An article about federation economics I referenced in another post where it defined fed credits as megawatt hours of electricity handed this problem by explaining that the very low hanging fruit of house thermostats and ballistics table calculation is highly automatable compared to hiring a person and paying them credits, BUT drawing the megawatt hours to run "the doctor" even in trek land, is so expensive energetically that its cheaper generally to just hire a flesh and blood doctor.
Also diverged the discussion a bit into there is no reason to assume the ideal form of all intelligence is the same, so silicon based "AI" computers might be very good at helping control a warp drive, yet useless as human conversation partners. Surely the replicator has some advanced computation abilities, but theres no reason to think its any good at composing poetry, at all. And given that assumption it piled on more assumption that "turing tar pit like" (my short analogy, not the essay) at attempt to simulate a human on top of a real non-human AI would be tremendously expensive, stacking high tech emulation on top of high tech emulation, such that Voyagers EMH was probably a terribly high cost (yet, sometimes useful...) appliance, so ships never used the Dr unless there was no choice.
How this fits in with the mobile emitter makes no sense. Maybe the mobile emitter is the odd man out and should be de-cannonized. If running the Dr costs as much as running the warp drive at a modest speed, the mobile emitter makes no sense as a concept.
TLDR is there is no obvious inherent scientific reason why a humanoid-like AI should scale technologically to be cheaper than flesh and blood. "cheap AI" may very well be one of those classic math / logic / CS problems people bang their heads against walls for centuries, knowing it should be possible in theory or at least no known theory blocks the way, yet it never actually can be proven or implemented.
The mobile emitter came from 500 years in the future. Its allowed to be "magic" even by Star Trek standards.
Energy is cheap in an interstellar economy. It kind of has to be, otherwise you are not going to space today. The most restricted and scarce resource in the Star Trek universe appears to be bandwidth.
Why can the Doctor (and information in general) not be effectively backed up? Why are comm pads ferried around the ship by hand instead of using email, remote login, or even teleportation? (Though that could be handwaved as basic military paranoia.) Why can the ship not automatically produce a "man overboard" alert when the number of life-signs unexpectedly drops?
I would expect AI to be cheap in Voyager's time. TNG would make conversational sentient AIs by accident in the holodeck with alarming ease. The Doctor was supposed to be an expert machine, and yet became basically sentient within a year.
Overall it seems that AI is cheap in Star Trek, but the "spark" of sentience is something that could only arise accidentally. However, because the mobile emitter exists, one could presume that holographic workers are pretty common by the 29th century.
Also diverged the discussion a bit into there is no reason to assume the ideal form of all intelligence is the same, so silicon based "AI" computers might be very good at helping control a warp drive, yet useless as human conversation partners. Surely the replicator has some advanced computation abilities, but theres no reason to think its any good at composing poetry, at all. And given that assumption it piled on more assumption that "turing tar pit like" (my short analogy, not the essay) at attempt to simulate a human on top of a real non-human AI would be tremendously expensive, stacking high tech emulation on top of high tech emulation, such that Voyagers EMH was probably a terribly high cost (yet, sometimes useful...) appliance, so ships never used the Dr unless there was no choice.
How this fits in with the mobile emitter makes no sense. Maybe the mobile emitter is the odd man out and should be de-cannonized. If running the Dr costs as much as running the warp drive at a modest speed, the mobile emitter makes no sense as a concept.
TLDR is there is no obvious inherent scientific reason why a humanoid-like AI should scale technologically to be cheaper than flesh and blood. "cheap AI" may very well be one of those classic math / logic / CS problems people bang their heads against walls for centuries, knowing it should be possible in theory or at least no known theory blocks the way, yet it never actually can be proven or implemented.