remote control

There's no need for an embedded human. See the following:

Remote surgery

When Mehran Anvari picks up a surgical instrument and cuts into
somebody’s flesh, he doesn’t use his own hands. In fact, he’s not even
in the room. He operates on patients that are 400 kilometres away.
http://www.bbc.com/future/story/20140516-i-operate-on-people-400km-away

The robots can be left to get on with anything their neural nets can cope with. Humans work from their office acting acting as supervisors. They use VR glasses and sensors to sense the environment from the robot's point of view.

This works especially well in mines and other hazardous places unless of course you have disposable human slaves. The 'manufacturing cost' of a human being then has to be compared with the cost to provide an electronic brain.

I've been a firefighter for over 30 years. A working, reliable, exosuit would be a godsend.

Automation really can't do the job (and until you have full AI, can't) due to the variable nature of what first responders do; you might need to perform CPR on an infant and then an hour later are ripping apart a car to get at the occupant after a crash and the hour after that searching a smoke-filled building looking for survivors in the middle of a structural fire. While theoretically telepresence could do it, people don't become first responders to sit in an office.

Idk what 'Warlords' is, but assuming it's some technoverse where technology has bizarrely only improved in one or two fields and those only in being able to present what is capable today more readily plus some weapons technology..actually.. not assuming that, coz it could have any number of odd precepts =)

Mining where?
Machines can be built to operate under constant environmental constraints that people just won't (normally) put up with(that is to say, they will want to get out of those suits) which means you need a living environment, habitat for them, vacations if nothing else.

The only real reason to use people for anything in the long term, is because people will(want) to do it, not because it's 'cheaper.'

Fact is that even today a significant proportion of human industry is redundant (99% of advertising spending is spent trying to attract consumers to one particular variation on a product or service of many that can do that job, for example)

At some point, hopefully long before a century and a half has passed, the general comprehension of 'economy' will reorient to "those mechanisms of exchange whereby things that need doing get done" as opposed to "i have to justify my existence to society so i must convince it it needs something doing" at which point...

Pretty much any argument other than habitation provision, workforce replenishment, mission duration, transport costs & h&s can quite readily be answered with 'whatever a robot can do, a human with robotic support can do also.'

But tele-operation is a thing, and realistically it's much easier to build a single partial rig and a transmitter than it is to maintain conditions suitable for people on-site.

What you said about 'nothing too crazy.' 150 years ago radio communication did not exist, farming was manual labor, modifying the human form was all about surgery and breeding programs, in 25 mankind will be able to reliably synthesise biogenesis...

Question is really, in 150 years will humanities loftiest ambition to be to sit in a tele-op rig? Or to alter itself so that it can do those things it's products would otherwise do better.

AI is good and getting better by the day. If there are power packs capable of automating a human exo-suit they would work even better for Autonomous Mobile robots. Any task a human could do can be done quicker, to a higher precision, and with greater precision.

Three possibilities:

1. AI has been outlawed, or prohibited by powerful socio/economic players. Be they government/labour unions/religions/companies/etc...


2. AI does not want to dig in the dirt, let the humans do that. They even seem to enjoy it.


3. The materials for constructing exo-suits are common, the materials for constructing suitable AI's are not. Some applications will have robots, but there is a greater need then availability.