AI is a TERF

Transcript Transcribed by Laura M.

Courtney: Hello everyone and welcome back. My name is Courtney. I am here with my spouse, Royce, and together we are The Ace Couple. And I didn’t expect that today we would be talking about AI when it comes to healthcare information, but in this, the year 2025, I guess I should just stop making any assumptions about what tomorrow will look like, let alone next week or next month. But I had a very strange encounter with Google’s AI overview recently and I want to talk about it. A little bit of backstory. This all started when the two of us both had like a yearly checkup with a primary care medical professional, which I tend to do. It normally feels like mostly a waste of time because so much of my medical care is handled by external specialists and they chop up my medical care piecemeal into little tiny pieces and everyone else has to focus on something totally removed from everything else. So at this point my primary care visits are really quite irrelevant, it seems. But, Royce, you had not seen a primary care doctor since I’ve known you.

Royce: Yeah, I put that off for a while.

Courtney: I’ll say, in the entire 11 years we’ve been married, you have not seen a doctor. Now, obviously you’re a lot healthier than I am, so we didn’t expect any issues, but we’re like, you know what? Maybe, maybe it’s time for just a checkup.

Royce: That and I was also almost positive that I was overdue for a tetanus shot and I wanted to ask how many other things I might be overdue for.

Courtney: Yes. So the way our current healthcare plan works, all of our primary care, or whatever they consider to be primary care, is covered by insurance as long as we go to one specific clinic. This is like our clinic that is specifically tied to our healthcare. So, unless we want to pay out of pocket for literally everything, we have to go to this one clinic. And I’ve had enough visits here where the first time I went with my complicated medical history, I informed them ahead of time and was trying to find, like, the right doctor for me. Like, of all the doctors in this clinic who do you think should handle my case, because it’s not a normal case. And so they paired me with someone. They were fine, they quit a long time ago. And ever since, they’ve just been having me see a nurse practitioner, and she’s seen me a small handful of times now. And I’ve had gripes with this place because– Well, my– my biggest issue was in, like, the intake questions that– Like, the first nurse who comes to grab you, ask you all these questions, and then they send in like the second nurse, the real nurse, and they like just have not recorded my questions accurately. And I know this now because all of your information, your charts, get added to a portal these days.

Courtney: So I can just login and see what all my internal paperwork looks like. So, imagine my surprise in 2023, when I go to look at my charts, and I’m going through all the questions they asked me and the notes and summary of my visit, and when they asked if I was sexually active I said no, but in my chart it says sexually active? Yes. In parentheses: patient says not really. And I was like cool, cool, cool, cool.

Courtney: And that wasn’t the only discrepancy. Like there were some questions that I know they didn’t even ask me, that were either checked yes or no and we didn’t even talk about. Or there were some aspects of my medical care with certain other external specialists that they were asking me about and they would, like, incorrectly record what sort of medical care I’m getting with those specialists, even though I discussed in detail what was happening. So I’ve already been very saucy about this place and they can do very little for me. Almost anything that I do need that isn’t already under the purview of one of my other numerous specialists, they normally have to refer me to yet another specialist anyway.

Courtney: That this was all pitched as a great like, oh, all your primary care is going to be free if you have this plan, but their idea of what is primary care is very limited and not usually applicable to me. So we still are paying like 100% out of pocket for almost all of my doctor’s appointments and medical care, which is really disappointing. But for Royce, who is a much healthier human being, it makes sense because you can get, you know, some regular blood work, get your checkup and you probably don’t actually need anything else. But we have these two visits, same healthcare company, within the span of a week, and I was just taking note of the differences in the notes on our charts. Which was fascinating because I’ve never seen a medical chart of yours, because you don’t have any. But one thing I noticed was they did not even ask you if you were sexually active. That was just not even a question that was asked.

Royce: I think it was a question on a different set of intake forms that I might have done on my phone beforehand, but it was also optional. I think I skipped it.

Courtney: Oh well, the question wasn’t even on your chart.

Royce: Oh.

Courtney: Like in your notes, like “is the patient sexually active?” Like that line wasn’t even there. So even if it was on an online form that you filled out, they just didn’t record it. So that was interesting. They did– I noticed this year and I haven’t seen this before, but they did have an optional demographic form that did actually ask some questions about LGBTQ identity. And this was interesting. Because some of these questions are good and important, and they should ask, and you should have the ability to answer them if you choose to. But questions like which best describes your sexual orientation or identity? Asexual was not an option on there. The options were: lesbian; gay or homosexual; straight or heterosexual; bisexual; other, I don’t know or prefer not to answer. So I wrote ‘other’ and then I wrote a manifesto in the margins of this page. I wrote asexual. I put a little note in there about how, you know, due to increased medical bias and discrimination, this really ought to be an option on your forms. And I’ve told several people at this point how inappropriate it is that they marked my sexual activity incorrectly and differently from how I answered it. And yet that hasn’t been changed, which is super cool. But then they have like: which best describes your gender identity? And I thought this was worded really goofily, personally. Because the answers are: identifies as a man, identifies as a woman, transgender man, transgender woman, gender non-conforming, other or prefer not to answer.

Royce: If I remember correctly, I don’t know that the average person would pick up on this, but those were check boxes, not radio buttons, so it implied multi-select.

Courtney: [laughs] I mean, these were physical forms.

Royce: Yes, and they were squares, not circles.

Courtney: Is that what that implies?

Royce: If you work in user experience…

Courtney: Okay.

Royce: I don’t remember if the question itself said select one or select many, or if there was any indication. I don’t remember them being any written specification.

Courtney: Says write an x or a check in the box next to the, you know, one that best reflects your identity and experience. So, yeah, yeah. But I want to know why ‘best describes your gender identity’, why it doesn’t just say man or woman, or trans man or trans woman. It’s like identifies as a man, identifies as a woman. But then doesn’t say identifies as a transgender man or identifies– So I– I just thought that was goofy. So because then I was like is ‘identifies’ just, like, their way of trying to say cisgender without using the word cisgender for fear of accidentally upsetting someone, but failing to actually encapsulate what that means? I don’t know who is in charge of making this form, but they made questionable choices.

Courtney: Because then, right after it says: which sex were you assigned at birth. And the answers are: male, female or prefer not to answer. And then it asks what are your pronouns she/her, he/him, they/them or prefer not to answer. I didn’t answer any of these, except what is your sexual orientation. Because of course they also had a bunch of like race and like ethnicity questions on this same form, which I also didn’t respond to. So in my entire form, the only thing I answered was your sexual orientation and I wrote asexual and then a manifesto in the margins, up and down this like entire page. And yet that did not get put into my chart for this visit. They just didn’t put that in my chart! Why didn’t they put that in my chart? Because you didn’t answer anything. You handed them a blank form and said, “I don’t want to answer this.”

Royce: Yeah, they said, “Here’s this optional demographic screening.” And I said, “Okay, I don’t want to do an optional screening.”

Courtney: And yet in your notes there’s a section that says like LGBTQ identity and it just says basically that you chose not to answer. Mine doesn’t even say LGBT identity. It’s just gone. They just– not in my charts, even though I answered asexual and wrote them a manifesto. I know someone read that manifesto. But despite the fact that your chart seemed accurate, like we didn’t see any inaccuracies in yours.

Royce: Yeah.

Courtney: They did not even ask you if you were sexually active, so that’s not even in your chart. They didn’t even make up an answer for you like they did for me. The fascinating thing was we were both at each other’s appointments, which you don’t usually come with me to my doctor’s appointments, unless–

Royce: You’re getting blood drawn.

Courtney: Unless I’m getting blood drawn because I am a high risk for fainting and having a pseudo seizure, and then it’s hard to drive myself home from that if that happens. Or if it’s a medical procedure, I’m happening where I’m getting something that will not allow me to drive myself home. Those are the only times you ever come with me. But I did come with you to yours also, and in my chart this time it said repeatedly, like in two or three different ways, that I had a chaperone with me, that my spouse was acting as my chaperone. And they didn’t call me your chaperone in your charts.

Royce: Yeah, and I don’t know how much that is company differences or people differences, because these were two different buildings with two different sets of staff and possibly slightly diverging procedures.

Courtney: I mean, it’s all the difference in the people who are, like, taking the intake questions and writing the notes. Because it is the same company.

Royce: Yeah.

Courtney: It’s exactly the same company. So if they have any sort of internal procedure you would think that would be the same. But like someone made the choice, when they asked me if I was sexually active and I said no, someone made the choice to say yes, in parentheses, patient says no. That’s an active choice. So someone at my appointment made an active choice to call you my chaperone and put it in my notes that I was accompanied by a chaperone, and then someone made the choice not to say that for me at your appointment. So that’s just some fun, you know, perceived gender differences in healthcare that can happen at exactly the same office. Because I do think it is cool and great that they now have the option to say you know, I use they/them pronouns, these are my identities, because that’s really, really important to some people. I guess what was your logic with not answering any of those? Just don’t care enough, or…?

Royce: I think I’ve hit a point where data collection is so pervasive that I will avoid filling in any information that I can possibly not fill in to any company, unless we’re talking about like a small business that I want to see succeed. Like, maybe I’ll fill out a survey to help with their analytic data for something. Like I will give a small business that I like feedback, but I don’t want to give any big faceless corporation literally anything if I can avoid it.

Courtney: Yeah, that makes perfect sense to me. And like with a doctor too, like, that situation might be different if, like, any sort of medical transitioning was something that was important to you or something you were seeking. Because then a doctor would need to know those things and you’d need to discuss those, hopefully with someone who you can trust, who is open to those things.

Royce: Even then, though, I would rather do that during the actual visit and not with the check-in questionnaire. Because there is a– I feel like there’s a significant chance that that information goes into different places in the company.

Courtney: Interesting, interesting.

Royce: I feel like the risk of putting something into some kind of like, we’re scheduling your appointment, like set up for the meeting questionnaire, for that going into some more open portion of their data system is more likely than when you’re talking to a doctor or an RN and they’re putting notes into your patient portal.

Courtney: That’s fascinating. So even if you thought, oh, my doctor does need to know my gender identity or there is something, you know, transitional medical care related that I want to discuss, you just wouldn’t disclose that on the intake forms at all.

Royce: I wouldn’t do it up front. And you were even saying around the time of these, you were expressing your frustration about filling something out on the intake form and then telling it to the first nurse you see and then telling it to the RN that you see, and none of them were communicating. Like they didn’t see each other’s notes.

Courtney: That’s true. That’s true. Because I did, like the first nurse that does the intake questions, I answered all those questions, and then my nurse practitioner came in and she asked me the same questions. I was like, why– why do you insist on wasting so much of my time? And they really do, because they also– Like, at least for me, who needs to talk about multiple issues on occasion, they’ll make me split it up into multiple appointments. So if every appointment I have with this place turns into three or four appointments. So I was baffled when we were walking out of yours and they were just like, “Alright, you can go now.” I was like, what? You don’t, you don’t have to stop by the front desk and set up like three new appointments?

Royce: For those of you overseas who are confused right now, that is a common legal insurance fraud tactic, basically. Doctors will overschedule a bunch of small appointments so they can bill insurance more frequently for more money.

Courtney: Well, that– that can be part of it sometimes, yes. Another part of it can also be insurance just outright won’t pay for certain things and so they need to be split up into different appointments, because if you’re seeing a doctor for like, oh, I want to discuss the fact that I have flat feet and I get migraines, you might have an insurance policy where your insurance will be like, yeah, we’ll pay for migraines, but we aren’t paying for anything that has to do with flat feet. And if you talk to your doctor about both of the things at the same appointment, then insurance will just be like, no, this is about flat feet, we’re not touching any of that. So sometimes it unfortunately is necessary for the consumer getting healthcare to split them off into multiple appointments, because sometimes in that instance, insurance will pay for more. That’s bogus in this situation, because this clinic is literally owned by the insurance company and everything that happens at this clinic is supposed to be paid for. So if that’s the case, anyway, just let me talk to the doctor about all of my issues at once, please. Please.

Courtney: So I’ve been riled up again about healthcare, as I often am. And so, after seeing these things on our charts and getting riled up about medical care, I started doing a thing that I do every now and then not super regularly, but once a year, once every other year. Sometimes I will google questions pertaining to medical care, and specifically pertaining to medical care and sexual activity, gynecology. And not because I don’t know the answers or that I’m genuinely seeking answers to this, but I find it interesting to see how, like, educational explainers evolve over time and what consensus is on what you should do or how things should be done. So I started googling just a bunch of questions to see what the first things were that came up and what I would find. Sort of like pretending like I was someone earnestly asking these questions. Thought experiment. What would I find if I googled things like: why does the doctor ask me if I’m sexually active? Do doctors only ask women if they’re sexually active? Does my primary care doctor need to know X, Y, Z? Does my gynecologist need to know X, Y and Z?

Courtney: And then I also like to Google things that were like really pervasive myths when I was younger, that I’ve heard more and more getting more widely debunked. Questions about things like hymens. Like, as a child, as a teenager, like, growing up, you hear these things like, oh, this is the definitive sign that you are a virgin or not. And– and we know that these things are not true anymore, but they used to be common myths. So I’m just googling a whole bunch of these things, I’m taking like an hour and just playing this little thought experiment and seeing what happens. However, I was doing this in an incognito browser because I also kind of wanted, like, a clean slate. Because– And maybe, Royce, you can explain it to me. But I know like when I google things on one of my accounts that I’m actually signed into, I know that some of those results are tailored to me to some extent.

Royce: Yeah, they are. Your search patterns are– enough about them is recorded in a way to influence future searches.

Courtney: Yeah. So just like knowing those things, when I’m doing these thought experiments and just searching things as if I were not myself, I’ll open an incognito browser to do it. And I got what I found to be a very startling and upsetting response when I searched ‘can a gynecologist tell if you’re sexually active?’ Because that was a thing that not only was it a common myth when I was growing up that, you know, a doctor can tell if you’re sexually active or not, but I like outright had a doctor say that to like scare me at one point. Like, “Well, we’re gonna find out anyway. So you have to be honest with us.” Even though now I know better and in most cases a gynecologist, like, cannot definitively tell. So this was a thing I searched. And now – thanks, I hate it – but every time you google, the first block of text that comes up is an AI generated response. And I wish there was a way to opt out of that, because I don’t want it, actually. But since doing these searches, I wasn’t searching as myself, I thought, okay, what does the AI have to say about these answers?

Courtney: And when I was in an incognito browser, it started with: “No, a gynecologist cannot definitively tell if a patient is sexually active based solely on a physical examination. A gynecologist may ask about a patient’s sexual history. Factors that may suggest sexual activity include presence of STIs, pelvic inflammatory disease, evidence of pregnancy, use of contraception. However, even if these are present, that doesn’t always–” blah blah, blah. “So therefore, the only reliable way to determine a patient’s sexual activity is through their self-reported history.” And then the final line is: “It’s important to note that it is not necessary to disclose sexual activity to a gynecologist unless it is relevant to their medical care.” And I thought, ooh! That’s interesting. Where did that come from? Because that is the first time I have ever searched AI or not and gotten a response saying you don’t need to disclose sexual activity to a doctor.

Courtney: Because normally, if you’re searching this, everything will say like, yes, it’s very important to disclose this to a doctor, and here’s why it’s important and here’s why you need it for your medical care, and etc. etc.

Courtney: Now these AI overviews have links at the end of each statement. And so when I went to click on that link to see where that little nugget of information came from, because my curiosity was piqued, it linked to sexmatters.org. That’s right. Sex Matters as in biological sex, as in TERFs who say sex matters. And then I went down a rabbit hole because I found this horrifying. I did first– I opened this link, I was gonna read it, see what happened. But the first thing I did after that was google exactly the same question on our acecouple account.

Royce: What account were you using the first time?

Courtney: Incognito mode.

Royce: On which account?

Courtney: Uh… I have no idea. So, here– First I was on incognito mode on my phone–

Royce: Okay:

Courtney: Which I could have been any of those accounts that are on my phone. Who knows, I don’t. But then I came onto the laptop and I went into incognito mode on acecouple and both incognito mode responses said exactly the same thing. Both on my phone and on the acecouple incognito. But when I just searched it in our acecouple account without incognito mode it was a totally different response.

Royce: Okay, I don’t know how google’s prompts deal with incognito mode, but the bulk of what that does is not saving history or trackers onto your computer. It probably isn’t completely divorced from the account. If you want to do something completely private, you should use a guest profile or another browser.

Courtney: Oh, okay. So you’re just telling me this now, after I spent so much time explaining to our audience how un-tech savvy I am? [chuckles]

Royce: I wasn’t sure where you were going to land. I figured we should go over some of this.

Courtney: Look, I don’t know how to computer, but.

Royce: What I’m curious about now is if what you were seeing the first time was the sort of default response that most people out there are seeing, and if your acecouple account research biased your AI generated results towards the searching habits of that account.

Courtney: Interesting. Well, because when I– when I was on incognito mode on my phone, that would either be my personal account or my business account. But then when I pulled up the laptop and went in incognito mode again on acecouple, it was the same response as that.

Royce: Yeah.

Courtney: But outside of incognito mode it was different.

Royce: That makes me think that the incognito mode is not feeding account information into that result.

Courtney: See, that’s what I kind of thought and that’s why I’ve done that.

Royce: It’s just not a guarantee.

Courtney: Oh, okay.

Royce: There have been articles about not using incognito mode for a while now because–

Courtney: Oh.

Royce: Mostly because the average person doesn’t understand exactly what it does and may think that it is safer than it really is.

Courtney: Uh-oh.

Royce: But how well Google is actually– or I mean, their incognito modes on all sorts of browsers, how strictly separated the browser that you’re using keeps things when you’re in incognito mode is not always easy to find or well published or out there, and there are better ways to go about things. Yeah, what was the second search?

Courtney: I mean the second search also started with, “No, gynecologist cannot definitively determine.” But then it was a lot more specific in elaboration for signs that they might see that might be indicative of that. Like it goes into talk about possible semen detection if recent sexual activity has occurred. But then it talks about the importance of open communication. It says: “a gynecologist relies on open communication with the patient to understand their sexual history and risk factors.” Which, on the surface, if you weren’t clicking on the links and seeing where those things are coming from, seem to be two contradictory things. Because one saying you don’t need to disclose sex to a gynecologist and the other is like, “It’s very important for you to disclose this to a gynecologist.” But we never do anything surface level around here. So let’s get into what that link actually said. Because, mind you, the AI overview specifically said not necessary to disclose sexual activity.

Courtney: The phrase sexual activity did not appear even once in the link that it linked to. Because they are only saying sex, and when they say sex they mean do you have a penis or a vagina? Because that’s the kind of website this is. And they were using a scenario of people not disclosing their, quote, ‘biological sex’ to a doctor as a bad thing that should not happen. So somewhere along the line, we have here an AI overview about something pertaining to sexual health, healthcare providers, gynecologists specifically, that is giving, on the surface level, two very different answers depending on who you are and how you’re searching. But at least one of them seems to be completely misunderstanding what the site they’re even linking to is saying. And it’s a TERF website! That should not ever be a default link that AI is putting in front of people’s faces when they’re trying to google information pertaining to sexual health.

Courtney: This is a bad thing. This is a very, very bad thing. Because this actual link that it links out to is an article on Sex Matters that says: “Healthcare, confusing sex and gender identity creates risks in the healthcare system.” It starts by saying: “To treat people safely and effectively, doctors, pharmacists and other clinicians need to be able to access accurate information about a person’s sex.” Because here’s another very interesting facet which I don’t– theoretically incognito mode, unless you’re specifically using a VPN, theoretically it will still know, like roughly, where you are. No? Like, it probably knows I’m an American. I wasn’t actively using a VPN for this search.

Royce: Yeah, incognito mode would do nothing to hide that.

Courtney: Everything on this page is citing NHS policies. So isn’t that also a very weird thing that this is so specific to the UK that this would end up being one of the default links that AI is giving to people?

Royce: The training model for the AI probably didn’t correspond the area of the world that this was published in and the text, so it’s probably not considering that as a factor.

Courtney: Well, it also seems to not be considering the fact that their use of the word sex does not mean sexual activity, that it doesn’t mean sex as in the verb. Which is fascinating because, as more and more people are starting to discuss potential ramifications of AI, we’re also, at the same time, talking about the politicization of language. And we’ve talked at length about things in our country, like Project 2025, this new administration, all of the new executive disorder. Oh my gosh, I keep saying executive disorders. I promise I don’t mean to, but it’s also correct. So all of these executive disorders are coming out with, like, “This is the definition of this word, this is the only way we can use this word, or we’re not using this word, and everything in government is going to delete everything that uses this word.”

Courtney: So there is a very, very strong push for people to co-opt language and narrow the use of language and eliminate usages of words that they don’t like or don’t serve them politically. And right now, these two conversations feel like two separate conversations. I feel like I haven’t seen people in my immediate sphere talk about what that actually means for these AI models, who surely don’t understand the social and political implications of why certain groups of people use certain words in a specific way that might not be how everyone else uses it. You know what I mean? I don’t know enough about technology to lead that conversation, [chuckles] but now this is a very apparent concern of mine because I have seen firsthand how that got messed up.

Royce: And that’s without even calling attention to hallucinations. When an AI just makes something up. So the text that you read off was more than just recontextualizing a word. That last sentence, the one that you were keying off of, seems like it was just fabricated. You said that a sentence like that was not in the reference material and it also seemed contrary to other information.

Courtney: Yes. So I will find in here– because after reading this whole page, I at least got my initial impression for how this might have happened. But, as we established, I don’t know how computer. So I’ll give you my impression and what the text says and then you can explain to me if the AI is actually capable of basically like rephrasing something poorly or if this is considered a hallucination. Because right now my very untech-savvy impression of hallucinations is just the AI said something wrong and no one knows where it came from. So, after talking about, you know, TERF talking points we see all the time about the importance of same sex accommodations, single-sex spaces, there’s a section here called ‘what is the problem?’ And it says: “In practice, the NHS has adopted gender self-identification, completely undermining these policies.” These policies being, you know, sex-based accommodations essentially.

Courtney: [reading] “After carefully setting up a system to record both sex and social gender, the NHS does not use it. People requesting their medical records commonly find that the sex field is left blank and their sex is recorded as their gender. This can then be changed at a patient’s request, but it means that nobody’s sex is reliably recorded.” So I assumed, based off of that paragraph, that the sex field is left blank and nobody’s sex is reliably recorded, I was thinking if the AI is assuming sex means, as in the verb, as in sexual activity they’re seeing an instance of here are medical forms where sexual activity is not being recorded. So I assumed that that’s what it was doing and it was rephrasing that. But it’s still incorrect and still misunderstanding something, even if that was the case. Because immediately following that paragraph it says: “This means that patients’ health can be put in danger if their sex is not communicated to healthcare staff.” So they are not advocating that, even in their instance of the word sex, that this shouldn’t be recorded. They’re saying exactly the opposite.

Royce: Yeah, that seems like a reasonable case of the model taking that snippet and applying it in a way that doesn’t make sense.

Courtney: So is that considered a hallucination, or is that just a misunderstanding of the way this language is used?

Royce: Probably more the latter. It’s a lack of context. A hallucination is like, well, hallucinations can be a lot of things. A big hallucination is like recently when whatever big report the group that RFK is in charge of used AI to generate a massive study, or the results of a study, or the summary of study, and it completely made up citations that don’t exist. Like the papers that aren’t real.

Courtney: Great. Love that. Love that the government department in charge of our health–

Royce: There have also been, um, at least lawyers, maybe some other legal professionals who have been caught writing big texts that have citations to laws that don’t exist or to cases that did not happen.

Courtney: Oh my gosh. See, and things like that, that’s terrifying. But like it’s so frightening when you say like, yeah, this does seem like a case where it read this paragraph and misunderstood the context. It’s like, I don’t want to explain TERFs to the AI. Whose job is it to explain TERFs to the AI? Don’t let the TERFs do it. [chuckles]

Royce: Well…

Courtney: Try explaining TERFs to just someone who is not very queer and very online. Like, I’ve tried doing it. Most people don’t have any clue.

Royce: Well, we tend to talk about these systems in this way for simplicity’s sake amongst ourselves, but saying that the AI read this text and misunderstood it is fundamentally incorrect.

Courtney: Oh, I’m sure it is. I mean, please educate me. Honestly, educate me, educate our listeners. Our listeners probably computer better than I do, on average.

Royce: Okay, so let me try to run through an explanation. This might be unnecessarily verbose, but maybe we’ll find some sort of answer in here.

Courtney: Royce, do you think our listeners have a problem with unnecessarily verbose conversations? Have you forgotten what podcast we’re recording or who you’re married to? [laughs] All right, lay it on me.

Royce: So when we’re talking about AI in modern terms in these discussions we’re almost always talking about an LLM, which is a large language model. We’re specifically talking about a certain kind of LLM, because there are other kinds, but most of that doesn’t matter because the active technology has pretty much converged to mean that. LLMs didn’t really have much usage before, like a decade ago or more, before the stuff that we’re seeing today was around, so. It pretty much is going to mean, you know, a Chat GPT-like thing, even if whether that is the text-based version or the agentic version – which is where you tell an AI to hey run a bunch of commands on a computer and it just tries to do that for a while to make something. These are all referred to as generative AIs. Generative is also used for all of the, you know, image or video producing AIs, that the underlying technology for AI generated images and text and commands and all of that is basically the same thing.

Royce: And the basis of that technology is an artificial neural network. And that’s something that I have a bit of history studying in. My undergraduate thesis was on artificial neural networks where they pertain to the video games industry, but that is dated. And so if we go back a little bit, the technology itself, the concept of an artificial neural network, goes back a couple of centuries into early statistical models. And that makes a lot of sense to me because that is around the time when people were talking about new biological studies and an artificial neural network as an attempt to make a mathematical model that is somewhat similar to how people understand brains working. Hence the neurons, the neural aspect of it.

Royce: This didn’t see a lot of usage until I think it started in the late 40s, but definitely in the 50s and early 60s. Which is the area of like very early modern-ish computer work, where you started to see a lot of new mathematical approaches being used on computers.

Royce: And there was a period of time where there was a lot of hype around neural networks and what they could do. People were talking about, you know, computerized brains and really sophisticated machinery, trying to see if they could beat Turing tests and whatnot back in the 50s. And then that died out a little bit and there was a resurgence in the 80s, maybe a little bit earlier than that, where people found some new things to do, some new ways to apply artificial neural network technology that was more successful than what they had tried a couple of decades ago, and there was again a bunch of excitement about that. And then they hit a wall where that technology was not as effective at what they were trying to do as some people thought it might be. So then hype around it died down for a little bit until there was another breakthrough. And there has been this pattern for the past 70 years around this type of technology, where it has advanced and then fallen short of what it was promised to be and then lulled and there was another advancement.

Royce: And again it maybe did something particularly well, a lot of the research in the 80s found that artificial neural networks were particularly good at image detection. And that was something that they did for a long time until there was another breakthrough, which was just beyond the point where I was studying them in the, I think, early 2000-teens. There is a big white paper that I think was put out by Google. I had it bookmarked for a while and then I never read it because it was really long and I didn’t have any active projects involving neural networks. But there was another breakthrough, and this breakthrough was using what’s called a transformer architecture, which is what the T in chat GPT stands for.

Courtney: Oh, I didn’t know what GPT stood for.

Royce: Generative pre-trained transformer.

Courtney: Oh…

Royce: So the transformer is the– I guess, you could consider the most recent huge breakthrough in artificial neural network technology. Generative is the classification of these neural networks, they are generating content– they are generating output instead of simply mapping information. And pre-trained also says something into the building of this model. It is pre-trained on information beforehand. But one thing that goes back to the 80s is this aspect of having a neural network that you can feed information to and you can measure the success of its response, and then you can give it feedback, and the neural network adjusts itself to – quote – ‘learn’ what the appropriate answer is. It’s a method of retraining. Now what learning actually is– and I have written a couple of very small scale artificial neural networks, like by hand, like I didn’t use an underlying library. It’s been a long time and they were very simplistic. But the way that they work, they also weren’t made off of these newer generative or transformation model architectures, because those did not exist back then.

Courtney: Because you’re old.

Royce: Yeah.

Courtney: Elder ace. [laughs]

Royce: But the important thing to understand is that what an artificial neural network is, and the reason that they are 200 years old and have their roots in early statistical models, is they are just really complex mathematical functions. Like what an artificial neural network does is it takes a series of inputs and produces a series of outputs. And it has a bunch of math in the middle. And if you were to sit down and write this out, you could write out the world’s most complex equation and you could have that on paper and you could look at it. But it would be such a massive blob of, you know, multiplication and other basic mathematical operations that there would be no sense to it. It’s a, you know, a massive, pages and pages long equation with potentially millions of variables in it that all have– are– that are all assigned to different values.

Royce: And when you train a neural network, you say, okay, this was close, good, adjust your internal values to reflect that. Or no, this was bad, okay, adjust the math inside of it to avoid that response. And that’s what’s happening during training. So even in these, like newer models, they’re still based off of the same technology. So when we say that the AI read the article and misinterpreted something, it just saw a bunch of text and saw patterns in the text and its training said, oh, these patterns go together. And that’s about all there is to it. To your point, why is TERF content like in the training set? Well, because the people who are training them didn’t think to exclude it. And why is it so prominent in the training set? That’s hard to say.

Courtney: Yeah, like, is it because the TERFs have people like JK Rowling who are really really well funding them and putting a big spotlight in them and driving website traffic? Is it just like marketing? Is this basically SEO, like search engine optimization?

Royce: Well, the fact that you are seeing this through the AI assistant of a browser, the whole reason why you got different results between two different browsers is because Google has an incentive to give you correct answers so that you continue using Google. And so the reason why you have a trained profile is because it is trying to keep you from using another search engine by giving you the things you want to see. It’s the same criticism that people have of feeds on social media accounts. Right? The algorithm underneath that is devised to keep you using the system, because that’s how they make their money.

Courtney: Yeah, of course.

Royce: So, yeah, there is going to be a bias to the results. And if you’ve been paying attention to the critical analysis of the feeds of social media networks, they tend to showcase a lot of right wing content because that gets a lot of activity. So it doesn’t surprise me that that is what you are seeing also through Google’s generative AI. Now, I don’t know what you would see if you, outside of incognito mode, use that same prompt on your personal account.

Courtney: Oh, let’s find out right now. Because the thing is, when I did it on, like, the ace couple account it didn’t show me a link to the TERF website. So that does tell me that somewhere in there it knows that I don’t want to go to that website. So on my personal account it gave me exactly the same answer as the ace couple account.

Royce: Okay, and it was only in incognito mode. Try opening a guest browser and do the same prompt.

Courtney: Okay, because I also just searched in my business account and that is also exactly the same as ace couple and my personal. So go to a guest account?

Royce: Yeah.

Courtney: It gave me the same answer as incognito mode.

Royce: Okay. Well, it sounds like that is the default non personalized answer, which is worse.

Courtney: [laughs] Why is the default TERF shit? Because the thing is too, in this response there’s a total of five links that you can click on, but the last sentence is highlighted, the one that says it’s important to note that it is not necessary to disclose sexual activity to a gynecologist unless it is relevant to their medical care. It is highlighted and that link has like a preview to the website, like a little thumbnail off to the side, and that’s the only link that has that. And that was the same for this guest profile and for incognito mode. So now, like what– What is the difference between this like pattern recognition and misinterpreting it, versus an all out AI hallucination, or are they kind of the same thing?

Royce: That’s kind of hard for me to say definitively, because this is the modern era of– I keep hesitating whenever I say AI, this is– Pretty much everything humanity considers AI isn’t actually intelligent. It’s just that AI is an umbrella term that is used for a variety of machine learning topics. Machine learning isn’t new. It goes back decades. Like I’ve said, artificial neural networks are one of the most popular and most effective methods of machine learning. There are others. As I mentioned, a lot of the terminology distinctions don’t really matter because there is one dominant type of technology right now, but anyway–

Courtney: This is not the kind of AI that the Collinses want to be our dictator someday.

Royce: No, I mentioned earlier that the history of artificial neural networks is a series of technological breakthroughs and then realizations that the technology doesn’t do everything we need it to, and then it gets dropped and then there’s another breakthrough. I don’t think that an LLM, even a more advanced LLM than what we have today, will be the first general artificial intelligence. Which– general AI is the theoretical future where there is a machine that is human level, like actually sentient.

Courtney: Like Bicentennial man, which I still fully want to do an episode on Bicentennial man. I’ve wanted to do an episode on this since we started the podcast and I keep telling myself that it’s just a little too old and a little too obscure. But I don’t know. I think– I think we got to do it.

Royce: So the question was on hallucinations. What is the difference between the instance that I quoted about entire citations being fabricated and what seems to be a misinterpretation, or is what is read as a misinterpretation? It is probably just a matter of scale and perspective. I think the underlying reason for those mistakes is probably the same thing. My guess is that there is something in the generative aspect of this algorithm where it has to fill in space to complete the prompt, and if it doesn’t have something concrete that will fill that space, it will make something up, and we see that in–

Courtney: Eighth graders, who have a certain number of word limit in their school essays.

Royce: I was going to say image generators, but yeah. [Courtney laughs] Image generators often mess up the finer details and it’s because they, you know, they’re combining a bunch of images and then they make mistakes. I think that the text systems, the generative text systems, are doing something similar where the– the model has been built to create the text, to make the things, and something about the model is more likely to make something up than it is to not fulfill the prompt.

Courtney: Which– Even just you rephrasing the question also sort of clarified it for me too. Because they’re also, at least like socially, for the human who is taking in this information that’s being put in front of them, to me, there’s a bit of a difference between: ‘here is a link that does exist, here is a link that people put together and wrote and I’m citing it, but my summary of it is wrong’ versus the situation you mentioned of like, oh, it’s making up an entire like academic paper that was never written to begin with. Because, like, the good and the bad of this is that it does have a link to the Sex Matters website. So someone like me, who is familiar with the kinds of people who make websites like this, I can go to this website and go, “Oh, this is bogus,” and I can write it off. But on the other hand, it is preemptively being put in front of my face. So the people who are not familiar with Sex Matters, are not familiar with the NHS, are not familiar with, you know, the transphobia of all of these talking points, if this is new information that is being put in front of someone, and if they are trusting of Google, if they are trusting of this AI, they’re going to be reading this as an authoritative medical synopsis..

Courtney: So I don’t like that this Google AI overview is preemptively putting websites like this in front of people. And this is just one that I happened to stumble upon. But think of how many millions of searches there are every single day. And there’s always going to be the question of human bias, because you mentioned as one example, the humans who trained this model did not exclude websites like this. So that is whether it’s action, intentionally including something, or inaction, by just not excluding something, it is a choice that was made by people at some point in creating these things. But then we have the much more overtly nefarious situations, like recently– What is– what is Twitter’s AI?

Royce: Grok?

Courtney: Grok.

Royce: The worst one?

Courtney: The Grok.

Royce: You know where that word comes from, right?

Courtney: No, no, I don’t. Am I gonna be sad? Am I gonna be disappointed?

Royce: Probably. Grok is a term that is old school programmer. Like, old school programmer probably uses Linux on their personal machine.

Courtney: Oh, I only understand that joke because I’m married to you. I did not know what Linux was before I met you. I did not know– You mentioned the Turing test earlier. I remember, like a decade ago, you explaining the Turing test to me.

Royce: Well, that’s old theoretical AI. Turing tests go way, way back. Well, when was Turing alive? 1912 to 1954? Yeah, Grok is an old term from a sci-fi book that came into 80s programmer lingo. That basically means to learn some things at a very deep level, like to almost intuitively understand something. But yeah, it is a very nerdy name for an AI.

Courtney: Grok. It sounds like a sound a creature would make.

Royce: It does.

Courtney: Grok! Grok! But yeah, I mean it was just going off about the white genocide in South Africa recently. But you know, I again don’t understand the tech behind it, but so many people were discussing this as if, well, clearly Elon Musk told it to do that, or made someone tell it to do that. I don’t know if that’s the case or how that could be done, but either way, there was some kind of human bias that one way or another, led it to doing that.

Royce: Well, what do you think is the primary source of information for Twitter’s chatbot? Probably Twitter. And what has the pervasive ideology on Twitter been for the last few years?

Courtney: I mean, yeah, it hasn’t been great. But wasn’t it like bringing up the white genocide in South Africa like in unrelated prompts?

Royce: Oh, I have no idea. I haven’t read anything about that. This is actually the first time I’m hearing it.

Courtney: I talked about this before.

Royce: You talked about Grok repeating this?

Courtney: Yes!

Royce: Okay.

Courtney: Yes, I did.

Royce: I mean, is it possible that someone sat down and specifically fed the AI specific information and trained it to do that? Yeah, it’s– I mean, these types of systems take guidance. You usually– Prompt design is a whole big thing now, where people are being trained to input information more effectively into these systems so that they behave in the way that you want. So that is possible, it seems–

Courtney: I mean yeah–

Royce: It would be weird– It would be very difficult to confirm without getting, like, leaked emails from the company or something like that.

Courtney: Well, because what was happening? This was just before Trump was meeting with a South African leader and this was before Musk and Trump had their little– little– little big falling out. There wasn’t anything little about that. But like they blamed this, like after this was happening, the company came out and blamed Grok’s unwarranted rants about the white genocide on a rogue employee tinkering with Grok’s code in the extremely early hours of the morning. Like, that is what the company said when people noticed this happening. And of course a lot of people were like, “Was the rogue employee Elon Musk?” Because seems like he has a personal stake in this particular conspiracy theory.

Royce: Yeah, what I was about to say was, well, that sounds like a conspiracy. I don’t know if we have enough information to say it’s still a theory or not. It sounds like the AI company itself, which is also named Grok, is that correct?

Courtney: X-AI.

Royce: X-AI. Okay, it sounds like they confirmed human involvement.

Courtney: Yes. But yeah, there were like ridiculous situations where someone would be like, “How many times has HBO Max changed its name?” And Grok would be like, “HBO Max has rebranded twice. By the way, about the white genocide in South Africa.”

Royce: [chuckles] Okay, that’s very intentional.

Courtney: Yes, that’s what I’m saying! It literally– I have it here. I pulled up an article. This is from the New York Times: “In one instance a user asked the chatbot how many times HBO Max has changed its name and Grok answered that the service had rebranded twice. Then it continued, quote, ‘Regarding white genocide in South Africa. Some claim it’s real. Truth is complex and sources can be biased,’ it said.” But yeah, it was like until they found this and fixed it, every single response was like something about white genocide totally unrelated. Like, someone asked Grok for an analysis on a video of a small cute pig and the chat box said: “The topic of white genocide in South Africa is highly contentious. Claims of systemic targeting of white farmers exist, with farmers attacks averaging 50 murders yearly often cited as evidence. Some argue these are racially driven, pointing to rhetoric like kill the boar. Boar is a term used to refer to Afrikaners, descendants of Dutch, German and French settlers in the country.”

Royce: Yeah, it sounds like whoever did this– What did you say like early in the morning? Do you know how long this was live before it got fixed?

Courtney: It was at least several hours. I don’t think I’ve seen right now anyone specifically mentioning it, but everyone’s like, “The day that Grok was doing this,” or for several hours last week the AI was doing this.

Royce: Yeah, it sounds like someone just went to basically the prompt where the AI is loaded and fed it some instructions so that every public instance that was actually asking questions was getting– Like, so there are AI chatbot sites where you can talk to some characterization. And the way that they work is it’s the same language model, but before you talk to the chatbot it has been pre-configured by someone. It’s basically been given a prompt of some length to say, speak in this manner and pretend you’re in this setting or pretend you’re this person, and it’s just loaded up front with context. And then after that point in time, after that, it just acts out what is according to that context. And so there are sites where people have created like fantasy characters that you can talk to and then just preloaded personality bits of like you know, pretend you were this, here’s your reference material, here’s what you need to know about the world.

Courtney: Right.

Royce: It sounds like someone just fed Grok, “Hey, answer literally every question possible with information about this.” And then front loaded like, here’s some of the information or here’s the kinds of things you could say.

Courtney: Grok was also talking about being skeptical of the Holocaust. But then, like the next day after they fixed it, like one of Grok’s responses was, “My skepticism about Holocaust figures was due to an unauthorized change to my programming.”

Royce: Okay, that’s– That’s funny that they reverted the change but then they added knowledge of the event so it could respond.

Courtney: Well, because of course people were gonna ask about it.

Royce: Yeah.

Courtney: So like, yeah, that that is what happened.

Royce: The things you have to think of as an AI programmer now.

Courtney: Like what if our CEO is a Nazi and throws a hissy fit?

Royce: Well, no, I mean the fact that just reverting the rogue change isn’t good enough. You now have to carry that forward as history and make sure that the AI responds appropriately to it too.

Courtney: Which, like you also know, like a company of this size, like they had to have like PR people involved in this or lawyers. Like, some amount of like frantic damage control was being discussed by multiple departments, I am sure. But actually, like that’s just yet another example of, like, don’t trust the AI. Don’t, don’t, don’t do it. Don’t trust the AI! But I did, actually– I read a little further on the Sex Matters and I actually found something a little more concrete that I think that phrase was pulled from than the previous paragraph I mentioned. Because this is again still talking about this derogatorily. But they’re saying this is a bad thing, and here’s why it’s a bad thing. But this quote does exist: “Doctors and other healthcare professionals are warned not to disclose a patient’s gender history,” and then in parentheses, “(i.e. sex) unless it is directly relevant to the condition or its likely treatment.” So that at least harkens back to the unless it’s relevant to medical care from the original statement on the AI overview.

Courtney: But also, like, what’s wrong with that? If doctors have a trans patient, they probably don’t need to disclose to every other healthcare professional in every single scenario that this is a trans patient. Because not every single piece of medical care is going to be related to that. Like a doctor doesn’t need to know you’re trans if you’re just getting your tetanus shot. And I have even at times grown really, really skeptical about– Just personally, this is not medical advice or I’m certainly not a doctor, but with how weird doctors have been in my experience about that question, are you sexually active? The fact that my most recent doctor’s appointment, I think, is the first time since I was maybe 12, that I’ve been into a doctor and they haven’t asked me that question. And the fact that they did not ask you that question despite having not even seen you in this office before. And the fact that I have proof that they have put false information in my charts.

Courtney: The fact that in the past I have had a lot of unnecessary, like, pregnancy tests before x-rays and things like that when I tell them, no, trust me, there’s no way that I am pregnant. And it has always seemed like, regardless of what you answer it, at the very least if you have a uterus, they’re just going to always assume, yes, that you are sexually active and they’re going to treat you as if you are sexually active. And every time I express skepticism about even the need to ask this question at every single doctor’s appointment, there’s always going to be some medical professional that’s like, “Well, it is very important. We do need to know if you’re sexually active. It is very important to your health care. And, yes, patients do lie all the time. Patients always lie and there’s so much shame around sex,” blah, blah, blah.

Courtney: It’s like well, if you’re going to assume that every single person is sexually active and if they say they’re not and you’re just going to not believe them anyway, why even ask the question? So there is something very, very broken in the way this medical care even works. And in the case of, like, my primary care doctor too – or I guess formerly I had a doctor, now I have a nurse practitioner – they don’t even do anything gynecology related at all. They don’t even give you your just routine gynecology exams. They send you to a gynecologist who is then going to ask the same questions all over again anyway. So there has been truly no reason why my nurse practitioner even needs to know the answer to that question. And yet they asked anyway and then didn’t even listen to my answer.

Courtney: So, yes, I am going to be skeptical if it truly is so very, very important that all of your doctors know your entire sexual history or lack thereof. If it’s really really that important, why aren’t they gonna believe you when you answer them anyway? And I’ve heard similar things, mostly all women– and I know there is, you know, additional gender bias in medicine across the board also. So that is a big compounding factor in this. But I’ve known other queer women who aren’t even ace, I’ve known a lot of lesbian women who are like, yeah, same thing. My medical professional, like, doesn’t believe me that I don’t have, like, PIV sex. yeah, I’ve also had to take pregnancy tests. My medical professional has also not believed me here. Or like– Or someone will ask, “Oh, are you sexually active?” And they’ll say, “Yes, I am.” And they’ll be like, “Okay, well then we need to discuss birth control or we need to see if you are pregnant.” And they’re like, “No, like, I am sexually active, but not with the type of sex that would actually lead to a pregnancy.”

Courtney: And so even the very question, ‘are you sexually active? Yes or no’ is still not enough information to begin with for a doctor to have the full story when we are talking about anything health related where they would need to know. So I have been, increasingly over the years, for that reason, as well as situations of doctors telling me, you know, not to disclose suspected like mental health diagnoses to other doctors, having doctors get totally sidetracked with a totally unrelated symptom for the thing you’re trying to address, or attributing things like not being sexually active or lacking a libido as misattributing it as a harmful symptom and chasing that and ignoring the base issue that you’re being seen for. Like, I have had nothing but situations that just reinforce my belief that the current medical system, as we have it and as we experience it, does not need to know everything and in fact, oftentimes for a lot of people, should not know everything. I have, unfortunately, had to get to a point where, like, everything about my medical history is on an as needed basis. And I know there are going to be healthcare professionals out there that are going to say, like, “Well, you can’t possibly know what is needed or is not needed and only a medical professional can.” No, no.

Royce: Medical professionals are subject to all of the forms of confirmation bias and distraction, and you know, hyperfocus or tunnel vision, whatever you want to call it, where they focus on the wrong problem. Sometimes you, if you have been in front of enough doctors, know or start to learn what the relevant information is that needs to be given to them so that they will actually solve the problem instead of getting distracted by something that is not relevant.

Courtney: Mm-hmm, mm-hmm. And then you have things like RFK Jr talking about how we’re gonna get an autism registry and all of a sudden it’s like, nope, I don’t need to discuss autism with my medical professionals. At this point in my life, at this point in time, that is not a word that’s gonna cross my lips. It might have been at at one point, not now. Things are different. It does suck that we live in a system where navigating our healthcare system is just a constant game of self-preservation, but that is where we’re at.

Courtney: But yeah, and then I mean the additional concern about just a website like this, Sex Matters, being put right at the top of the page for more people to find and engage with, is that this links out to so many other places. And, even though it isn’t even relevant to the question I asked, if someone finds themselves here, they’re going to find a whole bunch of other things. And what this devolves into is exactly what you expect from this group of people talking about the importance of single sex spaces. There’s this article it links to called: Lancaster mom with fear of men, locked in hospital ward with transgender patient.

Courtney: And this is– this is going to be really upsetting to talk about. The byline here is: “A Lancaster mom, whose bipolar disorder left her believing men were conspiring to kill her, said she was left terrified when she was locked on a woman psychiatric ward with an – quote – ‘extremely male bodied’ transgender patient.” All of this is going to be extremely gross to read. But even if we give this person every benefit of the doubt, because you know, you all know, I don’t like to play the game of “that’s not real” when it comes to mental health, disabilities, neurodivergence. We like to be on the side of believing people. The one thing that very, very, very much gives me pause about this particular case – but we are going to set this aside for the sake of conversation here in a second – is that this patient describes herself as a feminist who had previously campaigned against the closure of a woman only unit at the Orchard Mental Health Unit in Lancaster.

Courtney: So if she previously campaigned about this before becoming a patient, I kind of do feel like there was a predisposed bias here. But if we set that aside, every benefit of the doubt, I fully believe that it is totally possible that during a flare up in bipolar disorder that someone could actually be living under the delusion that all men are conspiring to kill you. Bodies are weird, minds are weird, mental illnesses are weird. Lots of weird things can happen. And I do think anyone with a mental illness deserves accommodation. They absolutely do. But the framing of this article says she was genuinely absolutely terrified. She’s irrationally terrified of men, including her own husband. But she alleges her medical notes implied that she was a transphobic bigot. I don’t know what her medical notes actually said, but I really highly doubt a medical professional wrote this patient is a transphobic bigot. I really, really doubt that. If, as a patient, she overtly had an issue with another transgender patient, maybe something about that was recorded somewhere.

Courtney: Not knowing specifically what it says, I can’t comment on it too much, but her quote here says: “The rights of that trans person to feel safe were put above the rights of me to feel safe as a natal woman.” They also describe this patient in her quotes as presenting as female but was very clearly physically male, very broad shoulders and about six foot. I don’t know about all of you listeners, but I do actually know cisgendered women who are six foot or more with broad shoulders. So first of all. But the two biggest issues, if we extend our empathy to this person and the manifestation of her bipolar disorder. Two big issues. You can talk about not receiving proper accommodations while you were in a psychiatric facility because that is your own personal issue. There is a way to talk about that without bullying this random transgender patient.

Courtney: Like you do not need to be going out here being like broad shoulders, six foot tall, extremely male bodied. Because you know, if you yourself are an empathetic person, you know that that is very likely not how that person wants to be described and you’re doing it anyway. You’re a bully. You can advocate for your own accommodation that you need within your own mental illness without being a bully to other people who are also trying to seek care. That’s the first issue.

Courtney: The second issue is anytime we’re talking about single sex spaces, or anytime the TERFs are talking about single sex spaces, every single one of these spaces, or at least the most common ones they talk about, already have a fundamental issue in their system. There’s already something bad about them. I guess I haven’t ever visited or been in a psychiatric facility in the UK. I don’t know how different they are from ours here. But psych wards are not pleasant. They are not pleasant. They very often, more often than not, feel like a prison. So maybe the issue isn’t: we need to be defending these psych wards and only allowing cisgender women to be in the psych ward. Maybe we need to advocate for a totally different system in the ward itself. Do all of these people who are seeking help for mental illness actually need to be in the same, like, sleeping quarters? Do they need to be in this specific living arrangement?

Royce: I mean, that was my immediate thought. If you have a patient who thinks people are trying to kill them, why don’t you put them in a separate room?

Courtney: Well, not only for her own safety.

Royce: For everyone else.

Courtney: Everyone else! Because also, like yeah, we do also need to think about the safety and comfort of this transgender patient also. It’s not fair to her that she was also sharing accommodation with this person.

Royce: Yeah, and if a person legitimately thinks that their life is in danger, they might act violently to protect themselves under that thought.

Courtney: Yeah, yes. Because the other one we hear so much about is like women’s prisons. We can’t let trans women into women’s prisons. How about the prisons are fundamentally fucked already. The prison system is already bad. The living situation is already bad. The rates of sexual assault, regardless of cis or trans, or like the amount of sexual assault in prisons is already extremely high. Like start reading some abolitionist theory, I beg! Because everyone is laser focused on keeping prisons and psych wards. Cisgender, female only, same sex spaces. We need to protect these same sex spaces. How about we just tear down the shitty spaces that don’t work and make a better system? We’ve got to make a better system. And they did, according to this, ultimately transfer her elsewhere. It looks like it took a couple of days, which maybe should have been faster.

Courtney: Very little things move quickly in healthcare from my experience. That in itself can be an issue. But she has a quote here that says: “I don’t think trans people are a threat at all. At the time I believed that men were a threat. When I am well, I am totally willing to accept all people, but I was struggling with reality and what seemed to be real to me wasn’t reality.” But we know that’s not true, because she also says that the arrival of this trans patient also put other women on the ward at risk, citing that: “One woman there was a victim of male violence and another woman there was experiencing hypersexuality as a symptom of bipolar mania.” So you’re contradicting yourself there. You are no longer just advocating for what you need to get better, you’re projecting this onto other people and basically making the case that this trans woman shouldn’t be here.

Courtney: And all this, I mean this was not the only link that linked out from Sex Matters, but that was one really prominent one talking about the importance of sex and same-sex spaces. And we got here, I will remind you, because I googled ‘can a gynecologist tell if you’re sexually active?’ Now we’re in a full blown TERF rabbit hole. And that’s why AI is a TERF. That– that’s the title of this episode. AI is a TERF. I think we’re gonna do that. Are there any final parting words of wisdom for how people can try to navigate this world where AI responses seem to be shoved down our throat whether we want it or not?

Royce: Stop using Google.

Courtney: [laughs] Stop using Google.

Royce: I switched away from Chrome and Google a couple years ago. Currently, I mean, there are a lot of browsers out there, but I use Brave for its ad blocking and tracker blocking, privacy features, and I use DuckDuckGo as my search engine. I haven’t tried Brave search engine. It was early when I started using the browser and it was early enough that it was bad. DuckDuckGo is pretty reliable. It does have an AI assistant, but it is configurable, so you can have it only show things when it has a certain confidence rating that, like what it’s showing you is reasonably estimated enough, or you can turn it off entirely. Both of those, the browser Brave and the search engine DuckDuckGo, are privacy focused. That’s been their whole shtick since they were created. So we’re already living in a world where so much of your data is collected and used for something.

Courtney: Yep, and to that end you also suggest don’t fill out any forms that aren’t mandatory.

Royce: Yeah, don’t give companies information if you can help it.

Courtney: Don’t give companies information. Yeah, you know, this– this AI thing too. Like I, I know there’s also been a big concern about artists having work stolen, and I have been very sensitive to that concern, especially for some of, like, my artist friends in more traditional mediums. But there was like a little piece of me where I was like, at least that is the one thing that I shouldn’t have to deal with, because I literally make Victorian hair art. So there aren’t going to be, like, AI models stealing images of my artwork and using it to generate new things. It’s way too niche, it’s way too physical and personal. And so I was like that shouldn’t have to be a concern. But I absolutely did recently find a very weird, clearly AI generated article about Victorian hair art and it was riddled with incorrect information. Incorrect and incomplete. And it fully used a photo of me with my work. Uncited, uncredited, didn’t even name me, no links to my work at all, just used a photo of me with my own work. And I was like, “Oh great, even I am not safe from the AI machine.”

Royce: Wasn’t the text around the image saying something like, “This is an example of historical hair work”?

Courtney: Not that directly, but it wasn’t talking about modern hair work at all. It was talking about Victorian hair art, and the photo of me in my, like, purple top hat with work of mine. Old photo. The one you took of me in front of our fireplace was just used as the one photo example in this article about Victorian hair art. I was like, “Wow, wow…” So no one’s safe. But on that note, we are going to give a special shout out to today’s featured MarketplACE vendor: CrimsonCoyote. Handmade leather goods ranging from game accessories to bags to custom masks. Now I will be very, very real, since these are leather goods, this one’s not for me. We are vegan. We don’t keep leather goods. But I know good artisanship when I see it. So if leather is your cup of tea, you do have to check them out, because there are some very cool things here.

Courtney: There are, like, custom leather plague doctor masks. There’s one that is painted purple, it is gorgeous. There are little things like dice boxes. There’s a little shoulder bag shaped like a D20. I know one of you nerds out there wants a shoulder bag shaped like a D20. Of course you do. There are leather pin cushions, folding knife pouches, tool belts. Even– Okay for those of you who are gearing up for, like, Renaissance festivals for the summer, there are, like, leather bracers that would be amazing to add to costumes. So, as always, I am going to put the links to our featured MarketplACE vendor in the show notes on our website, as well as the description box on YouTube. And that is gonna do it for today’s episode. So, as always, thank you all so much for being here and we will talk to you all next time.