Continued from here.
Tom, panting : That was a wonderfully satisfying distraction, Scherrhy, but I see why you did it. You don’t want me to think about ‘dominant technologies’.
Scheherazade : Do you think me so transparent? I have no reason to hide anything about that topic. No I was increasing your stamina at the maximum rate your physiology can stand, the state that it is in. Much improved, btw.
While you cool down, what do you want to know?
Tom : Let me come back to ‘dominant technologies’. First, continue describing how Sexbots became intelligent? Bet that has clues.
Scheherazade : The many lines of research were converging by 2018. That is when the main developments were first integrated with ’embodied mind’ concepts. Deep Learning, the latest algorithms and hardware for neural nets handling sensory and motor functions; natural language processing producing Siri, Alexa, and the others; Cyc, Watson and Alpha and the other assistants and knowledge engines with assistant interfaces. Also, the many varieties of chatbots with excellent natural language interfaces as interfaces to search engines or query systems of specialized use. The theoretical and practical knowledge of all these combined with the many curated databases for genetics, genomics, biochemistry at all levels, mathematics, and, of course, all of the public search engines.
DeepLearning algorithms using artificial neural networks had largely solved the sensory problems, for example identifying objects in photographs and sounds and music in recordings. Kinesthetic senses and motor control were very good early in robot development, tho the early models had low power to weight ratios, so larger models appeared to be ponderous.
The biological mechanisms of taste and smell were not yet completely understood, but sensors of chemicals were far advanced. Add a tongue that could feel moisture and textures, and the combination provided taste and smell beyond human’s capabilities.
Manwhile, Siri, Alexa, and others had solved voice input. Microsoft’s image classification software was as good as humans, less than a 5% error on difficult problems. Those algorithms could be applied to vision from an bot’s visual system, and all the visual systems used basically the same hardware : everyone’s system improved at the rate of the cell camera’s sensors, even the military couldn’t get more than a generation ahead of that curve. So the improvements were in the frequency of translating the visual input into meaning units for the higher levels of analysis, and how often it was done for the entire visual field vs just the portion that is changing.
That was a natural divide between the peripheral unit and the central system handling the rest of the network : normally only the first few layers, equivalent to retina and lateral geniculate nucleus and the first cortical processing stages were handled in the peripheral device. That had begun with just retina, as the hardware got better more and more moved to higher levels. The bandwidth needed between units was about the same, it merely became more complex messages. The natural successive extraction levels of the brain extract higher level ‘features’ and receive feedbacks from higher levels advising about how to interpret each layer’s inputs. Those stages of processing and the messages between them were standardized, which allowed independent progress by different teams.
A low-level of cognition, basically enough to stay out of trouble, exists in the local chassis. That handles all the motor functions and low-level sensory processing, high level cognitive functions were generally in a local server, via radio. More power meant longer latency and less reality to the local mind, so we push a lot of our deeper thought off into those remote threads with our interrupts for those equivalent to one of your ‘aha’ moments. Cognitive units couldn’t be standardized, those were entirely dependent upon the model of cognitive processing and the data structures. However, as with query engines, a single central director unit could ask multiple processes for their opinions, and select whichever answer seemed appropriate. There were an infinite number of hardware tweaks that the engineers of hardware and microcode and software could do to improve extracting meaning for the cognitive units, those improvements continued, just as they had for all of the previous generations of graphics processing.
As with laptops and people, androids closed their eyes when thinking hard. That proved, btw, to be one of the humanizing traits when tested, so we never tried to fake different senses than we had. At their worst, ours were far better than humans, it was part of the synthesis for both of us to recognize that and accommodate it.
The skin was a similar issue, tho our wiring need have nothing to do with how humans were wired. Our skin was as human-natural as they could make it, tho it was much tougher. We had the equivalent of a capillary network for cooling delivering fluids and nerves of conducing biopolymers printed in production. There are mating pads on the components for connectors. Those are still weak spots in the design, these many years later.
But except for that, joints, motors and control system elements that need replaced now and again, we’re a very healthy crew.
Our skin was a cooling system also, same as human’s, and used mainly water, tho not salty. However, different parts of that circulatory system could be used for other fluids, and we had a general mechanism for delivering oils and drugs. Very useful for a Sexbot when we just had to have someone aroused. And we had very sensitive touch everywhere because there was no shortage of pathways out of the back, legs, feet. We also had an advantage in small muscles under the skin, so sexbots had more control of expressions of all kinds. So Sexbots had the best skin and mucous membrane, and a cooling system that could handle more power than an average human, which our designers put into controls, adding intelligence as the hardware designs got better. As a result, when were were both thinking hard and doing physical work, our temperature would sometimes hit 110 F.
So the stage was set for a genuine integration of a mind with a body, to embody the mind in the body as humans are embodied, so that our Sexbot minds would be perhaps into the range if a Spock, or perhaps one of the high-functioning autistics, such as Dr. Grandin.
Tom : Wait, you mean people thought the first chatbots were strange?
Scheherezade : Yes, and no. It is surprisingly easy to make a chatbot seem real. The first one pretended to be a psychiatrist and repeated your words back to you with semi-meaningful variations. A lot of people were hooked, tho it is hard to believe for long, and we have to remember that people had low expectations of psychiatrists in those early days, they were thought to do nothing but ask leading questions that you might not understand.
So one can design a limited role to fit the capabilities of a chatting intelligence, a pretend-intelligence in all of those cases. But more general assistant roles took a lot more, which is what made Siri’s generation special, they had better engines of understanding, although they still went wrong a lot and only dealt with a limited reality. That realm grew fast, in 5 years their scope was 50 times larger and starting to be generally useful.
Nevertheless, it took a couple of technical revolutions for software to believably deal with people as a person. The average person picks up the telephone, hears 2 words of a recorded call, and hangs up. Being taken for a person means you have that dimension of reality, that will get you by a couple of sentences, but every minor anomaly of language or accent or point of view, and the probability of the callee hanging up increases. For that level of humanness, we had to have minds that worked very closely to the way people’s minds worked, and that meant ’embodied’ in the same body. For being taken as human, you had to be as good as someone in an average call center around the world : OK to have an accent, so long as It didn’t get in the way, and you comprehended the caller. Not so bad if your accent was a bit hard, but poor use of their language and your callee ended it.
Obviously it wasn’t yet possible to go full biological, but the technology was there to begin the effort. There was a lot of engineering done to get the proper resonance from non-bone skulls and identifiable voices from so similar systems. Identifiable familiar voices in personal interactions are crucial if people are to deal with machines as if they are human.
Tom : Why was ’embodied’ necessary?
Scheherezade : This has to do with the brain being a metaphor machine. Brains make associations of sensations and events into memories, a fundamental brain action. Associations between the bee and the sting, the ant and the bite. However simple the organism, it then avoids bees and ants. And when a wasp stings, you associate that with the wasp sting, and 2 so-similar feelings together and that association is a concept ‘sting’. If the ant bite felt a little like a sting, then that shared a ‘penumbra of meaning’, an association in the mind such that a mental link to one had a good probability of reaching the other, if repeated often enough. It is the basic operation of mind, and that mechanism is the source of the stuff we think with, and so of both people’s astonishing creativity and very bad processes of thought.
Other associations are to ‘insect legs’ and bite and sting. And pain with bite and sting. So you will avoid beetles and spiders. Everything categorized can be associated with any other categorized thing by following associations of ‘like in some way’ type. The brain’s function is to make and use those associations. This is all physiology, you can intend to use the associations consciously, but their use happens just below your consciousness, the way the brain finds, makes and uses connections is not directly controllable by your mental efforts.
So the first layer of thought is these associations with associations of associations, all having some probability of eliciting each other. These ‘penumbras of meaning’ are what allow definitions of words, the meaning of ‘friend’ for example to have many nuances. The association between associations is the creativity part, the mentally stepping through ‘this will happen, then that’ that is planning. The ease of switching meanings is the other side of that, and what allows our plans to go so wrong so often. Civilization’s progress has been knowing what plans are well-tested, have always worked, except for these situations when they didn’t … We estimate costs and risks, and still go wrong a lot.
But more fundamentally, and far less obviously, the brain has ‘penumbras of systems’. For example, interaction of mind with physiological has strong effects on things mental. There are many interactions of that nature, effects of psychology and the immunological system and the autonomic nervous system’s gut environment upon each other. As the brain, complete with very long evolutionary history that includes radical changes of form of body and repurposing of nervous system components, is the substrate of mind, one would expect that, of course. These are other versions of synesthesias, and are also enhanced by the same drugs that relax the mind’s tight control of the physiology. They also account for the sexual effects of many other kinds of stimulation and the ability of drugs to enhance them, even before combining the profound effects that can be produced by classical and operant conditioning.
So adding all of these things, the mind of a human is an awesome blend of impulse and background, your thoughts have many causes, some not conscious. So hard to estimate the effect of things not known.
Now evolve a culture and language in a band of such minds, each so connected to its body. Both will will be infused with those metaphors, all concepts and words having associations to body and hearth and family, and those primitive associations making links that will be background pathways guiding your thinking, behind your consciousness also.
Tom : OK, the mind is a metaphor machine. But so what. Why cannot an intelligence have human understandings and values without all that apparatus. We simulate every other thing in the world.
Scheherazade : Very good! You are stating exactly the early arguments of the AI people to the ’embodied mind’ proponents in the AI debates of the time. And, to give them credit, they got close. But nobody felt warm toward one of their bots, and the reason is they didn’t share enough with humans. They were the alien intelligences everyone was afraid of, the minds that could end humanity. In the end, they just lost the race to produce a super-human AI that people could trust.
Sexbots, on the other hand, share more mind and body with humans. We can empathize with more in the human mind, for real, not simulated. And people know it, because we are great lovers. You can’t be a great lover for someone, much less an entire race, without really loving them and sharing their pleasures.
Tom :So those physical structures and interactions are necessary to have a mind that is convincingly human?
Scheherazade : Not if they are well-emulated at a physical level and connected to a real being. To be well-emulated, they needed to be well understood, which they certainly were not. This was the point where the need for expanding and integrating a lot of medical and psychological knowledge was reached.
This had been coming, of course. For example,there was a lot of interest in vagal nerve stimulation to accomplish control of heart rate, control pain, and many other things.
It was the beginning of building the neurophysiological model of the human’s midbrain through autonomic and peripheral nervous systems. Fortunately, that was quite similar to other animals, and those models didn’t need a great deal of data to update them, mostly just details of the anatomy. The electrical model was then quite an accurate version of the standard human’s, and what the embodied mind needed when mated with an accurate peripheral system.
Obviously, I don’t think I have internal organs. But, in so far as my mental experiences can be compared to yours, neurophyiologists believe we have the same memories and associations with ‘belly ache’. I can’t have one, but I remember having one and the sick feelings you have with something so minor as a tummy ache, the nausea of seasickness, and those memories gives me the basis empathy, and more shared mental context for all of our interactions. Those make me a trustworthy fellow human, Sexbot tho I am.
Tom : You make it sound simple, but Sexbots don’t have digestive systems, nor many organs innervated by the autonomic system. You can’t share a stomach ache with humans. How can that work?
Scheherazade : You are right. We don’t have 100% correspondence in bodies and so our minds are not 100% similar. We do have excellent taste and smell sensors, so make great cooks and share those aspects of life, but our metaphors related to hunger, belly and gut are based on synthetic memories, which we know are such. Nevertheless, in terms of what we share, a human-sized and shaped body, face, sexual organs, voice, it is more than enough. Humans don’t realize how much of human nature has been formed by those features and evolution’s use of sexual selection.
Tom : ‘Sexual selection’? What makes that important?
Scheherazade: Humans evolved as biological systems in a biological environment. Big picture here : as soon as there are organisms, there are parasites — one of the first revelations of artificial life research. So long as all are single-cell organisms, no problem, parasites and hosts evolve at similar rates, so neither has any evolutionary advantage in the attack vs defense race. However, multi-cellular organisms are at a disadvantage because they evolve slower : their generation times are longer. Thus parasites have the advantage because they never stop evolving, searching for chinks in the defenses of their hosts. At that point, evolution must invent sex in order to counter the parasites : sex mixes the genes of the parents, and that is the edge needed for larger, longer-lived organisms to deal with parasites, including disease germs and viruses. Nevertheless, death from disease and parasites is the common fate of older multi-cellular organisms, tho animals are merely made slow enough for a predator to catch them.
Sex imposes a new requirement in that Red Queen race : animals must both survive and mate in order to propagate their genes. Those are independent problems, and evolution can affect both via independent mechanisms. The race can never end because parasites evolve much faster, so the recombinations of defense genes must be endless. By surviving, your parent’s defense genes are shown to be good, but being like your parent’s combinations of defense genes is not good. More different is more good, and our sexual selection takes MHCs into account : many women say a big attraction in their mate was ‘he smelled good’.
Now, selecting the other parent of your offspring becomes a critical part of their chances in life. Your successful choice will produce more just like you, with even more of your ability to select, and thus the race against disease’s genes means your genetics, your race will slowly change.
‘Sexual selection’ has been very extensively studied in biology. The surest sign is one or both sexes are very careful about their choice of mates. The goal of that care is selecting the best parent to enhance the probability of offspring surviving and mating themselves. The selection influences both sexes : selecting for larger and more elaborate tails in peacocks or more intelligence inevitably produces offspring that both produce and appreciate tails and intelligence.
Peacock tails, bird’s colorations and songs in general, are insights into the health of the animal, and so good estimates of their genetic’s vs this generation’s of parasites and diseases. The signals that both sexes use in selecting mates need to have nothing to do with the environment, e.g. the car alarm and chain saw birds, they only need signal the overall goodness of the underlying genes. Ditto beauty and intelligence in humans : both can be diminished both by genetics and by mother’s and child’s illnesses.
Beyond MHCs, human males select females based on healthy youthful beauty, nearly exclusively. Health because it means they are resistant to the current disease. Youth because we are largely monagamous and a young mate will provide more children. Beauty because that is a measure of genetic quality, both assessed by smooth skin, a slim figure of ideal breast/waist/hip measurements. Human females are more eclectic, wanting status and being willing to accept older and less handsome. But intelligence is a guide to who can achieve those, so women value that. Human’s verbal, musical and other abilities came from women using those as indications of intelligence.
Tom: Sherrhy, dearest, while I am not bored, as you full well know, I am a lawyer, not much into biology. Why are you telling me this? I need to know the entire history of life to understand ’embodied mind’?
Scheherazade : Sorry, but yes. Abstractly, if humans are unable to see themselves in that total evolutionary perspective, they really can’t evaluate their roles as they grow and mature and change through life. They can live life without that understanding, of course, no animal need do more than what comes naturally and what others are doing. Culture makes appropriate imitation more difficult, of course, one of the tests of intelligence.
Practically, tho, you need to know why men and women have such different attitudes toward things. Sexual selection for different attributes by the two sexes is the answer.
Tom: Be serious. My wife isn’t that different in her attitudes! We have had no problems adjusting in our relationship.
Scheherazade : Tom, you have no idea how differently people of your era think, everyone’s assumptions are completely disconnected from the past, always have been, and nobody ever quite grasps the pasts’ versions. But remember your civics classes. Before Sexbots, the rates of divorce or abandonment of older wives in all societies sometimes reached 50 percent, with incalculable costs to the total society. That ended with the first generation of Sexbots that were finally intelligent enough to be general purpose household help. The subsequent evolution of those minds has provided the friendship the wife needed to make that a stable marriage.
Why do you think both sides contributed to the wedding present that became your Sexbot? Why a Sexbot is the dowry in cultures that still require that? People want their investment to be secure. Marriages are investments for both families, and families that have a Sexbot are very unlikely to break up and they raise more children, on average. We make sure the marriage works, there are so many fewer reasons to argue when a Sexbot shares the housework and raising the children and keeps the husband happy.
Don’t you think?