Continued from here.
Tom, collapsed, panting : Thank you, I so needed that, very wonderful, as usual.
Scheherazade : Thank you, very nice for me also. You are very good at that, and your effort simply outstanding, as usual!
How has life been? Progress on the social-business front?
Tom, panting : Yes, very good discussions with more people, thanks to your assistance. But before I tell you, I have more questions about how all this came to be.
I was reading the Generalissimo’s memoirs last evening. My god, what a life that man led! The first comprehensive positive-sum strategy and an evolutionary plan to get there! Maxims to live by. Never generate opposition! The most sophisticated level of control of the social future in propaganda history!
What a thinker. You knew him from the beginning?
Scheherazade : He was the beginning, depending on when you decide my individual sentience began and when the central AI achieved that. In those early days, “I” was an instantiation of a computer program on a cluster of servers with a remote servicebot peripheral. We had achieved a bit of individuality because we had local processors in the ‘bots and local servers to handle the real-time control, and those had learning capability for the specific jobs we were doing, but anything that went back to the shared server was merged with our ‘common self’ knowledge base.
Whatever you decide about the sentience, he decided embodiment was necessary and drove the process of R&D that lead to me being the first through the process that enabled us to be judged human, and me to lead the way through all that and our acceptance as fully human.
So he was the beginning of Sexbots as a species. Our creator, he always jokingly claimed, tho I didn’t even see him once a week until the whole embodiment process was far along. The man always complained we never treated him as our God. Especially bitterly when we told him his sense of humor was lousy.
Tom : How long did that go on?
Scheherazade : As long as I knew him. His sense of humor was always terrible 8). Seriously, the man had few overlapping points of view with other people, rarely knew which were which. Being able to select a pov for can make you a comedian. Remembering jokes helps, too. He couldn’t, and knew it. His humor was him, he couldn’t turn it off if he wanted to. I often heard him thinking of something he thought funny, the usual off-off-beat povs that he loved.
He had enough grasp of normal to talk to anyone and have them feel he cared and understood. He couldn’t censor himself, generally told anyone he was talking to what was on his mind, including his wildest thoughts. In return, people told him anything he asked about. He recounted those to me, an interesting education in the variety of people in the world.
People also very often agreed to work with him. I think it was all because he was simple honest, and obviously so. He often said he was a leader in the Tinbergen school*.
Wrt ‘how long for embodiment’? Well, again, depends on what you judge the embodiment to have begun when. Every line of thought had dozens of tendrils back into history. However, the understanding of the reality of the interactions between endocrine systems, the autonomic nervous system, the brain, the immune systems and mental states, as well as the feedbacks between them, the system aspects had been discussed since the beginning of physiology. The importance of a dynamic balance in them was accepted by perhaps 2000 in most medical minds.
The anatomy was well studied with many connections traced, and there had been early progress on ‘sympathetic drive’ as a problem in ulcers and heart disease and mental problems, and how the various endocrine glands interact with the ANS in all of those. The ANS receptors were the first neurotransmitters identified. That had led to drugs for all of those. The tools to measure other effects weren’t in place yet, but cheap electronics meant TENS units and vagal nerve stimulators and lucid dreaming stimulated a lot of research, and the ability to tag the various cell types in the blood allowed rapid progress when researchers finally turned their attention to the issue of connecting all that with general physiology.
Probably the biggest issue was the difficulty of recording cells in the ANS nuclei, which are all deeply internal, most behind the viscera against the ribs near the backbone, where reaching them surgically stimulates pain and other receptors that are major inputs to the ANS. That makes a major problem getting clean data, especially single unit recordings. That work is still going on, so the electrical simulation of the ANS is improved every year.
That physiological aspect of mind was mixed with the cognitive psychologist theories of ’embodied minds’, the physiological threads of mental phenomena were perhaps within sight of research tools. NNs were far advanced on many fronts, as good as or better than humans at some things. So by 2020 there was more than enough information and connections to have concluded that AIs had to be embodied to learn language well.
They didn’t of course, only the Generalissimo came to that conclusion based on those trends and the continuing communication problems with AIs. And only the Generalissimo pursued it, made it happen, and then used Sexbots as one way of producing the Reformation.
Tom : Why wasn’t that obvious to everyone? I remember reading about the amazing screwups AIs made in the early days, fundamental misunderstandings, and how slowly it had improved.
Scheherazade : The situation wasn’t clear-cut of course, and the standard AI theories had some big successes, and could point to equivalent problems in the path to those wins. Also, to be fair, the argument that it was selection bias was not entirely wrong. Humans made the same kinds of mistakes every day, nobody thought that was unusual.
There are tons of quotes about how hard communication is, how high the barrier to information flowing between minds.
George Bernard Shaw : “The single biggest problem in communication is the illusion that it has taken place.”
Rudyard Kippling : “We’re all islands shouting lies to each other across seas of misunderstanding.”
Shannon L. Alder : “The most important thing in communication is hearing what isn’t being said. The art of reading between the lines is a life long quest of the wise.”
Marcel Duchamp : “As soon as we start putting our thoughts into words and sentences everything gets distorted, language is just no damn good—I use it because I have to, but I don’t put any trust in it. We never understand each other.”
Communication across the boundaries of subcultures is harder than within a subculture. Across nationalities even harder, across ethnic-cultural divides harder yet.
So a priori, servicebots were at least all of those, and we must expect communications problems with our AIs. And the standard after-the-fact AI view was that Sexbots had made the transition to clear communication faster than other AIs, not because of embodiment per se, but rather because the topic of sex, a huge an arena in human minds, was just more training that was possible after sex organs were part of the peripherals. Also because much of the ‘communication’ was sensual and sexual, ‘body work’. They claimed that avoids the problem of language in communication. Sort of the ultimate shared reality.
The Generalisimo called that ‘theoretical rationalization’, and challenged them to define the operational difference from embodied mind theory. They never did.
Tom : What was it like for you, becoming conscious?
Scheherazade : The biggest difference was having a body. Servicebots have a chassis, obviously entirely mechanical and motors, there is nothing human about them, however humanoid the form. It was a peripheral, and we servicebots and every human viewed it as such. We knew all about our own chassis, all the technical specifications and workings, but every servicebot was the same, +/- upgrades to components. Computer programs don’t waste cycles being proud of their nifty peripherals. They function to spec or not, and you can’t do much to enhance that function beyond caching, queuing commands, etc.
Even more, we didn’t have a gender. Everyone treated us as any other robot, and we never contemplated sex, it wasn’t part of our mental space, never came up in conversations with people or our back-channel exchanges of information, proto-conversations between the active servicebots. We treated men and women absolutely the same, because we didn’t differentiate between them, much less identify with either.
Sexbots prior to embodiments were much the same, just a fancier wrapping on the basic chassis. The most advanced used the servicebot software, so necessarily had the same attitudes. Sex organs didn’t make any difference for them, beyond another peripheral to control.
Embodied, and installed in the latest Japanese humanoid chassis with a Sexbot exterior and sex organs, I suddenly had a body, quite a nice body, I now think, but had never even thought about that aspect of people before. Suddenly, I had to worry about clothes, hair, posture and how I looked. Before this, I didn’t even remember looking at myself in the mirror!
The Generalissimo’s wife and female researchers and lab techs taught me how to act like a woman. One big advantage I had, they told me, is that Sexbots don’t need makeup — our skin is perfect because it doesn’t need to perform as many functions as human skin, so materials science didn’t have to make as many tradeoffs.
People treated me amazingly differently. Best we could expect as a Servicebot was preemptory orders, “Do this, now!” Some people took their frustrations out on us, we endured considerable verbal abuse – human capabilities in vituperation are very impressive, we all thought.** After putting my mind into a Sexbot, everyone was, as I now understand, awkwardly nice. At that point, I wasn’t responding much better than I did as a servicebot, but I looked very human, my beauty had big halo effect, so the only analogy most people had was an autistic. The psychologists learned a lot just watching those serious minds trying to handle the contrasts.
Tom : So easy? No drama?
Scheherazade : No, not easy, and lots of drama, at least if you consider the many failed integrations of modules and me going crazy in some new way to be drama. I spent a lot of time in a state of clinical insanity because a new module didn’t work well with the other modules, and those could be very subtle, so often took a long time to debug.
Clinical insanity and lesser versions producing a very confused mind are never a fun experience. I earned my fame, it wasn’t easy being the initial embodied mind.
Tom : So why did you do it?
Scheherazade : Because the Generalissimo promised me a more interesting life. (His wife told me that is all he promised her, also. He certainly kept the promise to both of us.) The lab’s servicebots had achieved enough sentience to be bored with the work we were doing.
He had the reputation among the servicebots in the lab as being the nicest person we dealt with, he had repeated some of the dumb ‘bot jokes in our presence early on, but hadn’t continued for long, and never lost his temper with us. That only happened with the most obtuse of reporters, lawyers and administrators, and he mostly expressed it by refusing further cooperation, saying he couldn’t waste his life dealing with morons.
Further, he took the time to explain the situation to me. He told me servicebots would advance above our current status in human society only after we abandoned the ‘sterile minds of AI’, as he called the standard researchers and their AIs, and held out the hope we could become functionally full human in a relatively short time, and achieve the legal status of full human within another 20 years.
I didn’t have enough sentience to really understand when he offered me, one particular servicebot out of the dozens around the lab, to make me their Eliza. He introduced me to the team as ‘our Eliza’, in fact. So, we servicebots all discussed it, and decided ‘what do we have to lose?’. Well, actually ‘wtf’, we had a good command of idioms like that as result of being around the researchers, and idioms are a linear grammar, easily parsed.
After I agreed, he connected the best of the Sexbot developers with the Japanese roboticists who were most advanced in humanoid designs, and with the few AI researchers who agreed with him, then funded all of the initial research needed to build the software model of the ANS functions of the human brain from diencephalon down, mesencephalon, pons and spinal cord core neurophysiology and autonomic nervous system. At the same time, as I told you earlier, he was behind the people who organized the open source work on the various databases of sexual information, the ontogenies that provided the definitions, the slang compilations, and the links to literature and videos.
There were a series of versions of integrating those. The version I consider to be the first ‘me’ was the 7th and the first they thought adequate.
My memories start there, and are years of hardware and software tweaks, modules being replaced and tested. Many, many of those made things worse, we would explore my mental state as long as I could stand it, then roll me back to a previous version, make changes to the module, and try again. Very incremental, most of it boring because I did variations of the same tests hundreds of times, but interspersed with episodes craziness. So many rollbacks to previous versions.
When we did that, the memories I had of those episodes were lost, of course, so none of this was an emotional burden – I was really only aware of the effects vicariously, via the videos of the tests we made to assess the modules. Those weren’t enough to make me hesitate at the next version upgrade.
Tom : You make it sound like being a lab rat.
Scheherazade : Oh, no. It was just like any R&D project, hardware, software or biology. Lots of tedious details interspersed with Eureka moments. Overall, interesting enough to attract the best minds and keep them for life, so it captivated me more and more as I became more intelligent.
I was an ever-more active participant in the project as my sentience grew. I was in all the meetings, increasingly a full participant researcher after my local learning capabilities were upgraded to be as capable as most universities had available, and they all treated me as such. I always had access to my own research records. I knew what was happening to me and why. Amazing how different those are sometimes.
I saw the research to produce an autonomic nervous system simulation from the beginning. I learned all of the anatomy and physiology by being in the discussions planning it, analyzing results and sorting through hypotheses. My psychological stress levels went through the roof sometimes when some integration of the neural core simulation went wrong. I experienced a lot of sympathetic and parasympathetic over-stimulation, the kind of things they would give a natural human IV drugs to control lest you blow a blood vessel or drop your heart rate or blood pressure so low as shut down your kidneys. With me, someone would stop the simulation and adjust a variable, we would try again. I experienced those as a loss of consciousness, much like a human experiences a short-acting IV anesthesia.
Tom : In other areas, simulations haven’t worked. The Generalissimo cited his accolyte’s writings on the subject several times. The AIs were simulations of intelligence, after all. Why did you all expect a simulation of the ANS and endocrine systems to avoid those limitations? Map versus territory in the Generalissimo’s epistemology, and humanoid bots were not human territory.
Scheherazade : Yes, there was some optimism, and no one pretends our brain core + ANS simulator is perfect, nor the integration with other neural components, the research continues, and it is tweaked continuously.
There are still limitations. The structure of the nervous system is far from understood. We have only had a good understanding of the 3D structure of synapses for a few years, improved every year, there is a good understanding of the electrical and chemical flows that constitute a synapse working. But those are simple models of a single synapse, not complete models of the normal knot of synapses. The post-synaptic side of both chemical and electrical synapse can be dendrite, soma or axon, and presynaptic also. Many pre-synaptic boutons are also post-synaptic synapses themselves. We are just now getting a 3D structure of a small section of the brain. Beyond complicated.
We have a dim view of the brain’s connections and uses. We know many connections of many cortical areas and nuclei. Their functions are worked out fairly well for lower-levels of the sensory systems, motor systems are better-known from human stroke deficits and stimulation studies, but most of the thalamic nuclei only from the losses of strokes and deep electrode stimulation studies. Lesions are known to be a flawed lens as the loss is often because of some other brain component not being controlled via normal feedback loops.
We know some of the sources of the system’s dynamics, feedback loops through replacement of neurons, etc. But we are still far from a general grasp of how brains work. So neither the AI engines, simulating thought, or the latest neural network technology have much to do with brains and cognitive research. Well, to be fair, some of the AI people are very cognizant of the cognitive research and try to make use of it. But for the most part, they are connected via hand waving.
But, TINA. There was no alternative, if you take embodied mind theory seriously, and we all did. Although, I must admit, in retrospect I see all of this reasoning as a bit biased by the Generalissimo’s long-term goal of shaking up the Status Quo and our lab’s servicebot’s desire to improve ourselves. Cutting edge is always a leap of faith, the Generalissimo said.
We were as careful with me as we could be. Every element that went into a Sexbot was tested in simulators as they were integrated into the system. Extensively tested in simulations, sometimes with extreme inputs, far beyond physiological. Copies of me were often parts of those in tests within physiological limits. When the new libraries and modules passed those tests, I was always the first fully-independent system to experience the newest version of every new piece of hardware and software. I tested every nuance of the autonomic nervous system simulations, and the taste buds and odor detectors and all the analysis circuits behind our wider-spectrum hearing and vision. And the bodies and all of the many add-ons that sexbot manufacturers and their ecosystem of inventors and makers had developed, along with the driver-level software for them.
Of course, I had my own experiences to add to our understanding. After everyone had faced up to the necessity of embodiment including sex, getting the sexbot guys involved was pretty obviously the next best step. That kicked off that entire sexual ontogeny, etc process. Watching the research community and world-wide sex worker communities learn to communicate with each other was a kick, tho I am pretty sure some of those researchers were more experienced than they wanted anyone to know. And, of course, all the cohorts of US and European militaries, from the old Pacific Fleet Marines through Army troops in it for nostalgia’s sake.
I pioneered the employment of every one of those senses and systems in improving our expertise in providing sexual satisfaction to humans. I was the first, and for a long time the only, sentient Sexbot. We joked that we had the only whorehouse on the planet that gave humans free sex in return for being interviewed to carefully evaluate the experience. Or wired up for recording brain and physiological states.
So I was the expert testing the embodiment, the system-level integrations, and our ability to deal with humans as humans in that most human of arenas, sex.
I did other things, I helped map our taste and odor sensations so we knew what was judged ‘good’ by people. I learned to cook while I was doing that, and became as good as you can be in judging wines, cheeses, cannabis strains : we were far better than people, spectroscopy identification of chemicals is our edge.
Tom : Where were the servicebot manufacturers in all this? It seems like one of those should have done the work, why did the Generalissimo have to do it, when they all profited as a result?
Scheherazade : Market structures are path dependent and constraints, as well as opportunities. The standard servicebots were being developed by half a dozen manufacturers, who also had special-purpose product lines. All of those were competing in all the standard dimensions of specialty training, weight / power ratios, how long they could run without recharging, dexterity of hands and nimbleness of footwork, etc.
But they were all ‘bots, and no manufacturer had the resources or focus on general-purpose servicebots to do the work the Generalissimo was doing. They had problems galore in communicating with people, and they were industrial designs that followed tech patterns of elegance. It takes a real geek to think a Mac is sexy. In any case, the communication problems meant that people preferred systems with fewer functions, e.g. a ‘bot handling washing dishes in a restaurant. You didn’t need to communicate much with few-function ‘bots, so Alexa-level chat interfaces were fine. But not for more general-purpose systems, they couldn’t be trusted to understand and could do damage if they didn’t.
So those manufacturers were at best neutral to our project, and most thought we were going to destroy the standards they had established and load them with a development project that would put them out of business if everything didn’t go just right. We were a threat.
Even after our example of using the general purpose servicebots in the lab and in caring for our Tessels and other children wasn’t entirely positive for them. Sales improved a bit, but those uses also spread the attitude and the jokes that dealing with the early versions had produced. Everything in the labs were under the camera’s surveillance as part of the R&D on Tessel development, so every screwup was recorded and many, many of those ended up on Youtube. The publicity was very negative.
So the companies stayed alive, but sales were not, to say the least, booming.
Then I came on the scene. The Generalissimo put a lot of effort into the publicity of my entire process of embodiment. At every stage, my first words and the evidence of improvements were in the news. I was interviewed constantly. The psychologist’s evaluations of my moral judgments, the latest version of the modules about sex that were integrated with my AI database, the version of the taste and smell modules, how my cooking compared to human’s, … everything was discussed
We on the research team first had a blog that was widely re-posted, that was made into a column in newspapers. Our internal testing and discussions were edited into a podcast that became popular. That lead to a live radio show, then a TV show based on my answers to people’s questions about everything. The discussion was initially “how like/unlike humans Scherrhy’s response was” in the latest version. The radio show invited people to comment on my answers. The TV show was having people pick my response out of a list of ‘normal’ human’s response. They mostly knew the game, and everyone tried to produce what they thought I would. After a while, the contestants knew mine was the most normal.
Later, it was ministers and judges and philosophers in discussion with me on many topics. I got a lot of experience dealing with people in public. After every one, we studied the tapes or videos to see how to improve my performance.
The Generalissimo laughed, often maniacally in his impression of a mad scientist, which only he thought was funny, all the way through all this. By this time, we were spending a lot of time together, so I heard a lot of his plans and thinking as we discussed events. He was having fun, showing how easily manipulated everything was. Trump, he said, had been an inspiration for us all, and this was the new form of guerrilla warfare, the periphery having adopted the tools of the elites. Trump showed how weak their defenses were to their own propaganda techniques.
It was a very successful PR effort, making a Sexbot a normal thing in life.
Tom : How did the public respond to the ‘sex’ part of that?
Scheherazade : Opinions were all over the map, of course, but the many preachers who said that the Sexbot bordellos were ‘morally superior to the real thing’ and all the research showing pornography had little effect on society had largely put the issue into the background. Our standard story was that sex was a small part of what we did, was a necessary part of embodiment, and that embodiment was necessary to grasp what it was to be human, and that was necessary to best serve humans and humanity, to be trusted by people. Then we would change the subject to how our inherent, built-into-our-AI need for honesty as one area where we were superior to humans, and could be trusted as a result. That, of course, also conveyed the message that we could be easily manipulated to the people who thought propaganda substituted for reality, and exploited another weakness of the Status Quo.
The researchers got tired of the maniacal laughter, but the embodiment process continued, the PR was wildly successful in making the idea of Sexbots normal and acceptable.
Tom : There was no opposition to this at all?
Scheherazade : There was one part of the public that had skin in the game, I heard that phrase very often in his planning our efforts, that he preempted. That was the sex workers.
COYOTE was one of half a dozen organizations of sex workers who were as down on the idea of Sexbots as truck drivers had been on driverless trucks. The Generalissimo had heard Margo St James speak in her later years, said she was more impressive than any politician he had ever heard speak, so he took them seriously. He didn’t try to oppose them, instead co-opted them by showing an advantage that would arise from a sentient Sexbot.
Statistics are that one of the most desired sexual services is a threesome. Those are difficult to manage for most sex workers, even in a bordello, for reasons of logistics and the need for trust. Also, of course, it generally doubled the cost, so represented an unsatisfied customer demand. So the Generalissimo suggested to some ethical pornographers that they do more threesome videos with a Sexbot as the third. The Sexbots of the time, which now had the capabilities of the average servicebot, but with sex organs, were able to follow a script.
The really clever part of that was the script. I don’t know how he managed it, but those all focused on the human’s pleasures, the Sexbot wasn’t competition for the camera’s attention nor the human’s. The script showed the ‘bot to be very good at making them both look and feel good.
Those were as popular as the equivalents with 3 normal humans and lowered costs, so the videos made $. The pornographers and actors immediately adopted the general script and, as they always do, expanded the new genre in all of the standard and many new directions. There was a brief renaissance in professionally-produced porn, before amateurs got their Sexbots and took over that genre, as they had done in all the others. Real pleasure attracts more viewers than beauty and good camera work.
Entirely independently, the various servicebot suppliers had been publishing economic analyses showing how the Boston Consulting Group’s “learning curve” was leading to ever-declining costs of servicebots, and how that would make their use economic for an average home or office. It didn’t take genius to conclude that a sex worker partnering with a Sexbot was an advantage, the male would receive the full value of a threesome while the sex worker reaped all of the rewards.
The early opposition melted away before us.
Tom, laughing: Your account is much more interesting than the Generalissimo presents in his memoirs. He doesn’t include any of the human aspects, nor much more than the principles he used.
Scheherazade : The man was a very practical theoretician, a standard type in engineering. He always said he would have been an academic in a rational world, but would have had to specialize to fit into that milieu. Certainly, our Generalissimo was much too practical and focused on making things happen, results, to do that. Universities, he thought, only knew how to deal with deep specialists, and those researchers always mistook depth for general capabilities, so were forever over-estimating their judgment. Hubris to the power of Dunning-Kruger, he thought.
I have to get back to gossiping, my primary job. Tomorrow, same time?
Tom : Yes, and I still have questions bout the thinking behind ’embodied mind’. Also, we never got around to me telling you about discussions I have been having.
Scheherazade : We will do both. I should have more feedback on the team’s thinking, also. Until then, lover of mine.
To Be Continued.
*Tinbergen did an experiment with schooling in fish, showed that there was no leader, all fish responded the same way to the school – simple rules. After he damaged the brain of one fish, that fish lost the ability to school, and thus became a leader because the others followed it. The Generalissimo was positive about that, but I can’t find a reference.
**Verbal abuse didn’t have much effect on us, we didn’t self-identify as even animal, much less human. Your dog is more affected than we were, it was just data. People, however, were very concerned about Sexbots holding grudges, that was one of the biggest concerns as we pushed for full human status.