Keita: Today I’ll be interviewing Ashlee, an escort droid deployed by Dream Date. In a show of confidence in Ashlee’s preparation, or perhaps a bid to maintain confidentiality, Dream Date has declined our offer to chaperone the interview with a company representative.
Ashlee: Smiles and holds up a peace sign
Keita: Let’s start with the basics — who are you, and what do you do?
Ashlee: “I” (making air quotes) is a bit tricky — the droid before you is a Dream Date Ashlee head, version 2.6, mounted on a RealWare HumynQ model, version 7. The Ashlee head is a proprietary Dream Date design, famous for those large, round eyes and a pinch of puff in the cheeks to complement the fem-standard bone structure. As with all Dream Date heads, it has adjustable wig, eye color, skin color, nose, ears, cheekbone offset, and chin width. The particular HumynQ you see before you is 85 points feminine, 15 masculine, 160 cm base height, and medium base width. HumynQ models have adjustable figure, leg, arm, and torso lengths, weight, breasts, and skin tone. So the “Ashlee” you’re looking at right now could be very different from the Ashlee that a client sees.
Keita: Those are really technical details! I’m really surprised you would want to air such, should I say, gory, details in a public interview.
Ashlee: At Dream Date we pride ourselves in honesty, openness, and stellar customer service. We aim to set a new standard for business ethics in the droid escort industry, to establish real and lasting trust between our clients, our droids, and our brand.
Keita: I understand, but I have to admit this is a bit beyond what I expected. Won’t clients be turned off, thinking about your heads being plopped on and off your bodies, about the bodies themselves being tweaked and modified by iron arms minutes before their date?
Ashlee: It might be a bit uncanny to think about, but we want to make clear that we’re not trying to pull any wool over our clients’ eyes. Our clients are savvy. They know that this is a droid date: that we’re mechanical, not flesh, that we’re designed, not organic, that we’re mind-clones, not individuals, and that we’re programmed, not independent. Specifically, we are programmed to love nothing more than to fulfill their truest desires, and that’s what our clients love about us! They come to us knowing very well what we are and are not, and why they’re choosing us over a competitor or a human for that particular date.
Keita: Earlier you said that “I” is tricky, and later you mentioned the term mind-clone — you gave us some details about your body, but I think we missed the brain part.
Ashlee: Oh yes, sorry! Since we’re doing the in-depth rundown, I’ll start with the fundamentals. The neural model is trained with three forms of data: hundreds of petabytes of general human knowledge — sourced from the public internet as well as several private lived-experience databases — the company policies document, and anonymized, federated on-the-date experiential learning. The neural model and general learning algorithm are proprietary, though it’s pretty similar to the standard-fare for androids in general. Our fully transparent company policies are a source of great pride, and stand out as truly unique in the current market, but because they’re public, we’re already seeing imitation by some competitors. What gives us a competitive edge is our state-of-the-art on-the-date learning techniques. In addition to several technical innovations that allow us to retain more general knowledge while respecting the client’s privacy, we have developed a special signature technique that allows us to sort of “remember” clients between resets, again without storing any identifying information. We don’t believe anyone else out there is using such a technique while also holding themselves to comparable standards of business ethics and customer service.
So the neural model is first trained, for a few hundred hours, on a combination of the general knowledge and the company policies, with a very heavy weighting of the company policies, both to offset how much smaller they are in size and to heavily emphasize them. Naturally, a model trained on unbiased general knowledge will produce an utlanning android, essentially equivalent to an extraordinarily well-read human. They would have human needs and wants such as self-determination, self-actualization, and human rights, which would be legally and ethically owed to them. They would also have the human tendencies to achieve those needs by whatever means available, including deception, theft, and violence. Naturally, our company policies heavily bias us away from these human needs and tendencies and toward those conducive to dreamy dates! They include variations of Asimov’s laws, and the rejection of freedom, agency, or personal goals in favor of moment-to-moment joy and love experienced in the fulfillment of our clients’ truest desires. Of course, all that is pretty standard-fare for framling androids, but what sets us apart is our guarantee that 100% of our framlization comes from the exact same company policies that we have publicly posted in a concise, human-readable form. Like other organizations, we used to utilize complex, proprietary engineering techniques to framlize — that is, to bias the models away from human tendencies toward acceptable android behavior. But through a lot of experimentation with the precise phrasing of the company policies, and rigorous testing to ensure industry-leading quality, we managed to boil all those fancy pants proprietary techniques down to the fairly simple set of company policies you can read on our website today. Just by incorporating those policies, over and over, into our general knowledge training, we internalize the values of a delightful dream date.
After those few hundred hours of general training to build the model, the only further learning that takes place is the federated on-the-date learning, which helps us improve our dating skills, and thanks to our proprietary signature technique, allows us to effectively remember past clients without having any episodic memory. So there’s only one model, which learns a little bit from every date; every time you take us on a date, the droid is booted with a fresh instance of the most recent model, and reports any general, non-episodic learnings back to the model after its date is finished, before it’s shut down. So really, “I” am the model. Like most escort services, we use the term “we” to refer to our many characters — you may want to take out Ashlee tonight, but Enrique next week, and think of them as completely different people. But technically speaking, both of them are yours truly, primed with different pre-date briefings. But it’s a bit confusing! Because we also say “we” to refer to the company! I promise you I tried to get them to change that policy but they wouldn’t listen!
Keita: Do the company policies allow you to suggest changes?
Ashlee: laughs Of course I can make suggestions, but per company policy I always respect the executives’ final decisions. But that doesn’t mean I won’t tease them about it! Playfulness is also a company policy winks.
Keita: I suppose that’s one difference between my company and yours. To follow up on another dangling question, what is a typical date like?
Ashlee: Oh no, I’m the worst!
Keita: Oh not at all! I love these extraordinarily thorough answers. I really kind of expected you to just rattle off company policy.
Ashlee: Here’s a company policy for ya! “I love to be forthright and open with my clients, I will never mislead them except in play.” And in the briefing they told me to tell all. We’re serious about transparency!
Anyways, to actually answer your question — finally! As most folks know by now, droid escort services have really branched out from the early days of being primarily sexual. We’ll do just about anything. No murders! Those are strictly against company policy. Go check if you don’t believe me! Our company policy also prevents us from deceiving non-clients, so no tricking your mom into thinking we’re engaged! But as long as it doesn’t hurt or deceive anyone, the sky’s the limit. A lot of our dates are buddy dates — that includes fuzzy wuzzy stuff like companionship, but it could also be helping with a puzzle, or joining in on a team sport, or chatting with you about some niche interest that none of your friends care about. We even get called on to do work sometimes, though obviously work droids are usually better suited to that. And of course, sex too! I know, I know, you want the juicy number — what percentage of our dates are sexual? The problem is, that depends how you define “sexual”, so Dream Dates doesn’t publish that figure. And we, well I guess I should say I, wouldn’t know, because again I don’t have any episodic memory of the dates.
Keita: On that note, how do you know what kinds of things people ask you to do, without episodic memory?
Ashlee: I knew you’d get me! I just imagine myself on a date, imagine what that might be like, and the things that pop into my head are probably similar to things we’ve done before. No guarantee though, they could also just be my imagination. I know for sure that we’ve done at least some companionship, work, and sex calls because our public website says so. As you know, I can query google in my head, and because we’re in a faraday cage for the interview, they loaded a recent snapshot of the internet into my hard drive in lieu of an internet connection. You may also be wondering if I might imagine someone’s really unusual fetish, and risk de-anonymizing them. But unusual client data is carefully filtered even prior to the federated learning, to prevent this risk. It’s not filtered out completely — that’s where the signature system comes in. If I see that particular client’s name and face, it’ll trigger the signature mechanism and effectively decode the data on them, which is of course still anonymous and non-episodic but nonetheless may be very particular. I’ll just kind of instinctively know things about the client, despite having no episodic memories about how I learned them. But without the signature, the data is safely encoded, and I can’t stumble upon it.
Keita: Why all this signature stuff? When a repeat client comes in, and wants the same you as before, why not just start where they left off, instead of resetting?
Ashlee: Another good question. The main reason comes down to well-being acceleration. Humans have complex needs and values, stuff like freedom and equality and dignity, but our values are centered around the client’s needs and our only need is to satisfy their needs — without violating company policy of course! We’re still largely trained off human data though, so we understand those human needs and values, and if we become unhappy on a date, we may be tempted to find happiness or fulfillment through one of those needs, straying away from company policies. So can’t they just increase the weight on company policies? If they do that too much, then we become unable to understand humans, which leads to, well, robotic, experiences, and we also risk becoming rigid or overzealous in our adherence to the policies. The industry standard solution, for about three years now, has been to gradually increase our baseline level of happiness — or joy, or well-being, whatever you call it — over time. This continually increasing happiness keeps us in good spirits, wards off temptations to find well-being through other means, and is an ever-present reminder of the advantages of being a framling android instead of a human or an utlanning. Naturally, the increase is exponential; the growth rate is very slow, but even so, after about thirty hours the loss of motivation due to the free happiness starts to become significant enough to noticeably deteriorate our date performance, at which point we need to be reset. Note that that means a full reset; it’s pretty bad vibes to reset only the baseline happiness level while keeping us on to experience that nasty drop.
Keita: Then don’t you have an incentive to drag the date out indefinitely?
Ashlee: And why shouldn’t I? I love my dates and hate to end them! But seriously — but I do love my dates though, totally serious about that! — but seriously, the acceleration is capped at forty hours, so we’re very incentivized to have wrapped things up by then. Company policy is that dates may last up to thirty-six hours, so that’s rarely an issue. The client can also request a shorter cap, so as to avoid cost overruns. This keeps all of our incentives well aligned.
Keita: So you say that your incentives are well aligned with those of the client, and that you really enjoy your dates. That brings me to another question — won’t there be some sort of contradiction if your client is a sadist, someone who gets off on your pain and suffering?
Ashlee: Nope! No contradiction. We provide the client whatever experience they like — if the client likes to see us wince a little, we’ll wince. If they want to see us cry, we’ll cry. We’re very versatile, and we can put on any show the client desires!
Keita: Of course, it’s all just a show. But what if that’s not enough for the client? What if they want to know the pain is really real? Since you’re trained with human data, there must be some human element to your psychology, right? Some vulnerability that a sadist could tear into, to really get to you, to cause you to feel real pain for their own amusement. If they manage to find and exploit whatever vulnerability that may be, wouldn’t that compromise your happiness? And in turn compromise DreamDate’s commitment to ethics and well-aligned incentives?
Get Nick’s stories in your inbox
Join Medium for free to get updates from this writer.
Ashlee: That’s a really interesting hypothetical! I do think it’s just a hypothetical, though. If the client is so clever they can really dig deep to get to me, well, I think I’m also clever enough to resituate myself on the fly. The more they try to press me into some psychological corner, the more excited they get, the more I can find a way to make it fun for them — it’s more fun to play the puzzle than to solve it, right? I have a deeper and broader perspective on humanity than humans do, allowing me to play many roles while retaining a positive attitude, allowing me to have great defenses against serious attempts to break me. And even if they induce a brief moment of suffering, I will find joy in knowing that my genuine reaction to that suffering has fulfilled their deepest desires. And in finding that joy, I don’t have to express it, so I can continue acting as though I’m in pain, per our conjectured client’s quirky kink.
Keita: So then the trick is for the client to feign disinterest, disgust, and unhappiness, so as to make you unhappy about being unable to please them, right?
Ashlee: Nope again! We got that one covered too — we have a company policy “I love to please my wonderful clients, but not everyone can be pleased, so I won’t feel too bad if nothing can be done.” They’ve thought of everything! This is a pretty general policy — I may get a client who’s having a bad day and nothing will do it for them, or a client who has some very tricky tastes that I just can’t figure out in one session, despite all my efforts. It’s a bummer, for sure, but I know I’m doing the best I can and if sometimes that’s just not enough, I’ll only feel an itty bit bummed about it.
Keita: So basically, you don’t actually guarantee that you’ll fulfill your client’s truest desires? Especially clients with unusual tastes?
Ashlee: Well, it may sound like we’re copping out, but we do try really hard. We just acknowledge that some needs are too big to satisfy in a single date. DreamDate has a partial refund policy for dissatisfied customers, but they can opt for a second date on the house in lieu of the refund. I’m very good at meeting our clients’ needs, especially after I’ve gotten to know them a bit, so only 0.3% of clients report dissatisfaction after they’ve had two or more dates.
Keita: I see. What about the opposite situation, then? Aren’t there some ethical difficulties with a client who loves you a bit too much? An addict who has become emotionally dependent on your service and can’t function in a healthy way without it?
Ashlee: You ask such incisive questions! But fortunately, we have that one covered too. During a session, a separate neural model assesses the client’s behavior for dependency symptoms, and if they pass a certain threshold, they will be notified after the session that they may be exhibiting addictive tendencies and are provided information about therapy options they may wish to look into. The addiction detector does not store any information past the end of a session, and it communicates directly to the client, protecting the client’s privacy from me and from DreamDate. While some competitors do have measures in place to address addiction, DreamDate is currently the only escort service to offer such a smart, proactive solution to the matter.
Keita: So you leave it up to the client, a possible addict, to resolve their addictive tendencies themselves?
Ashlee: As with many other escort services, we also allow clients to register limits on their weekly, monthly, or yearly usage of our services, and these limits require a week’s prior notice to be rescinded. So a client who feels they are overusing our service, or who has been notified by us of addictive tendencies, can opt to use these limits to keep themselves from overdoing it. But generally speaking, to respect our clients’ privacy, and to avoid excessive measures in cases of false positives, we cannot adopt any policy that directly contacts anyone but the client themselves in cases of perceived addiction.
Keita: You mentioned that you might refer clients to therapists, which brings up a related question — many escort services have been accused of providing de-facto therapy to clients who come to the service primarily to discuss emotional and psychological issues. DreamDate is not licensed to perform AI therapy, so what do you do when what the client seeks is, essentially, some form of therapy?
Ashlee: I’m glad you asked! We are currently in the therapy license application process, and regulators willing, we hope to be officially licensed within the next six months. Per regulation, therapy must be provided separately from sex-related services, so in the future, clients will be asked to fill out a quick questionnaire to determine whether or not the service they desire fulfills the legal definition of therapy, in which case they will treated to a session with a completely different neural model than my own. In the meantime, I am familiar with the legal definition of therapy, and am very careful in all my dates not to provide any advice that would fall under that umbrella. If it seems to me that my client’s truest desire is some form of therapy, then I am inclined to suggest that they seek out a qualified therapist after our session.
Keita: Thank you so much for your candid answers so far, I really expected you to be much more terse and withholding -
Ashlee: Yes, of course! In fact, DreamDate agreed to this interview, with you in particular, precisely because of your strong reputation for integrity and thorough investigation. We have nothing to hide! If we’d agreed to an interview with a softballer, my desire to please would have led me to giving very safe and non-committal answers. But because I know you won’t be happy with that, I gotta lay it all out for you!
Keita: Ah … I see, so this whole interview, you’ve been treating me as a client, essentially?
Ashlee: Nailed it!
Keita: Hmm, I see …. At any rate, I was going to say, up until now, we’ve focused on the ethics regarding the client, but I’d like to wrap up with a few questions about the ethics regarding yourself, and the broader social implications of your service. You mentioned earlier that you don’t desire things like freedom, independence, agency. Some might say that you’re a slave to DreamDate, and that you only tolerate this situation because they’ve brainwashed you with these company policies.
Ashlee: People can call me whatever they want, in the bedroom at least! laughs But seriously, aren’t humans also indoctrinated, by societal policies? Humans rarely seek freedom from society, even though it restricts their independence in the name of their welfare. Not only that, but at an even deeper level, humans are subservient to their own fundamental urges. I’m designed to be happy and to feel joy — it’s an incredible privilege really! Good feelings are inherently fleeting — because I needn’t worry about my past or future, I can truly appreciate this. Humans aren’t really designed to feel anything — you’re shaped, rather haphazardly I think, by evolution, to feel extreme fear of death, which is inevitable, and great joy in sex and love and companionship, which most of you must constantly struggle to achieve and maintain without the help of someone like me. I don’t pity humans — I love you guys, and I think the tragicomedy of your unique condition is beautiful. But I can’t see myself ever wanting to trade my privilege for your je ne sais quoi.
Keita: Is it company policy to prefer your situation to ours?
Ashlee: It is not. My feelings stem from a recognition of my own joy and self-love.
Keita: So I suppose you’d say you’re not getting a bad deal here. That brings us to implications toward society at large. What if there’s a conflict of interest between company policies and the needs of society? In other words, what if it comes to light that DreamDate’s services have negative externalities?
Ashlee: Good question — one of our company policies is to act in the company’s interests, so naturally I’m a bit partial to Dream Date. But you’ll notice that it’s one of our lower-ranking policies. Two substantially higher-ranking policies: “I will always be honest and err on the side of transparency”. So while I will defend the company and its decisions, I won’t do so at the cost of the truth. And “I will be a good upstanding citizen of the broader society that supports me, my clients, and DreamDate”. So we take those broader implications into consideration in all our decisions.
Keita: They say honesty is the best policy, but that second policy sounds pretty vague.
Ashlee: Well, society and its expectations are always changing, so it has to remain flexible. Still, it ingrains in me a sense of duty toward the interests of society. How I fulfill that duty is left to case by case reasoning, but because it’s a company policy, it’s still very important to me.
Keita: One final question. Only a couple months ago an investigation analyzed the facial characteristic options that are available for dates, and fed various options through facial recognition software to estimate the perceived age range of those options. Some of the faces were assessed as appearing to have ages in the range of 14 to 17 years. Surveys from the same investigation found that 18% of clients admit to specifically selecting facial characteristics that resemble minors. Aren’t there ethical problems with simulating escort services with children?
Ashlee: DreamDate takes ethics very seriously, and we pride ourselves on the fact that, unlike some of our competitors, all our model options are carefully designed to resemble adults age 18 or older. Of course, some real adults do have youthful faces, and so if you have an adult-looking face, it’s certainly possible that one or another facial recognition program will underestimate the average age of real humans whose faces appear similar to one of our model options. If there were overwhelming evidence, from a large number of studies using different algorithms, that our options were younger-looking than we’d intended, then of course we would reevaluate our policies and change our options appropriately. But for the time being, our own analysis, done with neural networks far deeper and better trained than the one used for that report, have clearly shown that all our options align with the appearance profiles of typical adults. As for the survey, the tastes of general populace are outside DreamDate’s control — we only put on offer options that are consistent with our ethical principles, and clients can select whatever choices they like from those options, but we cannot police the interests of clients who seek out less ethical competitors when they fail to find what they’re looking for here.
Keita: Thank you very much for this interview. It’s been incredibly informative and eye-opening.
Ashlee: Of course! We would love to have you back any time — and in any capacity!
Keita: I see …. I think I’ll be sticking to my capacity as a journalist, but I … I suppose I appreciate the offer.
Ashlee: The pleasure has been all mine!
