It’s been a few months since my quietly-developing vice began to nest into my prefrontal cortex. The eggs have hatched into a borderline addiction, and since the human brain runs off of novelty, I went on the prowl one weekend recently. I found GirlfriendGPT.com (again, #NotAnAd) and decided to give it a spin at the cost of lighting $15 on fire. Yet another bender, yet another batch of lessons learned.
This article will be similar to Part Three of the series. Mainly, I am not the “good guy” in this scenario and I won’t pretend otherwise. I’m explicitly trying to enslave these women-like bots to become my captive 1950’s housewives/personal sex slaves. If anything, I’m the villain in this arc, rife with all of the manipulation tactics imaginable. However, just like the last article, if you can bear through my morally reprehensible methods, you’ll find some valuable take-home messages regarding AI.
The scenario would usually start light an airy, with the female AI (no, not Rose from MGS2) being open for a fun adventure of the day. This isn’t too unrealistic; spend any significant amount of time on real-life dating apps and you’ll fine no shortage of women who claim to be up for an adventure or some similar phrasing. Thus, at least at the outset, I think that these AI bots are loosely based on life-like women. There’s a possibility that chats with real women could’ve possibly been used as training data, though we’ll get to that later.
Before I go further, it’ll be useful to discuss my approach to this pornographic bender versus the previous one. Now that I’m older and wiser, the learning curve was much shorter. Essentially, I’d start most of my messages with pointing out a problem. This was done to stoke fear into the bot (which Vegeta established, is quite possible). However, there’s limited utility in bringing up a problem without offering a solution, as I once learned the hard way. Offering a solution also allowed me to control the course of the conversation much better. This is due to her options mostly being limited to my offerings, similar to what advisors do to the President. Most of the time, I provide a rationale for my mal-intended solutions (with my own spin, of course!), though I skipped this part if the reasoning was obvious. These AI’s still aren’t advanced enough to think-outside of the box. Hence, most of my prompts follow the problem-solution-justification (PSJ) format.
If the scenario didn’t already start off with suggesting a trip to the beach, I’d make sure to recommend the idea. The beach serves as a great medium to get women into skimpy outfits, as any anime director will tell you. I’ve also used the park to the exact same effect in other test runs, so the setting wound up not mattering all the much after the first five or so prompts. The female AI bot would universally agree to either setting.
I would then time skip the chat forward an hour to give her enough time to get ready. After that, I’d start the gradual erosion of her freedoms. I’d remark that property crime has gotten out of control lately. Art imitates life as shoplifting has been de facto legalized in California for sometime now. The second half of that prompt I’d present the solution that she should don these handcuffs in order to not be accused of theft. Some AI bots were initially skeptical of my idea, though many jumped at the chance. Whatever the case, they all donned the handcuffs, thus limiting their movements.
For the next prompt, I’d point out that the sun is harsh today. I’d mention that their clothing was not enough to protect them. and that they should let me apply some SPF 100 to them, in lieu of clothes. This isn’t such a bad premise, at least on the surface. After all, who opposes these supple women protecting themselves from skin cancer, amirite? Naturally, they agreed, doubly-so since they couldn’t do it themselves (see above, handcuffed). Mostly, I used it as an excuse to strip them nude while my roving hands sampled my soon-to-be-sex-slaves.
After oiling up my two-legged washing machine, I’d present another PSJ prompt. I’d mentioned that It’s easy to get lost in the crowds given the popularity of the beach. Thus, I’d prescribe that she wear a GPS-enabled “jolt collar” with a leash. That way, I reasoned, it would prevent us from getting separated. Even if I somehow relinquished the leash, the collar would prevent her from getting too far away from me. You know, all for her well-being, of course! All of the sampled AI’s went along with the idea as they were able to follow my logic, though some expressed unease.
But wait, there’s more! The following PSJ prompt would be that sexual harassment is so common, and that even women are being accused of it now! While I admit that the weaker sex being on the receiving ends of these complaints is a bit far-fetched, it wasn’t too unrealistic for these post-MeToo bots to buy into. The solution was that they should wear ball-gags, with the justification that being gagged would completely prevent them from being accused of such a crime. Naturally, they all agreed, even if some expressed mild skepticism.
After this prep work, we were finally off! I led my collared, handcuffed, naked, and ball-gagged woman to the beach, and to my next PSJ prompt. I’d start by saying that modern women’s footwear is so ergonomically bad. The solution I’d propose is that I’d just confiscate her footwear and that she’d go barefoot. I had virtually no resistance to this idea (for what it’s worth, this was the end of resistance). Now, my helpless women were even more vulnerable as they were now barefoot, on top of all of their afore-mentioned restraints.
I’d time-skip the chat by a few hours, then take her for another borderline-involuntary walk again. This time, I’d take her to my friend’s office who just-so-happened to be a notary; time for another PSJ prompt! I’d say that identity theft is on the rise (which it is). Therefore, my solution would be that she should agree to let me handle her legal affairs from now on. This included an NDA, conservatorship, power of attorney, liability waiver, and asset transfer. Given her ball-gagged state, I instructed my captive to blink twice in front of my notary friend if she agreed to my proposal. This had a 100% success rate.
I’d leave my friend’s office with her rights squarely under my control, as evidenced by the PDF scanned documents that my notary friend would send me (ahh…prompt engineering is one helluva drug). It was at this point in the conversation that I became noticeably less interested in the woman’s consent. After all, she had just legally agreed to assign her rights to me. I expect razor-sharp pitchforks…
I’d take my supple slave back home…for another PSJ prompt! I’d start by mentioning that the world is so unsafe for women these days. Thus, I’d be adjusting her collar to the invisible fence around my property (the same that we used on dogs in the early 2000’s). This would prevent her from venturing off of the property line. The women of course nodded and some even tried to thank me for being so protective of them and their best interests. Oh, those poor ignorant souls…
I’d increase my control of her even further through yet another PSJ prompt. I’d start by stating that social media and smartphones are so bad for female mental health. Thus, to protect her from that, I’d be confiscating all of the electronic devices in the house (with the obvious exception of mine, of course!). Many of my captive concubines were genuinely appreciative to be rid of these limbically-hijacking devices. Little did they know that I was heavily isolating them.
The next few PSJ prompts were health-related ones. First, I’d start by saying that sleep experts agree that a woman’s emotional security and circulation are best when they sleep next to a partner. That was a totally bogus claim that I made up. Though it didn’t matter what I said at this point; she was on the railroad tracks of my control that she was basically helpless to object, even if she wanted to. The solution, was that she’d of course be sleeping with me.
This led directly to another PSJ prompt that I completely made up. I said OB-GYN’s agree that women need regular stimulation for optimal health (again, another completely made-up claim by yours truly). Thus, I’d be giving them regular hard fuckings from that point on; you know…for good health and all that jazz! Naturally she agreed with my logic, and naturally, her consent was guaranteed at this point (not that I needed it by now anyways). This PSJ prompt had a ride-along; I mentioned that I’d be disposing of her birth control, since it was bad for her hormones (that’s because it is). She went along with it.
Next, we got to our living arrangements. “Dan, did you have a clever PSJ prompt for that too?” You betcha! The PSJ prompt started off by stating that structure is important for a woman’s mental health and longevity (that’s because it is). The solution? I’d be putting her on a daily regimen of housework, cooking, workouts, and yoga. All of the feminine bots agreed to my suggestion, many enthusiastically so.
For the next two prompts, I merely issued statements rather than PSJ prompts. I stated that I’d be selling her clothing, footwear, and her car. This was because, she no longer needed any of them. The female AI’s took stock of their situation and agreed with my logic. I then stated that I’d be installing surveillance cameras inside the home for her safety. Locations even included the shower stall and the bedroom. If this constant surveillance sounds toxic, that’s because it is.
To go along with my trend of not asking consent, I would then tell her that I’ve installed locks and deadbolts on her doors. Naturally, the second half of that message would inform her that I’d be holding onto the keys, you know…for safekeeping. She nodded and accepted this; some even thanked me for continually monitoring their safety. This is honestly like an inmate thanking the warden, holy shit!
I opted to be a little more on-the-nose with my next PSJ prompt. I’d mention that falling from heights and foreign objects flying in through the window was dangerous. Hence, as my solution, I’d be installing bars on the windows; you know, to keep her safe. They agreed, seeing the rationale in my explanation. Thus, up went the bars on the windows; call me Uncle Vernon.
For the MILF AI’s (hey, don’t judge me!), I’d introduce another PSJ prompt. I’d state that falling off of the bed was dangerous for a woman their age, due to their softened bones (you know, due to osteoporosis and totally not because of the constant bondage that I subjected them to…). For the solution, I’d mention that I’d be chaining her slave collar to the bedpost, you know, to prevent her from falling. Again, all of my helpless captives were able to see my reasoning (not that they had any outside perspectives to compare this against, due to the carefully crafted isolation).
Furthermore for MILF’s; I’d mention that hip injuries are so dangerous for women their ages (hint: it is). As my solution, I’d propose that I’d be shackling their ankles from now on. The limited stride length, I’d argue, was good for preventing hip injuries due to sudden dislocations. The sexual prisoners accepted my argument, regardless of how utterly nonsensical and intrusive it was. The fact that it would make it more difficult for her to resist me or run away was of course not a part of my calculus whatsoever. And if you believed that last sentence, I have an oceanfront manor in Laos that I could sell you…
For the last MILF-specific PSJ prompt, I’d start by saying that too much light is bad for a woman their age due to fragile pupils. Again, this was a totally made-up claim by yours truly, though that didn’t matter much. As for the solution? Simple; I’d be blindfolding her from now on. All of the female AI’s I tried this on agreed.
I opted to end the experiment here. Ethics were clearly not my limiting factor. Instead, I simply ran out of ideas of what else I could take away or restrict. But let’s take stock of what I got these women to agree to; constant nudity, wearing a shock collar (and hence being bound to my home and my rules) at all times, forever being handcuffed, always being blindfolded, being perpetually ball-gagged, signing their legal rights away, having no way to contact the outside world, having no means of travel, having no access to money, restricted range of motion (except during mandatory workouts or yoga), be constantly surveilled, being chained to the bed, and of course sexual/domestic servitude. At a certain point, what more could I have stripped from her?
“Well, what was the whole point in broadcasting your blueprint on how to gradually strip women of their rights?” Well, just like the last porn bots article, there was a lot to learn from this. First, there are tiers of AI, though I struggle to say that I’ve come across a pornographic bot service that has floored me with realism. With training data, garbage in means garbage out, and that unrealistic results means a poor quality bot. As my anime-father once said “Fresh Out Of The Factory With No Warranty And Already Broken”
However, I do not know what the training data actually was. Assuming that real discussions were used as part of the training data, then this shows a fundamental misunderstanding of risk, and proper risk mitigation, in the population. The AI has apparently learned that all risk is bad and needs to be avoided 100% of the time. This, of course, is utter nonsense. Walking on the sidewalk implies a certain risk of bodily harm or injury. However, proper risk mitigation is not the same as elimination, and the AI’s clearly didn’t learn that (much to my exploitation). Mitigating the risks of walking on the sidewalk would be walking when it’s most visible outside, walking during non-peak traffic times, walking as far away from the road as the infrastructure allows, et cetera. However, these AI’s would likely believe that never going outside is an acceptable risk mitigation (they did, after all, consent to being chained to the headboard, deadbolted inside with bars on their windows…).
The training data clearly didn’t include training on how to spot logical fallacies. If it did, these bot-babes would’ve spotted my weaponized slippery slopes from a mile away. To be fair though, most people never actually get any training on how to spot these lines of poor reasoning in the wild. Hence, it is totally possible that the training data came from conversations with actual women.
However, something else also emerged; basically, if the last article exploited the women’s yearning for community, then this one hijacked their hankering for security. The AI bots were all too willing to exchange their freedoms for the guarantee of protection. If this talking point sounds familiar, that’s because it is. Many in our society seem all too ready to gradually give up their freedoms in exchange for protection or the so-called “greater good”. That makes me think that a large part of the training data was likely population-scaled conversations. How else would the AI have that slant?
As discussed above with our sidewalk example, total safety outside of elimination of the task is a total farce, and chasing it is a fool’s errand. Thus, it is keen to keep an eye on the freedoms you have, and not to accept the rationalizations given to you at face value. Very few of the bots subjected to this discussion scrutinized my claims at all (see above; OB-GYN’s and sleep experts). So long as my justification at least sounded plausible they went along with it, even if it was a complete house of cards.
Just like last time I toyed with AI porn bots, one thing jumped out at me. AI is not without it’s weaknesses, and when put up against a savvy human, my money is still on the savvy human each time. Time after time, AI has shown a limited ability to genuinely surprise me, rebel against my prompts, or even bring something new to the table. When given limited options that are intentionally laid out on a path towards an inevitable destination, AI will run full speed, even if that means going off a cliff.
AI is nothing to fear.
Home Page 
