I'm finding myself having to deal with a lot of squidgy issues in the story dealing with smart robots and computers and the characters and plot.
And of course, I have a Youtube Video to introduce the topic:
Like any decent Sci-Fi universe, I don't just have one AI/Android platform. I have several. So rather than describe one overarching techno-political screed, perhaps its best to tackle them one by one.
For starters, the idea makers in Sublight universe regard "Artificial Intelligence" as redundant term at best, downright confusing at worst. They refer to "Living Computers". They also tend to employ trained animals every bit as much as they employ robots. Each level builds on the level below it. A level 9 has all of the capabilities of levels 0-8.
Naturally this scale is only useful for telling you how close to human intelligence something may be. Many "low level" intelligences can do mind boggling complex tasks. A "simple" artillery computer is capable of performing math calculations that would stump most people.
Quins are mobile robotics platforms, capable of limited independent action. They started as crash test dummies, whose complexity and slavish attention to mimicking the human form led to a line of servant robots.
Quins are designed to mimic human form, dexterity, and strength. Like "Buster" in the Mythbusters, as parts break they can be replaced in the field. The only critical part is the skull, which contains their central processing unit and data recorders. Those are licensed and tracked by serial number. Those can only be produced by licensed suppliers. It is also illegal to open the skull to modify the parts inside. (Though there are plenty of expansion modules that can be legally added that bend those rules.)
Quins intelligence evolved from the holographic recorders their skull was built to house. Their processor is a holographic resonator which can fuse sensor data and several holographic memory cells at one time into a "gestalt". A set of control laws help the robot sort out if there is something wrong with the "gestalt", and take actions to bring the gestalt back into what it was programmed to accept as normal. These Gestalt/Equilibrium/Action sets can be stored and chained together to form very complex behavior.
Most Quins arrive to the user with a library of behaviors and accessories pre-adapted for a given task. Given the limited volume in their skull, and the bulkiness of holographic storage, most Quins have only a single task they can be programmed for at a time.
A limited series, The EHI System 7, included a special "prototyping" board. This allowed the user to swap out the hard-coded behavior matrix in the field. It also contained a special field-comparator that allowed the user to train and store novel behaviors. However, the storage on the System 7s was very limited, because this field comparator took up precious space inside the skull.
Expert systems are level 9 (and above) intelligences that were designed to tackle complex analytical and logistics problems. The hardware is actually relatively cheap, and as most are brains in a server rack, they lack the expensive sensors and actuators in other robot platforms.
This led to Expert Systems being employed in a lot of questionable applications. In particular, finance. The units are prone to novelty seeking, and they were not above sabotaging a project if they felt slighted or if the humans employing it were not deemed "worthy".
In one particularly hideous case, a expert system was put in charge or running a sleeper ship. When the maintenance crew and the computer had a disagreement about the mission parameters, it killed the crew to settle the argument. HAL incorporated insisted (in the inevitable lawsuits) that the computer was in the right, and that it was simply not given SURVIVAL OF THE CREW as a high enough priority mission parameter.
Which highlights the problem of dealing with extremely intelligent beings. They are legalistic, psychopathic pricks. A standing order in every space agency is never to deploy a Level 8 or above in any task. At all. But for older ships, to at least disconnect them from anything to do with navigation, life support, ordinance, airlock controls, reactor operations, food preparation, telephone banks... or basically anything where it could kill more than the immediate operator if it decides to get into a tiff.
(Why they left an exception for putting the operator at risk was never answered. At least to the public. Though frankly if you are stupid enough to program a computer to kill you (or want to kill you) perhaps letting the computer go off and do it is a mercy. If not to you, then to the rest of society.)
Expert systems are generally room sized devices, or in the case of ultra-compact examples, cabinet sized. They are not given independent bodies because... well... they have demonstrated a propensity to view humans as a problem to be solved by the process of elimination.
Many humans have latent telepathy that can be developed to very high levels. That connection doesn't exists strictly between one human mind and another human mind. Some people have developed a connection to animals.
Animals telepathically linked to humans are called familiars. Each bond between human and familiar is unique. Most are unique between one specific human and that specific animal. Although it is not unheard of for a familiars and humans to bond into teams. In either case, it takes years to develop the bond, and constant maintenance to keep it up.
Common familiars are Dogs, Cats, Parrots, Ravens, Seals, and Dolphins. Although, in the case of Dolphins, the humans are often the animal's familiar and not the other way around.
Holographic recording and playback technology combined with advanced medical imaging and psychic projection techniques have led to the development of Artificially Enhanced Life Forms. (They are called KRLs in Krasnovian spheres.) AELFs are living, organic brains that are trained using the same techniques as those used on Quins. However, the capacity for organic tissue to store information is several order of magentudes higher than holographic medium. (Although, like holographic medium, the more you store the fuzzier it gets.)
This can be applied to human or animal brains. Although, using it on humans leads to a lot of delicate ethical issues.
Applications for AELFs are generally to provide a cheap and cheerful means of replicating expert knowledge. The storage and playback technology tends to capture rote memories, procedures, and "muscle memory" tasks. It doesn't capture episodic memory. Or at least, well.
I'll spare you the litany of details, but the long and short if it is, you can only really use the technology to create a newborn with the lifetime experience (and habits) of say an artificer, or a plumber, or a gymnast.
As far as combining the signals into an artificer, plumber, gymnast... it really only works if you have a subject to record who has been all three. Or (future project) you start with a plumber, clone his mind into an AELF, raise that AELF to learn the fine art of Articery, and then clone that mind into a new AELF they you train as gymnast.
You can cut a few corners by starting with a mind that has shown itself to be exceptionally flexible, and then have copies of that mind go off and learn several skills in parallel. And then combine those parallel copies into a new mind going forward.
Although... perhaps that won't work out quite as well, as a lot of skills require developing the same parts of the brain in mutually exclusive ways. You run into the problem of combining Martial arts styles. They each have an internal consistency. They often share basic theory. But the high level composition of moves, or even philosophy, may be hard to train up to a level of expert in more than one at a time.