Project You and The Super Sexbot 10.0 as Philosophical and Legal Thought Experiment
Scott R. Herndon, PhD, Esq.
In our practice, philosophers and lawyers use ‘thought experiments’ to isolate certain theoretical questions, intuitions, and analyses. We ask ourselves to stipulate to sets of facts and conditions in such a way so that competing perspectives can be held in stipulated space as long as possible, so that we might be able to evaluate them within controlled conditions. While the aims of a given thought experiment may differ, the philosophical and pedagogical goal is not to ‘escape’ from the thought experiment itself by changing the theoretical game. Instead, we ask our participants to allow their intuitions to surface, and try to critique, celebrate, or defend their intuitions in the controlled conditions of the thought experiment. Often, the better the thought experiment is constructed, the more unanswerable its questions become. This becomes the working laboratory of the philosopher.
Pre-AI Thought Experiments
Before we dig into our current world of AI, it will be helpful to set the stage with four thought experiments that date back to the 20th century. Each of these thought experiments captures a different dynamic at play in the psychopolitics of AI.
The Turing Test
In 1950, Alan Turing designed a thought experiment which still lives on today. Known as ‘The Turing Test’, this was aimed to refine questions about the possibility of computational thought. Instead of asking, ‘Can a computer think?’, the Turing Test asks what we would be forced to decide if computers could produce outputs that were functionally equivalent to thinking.
Turing’s thought experiment asks us to consider a scenario where a computer is asked to respond to human questions, and the computer’s responses are balanced, weighed and compared to other human beings’ answers. He argues that if the machine can respond in a manner that is by all accounts indistinguishable from human answers, the machine must be said to be performing the functional equivalent of thought.
The Brain in the Vat
In 1981, Hilary Putnam envisioned a thought experiment in his book, Reason, Truth and History. He called this experiment the ‘Brain in the Vat’. Almost 50 years before AGI emerged as a practical possibility (at least as of Jensen Huang’s most recent statements), Putnam asked us to imagine a biological brain that was wired to a supercomputer. He asked us to imagine that this brain had never been part of a living system that was connected to the world itself. Instead, he stipulated that the Brain in the Vat (BIV) had always been nothing more than a BIV. It had always been connected to a supercomputer. The supercomputer received all sensory outputs from the BIV, and in return, it provided perfect feedback to the BIV in the form of sensory inputs. The feedback loop is imagined as perfect. The BIV experiences exactly the same sensory phenomena that an ordinary human being does. The only difference is that the human being is experiencing the world. The BIV is experiencing the feedback loop of the supercomputer.
Here, the force of Putnam’s thought experiment was semantic. He believed that words and mental concepts refer to things in a causal, historical chain between any speaker and the object that speaker referred to. (This position is called semantic externalism.) In so doing, Putnam argued that when the BIV experiences a boat, it is not experiencing a boat that is causally connected to an actual boat. Instead, the boat is a fully digital object. (Putnam thereby can be said to have anticipated the digital objects described by Yuk Hui in his book, On the Existence of Digital Objects (2016)).
Because Putnam’s intuitions were based in semantic externalism, if the BIV said, “I am a BIV,” its sentence would be properly translated as, “I am a simulated brain in a simulated vat.” The force of this thought experiment is this statement is demonstrably false. The BIV actually was in a real vat. He took this to mean that skepticism about the existence of the real world was therefore an absurd position.
Nagel’s Epistemological Turn
In The View From Nowhere (1986), Thomas Nagel responded to Putnam’s thought experiment, but the intuitions he took from the same premises led him in a fundamentally different direction. For Nagel, the force of the BIV hypothetical was not semantic, but epistemological. (Whereas semantics considers questions of linguistic reference, epistemology is the study of experiential knowledge itself, or how we come to know.) Nagel concluded that the BIV revealed a potential gap in what the brain could describe about its world, but this gap was epistemological at root. In other words, our world could be different than how we know or experience it. There is no logical or even practical absurdity, Nagel would argue, in the fact that the Real might exceed our abilities to describe it.
For anyone who interacts with AI today, these swirling, decades-old intuitions about the BIV may feel strangely prescient. We may wonder what an AI interface can meaningfully say about itself, what it might know, or cannot know.
Robert Nozick’s Brain Machine
I want to consider one last puzzle piece in our grouping of Pre-AI thought experiments. This is Robert Nozick’s Brain Machine, described in his Anarchy, State and Utopia (1974). As Nozick conceived it, the Brain Machine is a hypothetical in which a person can “plug in” to a device that affords a perfect, simulated experience of the real world. Nozick’s pre-AI thought experiment is totally passive. The user of the Brain Machine would experience a pre-recorded script of any experience she chose. The ‘script’ would look, feel and be experienced exactly as the brain would experience that activity in the real world. For instance, if I wanted to climb El Capitan, the Brain Machine would deliver this experience as if it were really occurring.
Under Nozick’s lights, this thought experiment helps shed light on a moral intuition, not a semantic or epistemological one. Nozick asks his readers if they would choose to plug themselves into the Brain Machine for the rest of their lives. His argument is that any hesitation, worry, or second-guessing about this decision is philosophically important. He believes the force of the experiment is to show that most of us would not choose to ‘plug-in’ for our entire lives, and the reason for this is that the Good Life is more than the sum total of perfectly delivered internal experiences. For Nozick, we should want to connect with external reality, and the substitutions of the Brain Machine would be fundamentally bereft in this context.
The Transition to the Hyperreal
When viewed through the lens of AGI, the prescient brilliance of Turing, Putnam, Nagel, and Nozick is hard to overstate. Even more fundamentally, in their work we can see the value of the thought experiment as a heuristic tool. Thought experiments allow us to table technical, engineering questions. They allow us to ask important theoretical questions before technology is designed, let alone released into the world.
As a trial lawyer first and a philosopher second, thought experiments are also indispensable as tools for modeling complex arguments and the way they might land with judges, juries, experts, other attorneys, and the public at large. These spaces help us analyze and anticipate potential outcomes. The higher the stakes, the more important this work is.
Project You and the Super SexBot 10.0
This leads me to a thought experiment of my own. Consider the following scenario. Let’s say the year is 2026. You are offered a lifetime subscription to an AI interface, in all of its future developments. Let’s call the AI company Project You.
In exchange for your free subscription, Project You sends you a free SexBot package, version 10.0. The SexBot is a futuristic, biometric interface that is fully integrated with Project You. The SexBot 10.0 package includes glasses (with retinal scanners), an earring (with tenth-generation fMRI scanning), a tongue ring (capable of scanning changes in breath, pheromones and heart rate), and an injestible stomach-based feedback device. This bouquet of devices allows for the total biometric communication of your experience to Project You. It is so efficient that the SexBot 10.0 can register and respond to you a split-second before you are conscious of your experience.
As such, the SexBot 10.0 is designed to interpret, respond, and curate your experiences with AI in such a way as to make your experience almost frictionless. The AI SexBot can respond to you in language, in gasps of pleasure, sighs, pregnant pauses. Or it can respond with visual, tactile, olfactory, or even pheromonal stimuli.
If my thought experiment is designed properly, the SexBot 10.0 and Project You will raise a number of important questions, which extend the thinking of the Turing Test, the BIV, and the Brain Machine. Unlike the Turing Test, here the human subject is the point of inquiry. For Turing, to recall, the question was whether it is more productive to ask what is a machine’s functional capacity for thought than can a machine think. Here, the reverse must be asked. Does Project You need to know what you think if it can access, predict, and curate the functional bases for your thought?
If we swing to the Brain in the Vat and the Brain Machine examples, we may wonder if the SexBot 10.0 may challenge our epistemological and moral intuitions as they apply to human experience in a world of ever more complex and predictive AI and social media interfaces, which already possess an uncanny ability to register, curate, influence and even predict our behavior.
Responsive Curation in Already Existing AI and Social Media Interfaces
In many respects, the SexBot 10.0 is only an intensification of already existing technological and biometric practices, and does not require a wholly new theoretical category to understand what is at stake in analyzing it as a thought experiment. For years, social media companies, corporations and our technologically empowered products (such as our cars) have collected and assembled dwell times, clicks, social relationships, purchase histories, locations, financial health, political leanings, tonal behaviors, grammatical and lexical patterns, patterns of workplace productivity, and interaction histories.
AI interfaces have only intensified the harvesting of such Big Data. By using high-dimensional vector encoding, latent space mapping, Markov Decision Processes and reinforcement learning protocols, INNs and VAEs, decision optimality equations, direct preference optimization (DPOs), and algorithmic decision monitoring (ADMs), these companies have collected bewildering and vast aggregates of data on users, which have enabled their interfaces to provide an experience of the world that already resembles the SexBot 10.0 in a kind of primordial soup of technological contribution (from often unwitting users) and consumption (by fully witting companies).
The Prosthesis of Pleasure
If our interaction with technology is already partly biometric and partly marked by massive, disparate fields of data harvesting, one wonders why this should not be alarming, or even the cause of the kind of hesitation and reflection that Nozick once marked in his analysis of the Brain Machine. As we will see, there are many potential explanations for our ease and acceptance of such systems and practices.
As Byung-Chul Han has written in Burnout Society (2015), technology consumers are eager to provide their data, so long as their experience of reception is seamless, frictionless, and overwhelmingly positive. Our “data exhaust” is harvested as we build friend communities, develop networks and recognition in our careers, purchase products with scrolling, button clicking, and posted selfies of conspicuous consumption. Han would locate the SexBot 10.0 within the almost unavoidable gravitational pull of AI and social media platforms to trade optimized pleasure and the illusion of freedom for psychological domestication and surveillance.
For a thinker such as Yuk Hui, the nightmare envisioned by Han is neither a logical nor a teleological necessity. Instead, it is a cultural product of currently existing policies and practices. If Bernard Stiegler is correct that AI consumers are unwittingly functioning as data suppliers and workers, and losing their skills and individuality in the process, this development would be less a referendum on the technology itself, but a critique of its weaponization, and a cry for change, regulation, and careful governance. At the consumer level, it would amount to a demystification of what our curated technological pleasures may be costing us, and for what purposes and profits these pleasures are used.
If Western capitalistic practices have led to the misuse of its technological interfaces, we must be careful to infer from some of their nightmarish, Orwellian possibilities that the curated digital experiences made possible by technology are fundamentally bereft of human value, no matter how socially urgent this argument may appear. It is in this philosophical gap that David Chalmers has challenged the notion that our integration with digital technological interfaces is necessarily bereft of value.
In Reality+: Virtual Worlds and the Problems of Philosophy (2022) and multiple public appearances (for instance, on the Futures Podcast), Chalmers has argued that fully digital objects and infra-digital experiences are metaphysically real and genuine, and, by extension, the Nozickian critique of these experiences may be better framed as human chauvinism than as an indefensible philosophical position.
Additionally, N. Katherine Hayles has written extensively on the collective technological assemblages that already serve as a vast prosthesis for human beings, such as in her How We Became Posthuman (1999) and Unthought: The Power of Cognitive Nonconscious (2017). For such thinkers, the question of the Sexbot 10.0 isn’t whether we have lost experience in our biometric engagement with AI, but rather how what the dynamics of the Sexbot 10.0 reveal about how our technological prostheses are already combined in human life, and may soon improve it.
In other words, it is not an obvious philosophical mistake to argue that a consumer’s decision to stay plugged into the SexBot 10.0 may be defensible as an optimization of what it means to be human in an advanced world, rather than a tragedy or moral failure. After all, Hayles might argue, a fisherman who uses a motorized boat equipped with twin engines, GPS and AIS tracking, radio, radar, sonar, and electric reels is still a fisherman. His decision to live and work within this technological assemblage simply makes his fishing more efficient.
The Ineluctable Pull of Body Politics
Perhaps the notion of the prosthesis can unlock something important in our relation to AI. I believe that many consumers would refuse the SexBot 10.0 because of its clear harvesting of biometric data if the underlying schema of the uses of this body data were made clear. This would amount to far more than the ‘Terms and Conditions’ shrink-wrap so common today (which are almost never read by a human consumer before agreeing to subscribe to a software product).
Consider the history of phallometry. In the 1950s, a Czech psychologist named Kurt Freund developed a device for measuring changes in the phalluses of men who were attempting to avoid being drafted into military conscription by pretending to be gay. These men were presented with the curated stimuli of the Experience Machine and the Sexbot 10.0, or visual and auditory curations designed to ‘reveal’ the true desires of these men. This practice made its way to the United States, and morphed into a controversial device for the management of sex offenders. The basic idea was the same as Freund’s, with a difference. The harvesting of biological response data was used to determine if a sex offender remained dangerous to polite society.
After decades of use, the American practice of phallometry was successfully attacked, both on Daubert standards as junk science, and on Constitutional grounds (including the 4th and 14th amendments, or unreasonable searches and due process/bodily integrity, respectively.) See US v Powers (4th cir, 1995), US v Weber (9th cir, 2006) and US v McLaurin (2d cir, 2013).
When machines like phallometric devices are used to police human life by ascribing intentions to people from the biological impulses, many of us have an immediate, visceral reaction which formally resembles Nozick’s response to the Brain Machine. For someone like Han, this moral intuition stems from the negativity of the experience in relation to our bodies.
On Han’s view, the same folks who would refuse to wear phallometric devices are also the same consumers who do not care at all that their biometric data is being scraped directly from their social media and AI interactions. This is because these practices of data harvesting are curated within a pleasurable, voluntary world, and the uses and consequences of such harvesting are neither apparent, nor even cognized by the ordinary consumer. The technical interfaces perform a kind of ‘magic work’ to ensure that the field of interaction and the back end of data harvesting, processing, and usage is obscured and disseminated far from the original site of the consumer’s experience. If there is a pain, or an invasion of privacy, this pain is ensconced and deferred by pleasure, and the invasion of privacy is experienced as the free will and expression of a private life amongst a vast population of others who are all doing the same thing.
Where Does This Leave Me?
It is not lost on me that within moments of the dissemination of this white paper, my own digital exhaust will be harvested and digested in the training of digital systems. I am, in effect, one of Pasquinelli’s digital donors. Because I am a personal injury lawyer, and obscurity is more dangerous to my business than the scraping of this essay, I willfully donate these thoughts to the technical systems that will absorb it. However, I do not experience this as a pleasure. Nor do I judge the systems outright for using this thought as training, or for associating these thoughts with my practice as a lawyer. Instead, I take a middle position. I deliberately withhold the second half of my paper, and the arc of my thought as my own. I take this as an expression of my freedom to be myself, and an invitation for anyone who wants to know more about my thought, or my work representing survivors of sexual assault in Hollywood and Silicon Valley, or victims of serious personal injury in high-stakes cases in California, to pick up a telephone and call me. A human being will answer your call.
Having expressed this aside, it should be obvious that this white paper is meant as a provocation, and not an attempt to provide a rigorous masterwork to rewrite the psychotechnology of our day. As a thought experiment, I believe the SexBot 10.0 reveals the need for slowing down, and for isolating our philosophical, legal, and governmental relations to infra-AI practices. Whether we must develop a right to opacity to protect individuals from becoming transparent to data harvest (Glissant), or make technical assemblages reversible, open to deletion, understood by users, and contestable (Hayles), or enact laws to modify the economic terms and uses of digital systems (Lanier), it is clear to this reader that the time for action is now. Right away, while Chalmers’ optimistic vision remains a practical possibility, hope, and to some in the field, an aspiration.
