Your Brain on AI: Interview with MIT Media Lab’s Nataliya Kosmyna

[00:00:00] People have problems with psychiatry, and I think that a lot of 'em are valid, especially given our past. But that is a bonkers intervention that would never be approved if the story of what's happening was accurately described in that way versus the way it's being marketed. How to be patient. Hello guys.

Welcome back to How to Be Patient. Preston, how are you doing today? Good. I, I am patiently waiting for this, um, paper to be discussed. Your brain on ai. Yes. We have a, we have a, we have a lovely guest here. We have a very exciting guest with us. This is what I have unilaterally decided to say is our sort of advanced journal club that we're gonna have more on the podcast where we bring people on who are actually doing the research and get to ask them questions that may be intelligent, maybe dumb questions.

Uh, and you guys get to hear right from the scientists doing the work, uh, and we get to talk about, and you can decide if they were dumb or intelligent questions. There are no dumb questions. Sorry I didn't [00:01:00] introduce myself, but there are no dumb questions. Yes, especially, sorry we're like studying before you were introducing me, but especially, especially in this age of ai, I just gonna tell you without training any studies, there are no dumb questions.

Please do ask questions. Everyone. There more wrong questions. Yes. Well, we have with us Dr. Natalya Kosmyna, um, who is a research scientist at the MIT Media Lab and also local to Boston where I am, as our listeners know, we are having her on as one of the lead authors for a study that was very, very much talked about to, uh, different levels maybe of accuracy over the summer and this fall.

Um, IE your brain on chat, GPT, which is a pre-print. Um, and we are so excited to have you on with us today. I am so very excited to be here also with you guys as experts. I think, um, I got an email from you several months ago we [00:02:00] scheduled and I was so excited. I honestly, this is, I, I know I don't want to write, you know, says that all the others are like podcasts or anything else, was not excited.

I really enjoyed them. But this one I'm very excited about because you can also bring in so much to the conversation. Of course. Uh, and complement it. And it's also important, uh, I think to just have more empirical data, more conversation and how your brain does not stop working, but there are some issues, challenges, right.

And we definitely can dive in all of that. So I am so, so excited. And as I said, well, deep Excite was on our side too. I think Margaret sent me like a Dunking GIF emoji when we found out. Send that you were gonna come on the podcast. I did sent that. I did not send that. You sent that to me. But yes, I was very, I was like, Preston, I'm sending a cold call email.

I'm excited. Let's see if we can have this guest on. No, I was like, yeah, I think I obviously, like, I remember I saw the email and I'm like, oh yeah, psychiatry. That's, that's super exciting. And I think [00:03:00] that was even before there was this, like psychotic papers were actually out. No, they were already out. But like I think there's so much more, it actually, you know, as it got out in, it feels like 20 years between these three months of, of emails and exchanges and I saw that, you know, your background specific like, you know, child psychiatry and uh, your person was like in a neuro psychiatry was like, oh my God, that be so, so good actually to chat with folks about that and get all different perspectives on the topic.

Yeah. Yeah. I mean, from my end clinically, Preston, I'm curious for you, but. In the child world, I think parents are, have been asking obviously about screen use and pediatrics. Child psych has given recommendations, but screen use means so many things. And now with AI included in that, I'm very selfishly excited to have you on and to be learning about your paper because I'm already asked about it from parents and like when we go to school visits.

Preston, I don't know what your experience is clinically. So I, I mean, in a lot of my neuro psych clinics we're dealing with patients with like cognitive disabilities and dementia syndromes and, and [00:04:00] these are patients that on the other end, it, it might not be overuse, but actually being taken advantage of by these systems.

So these are all populations that are really vulnerable to scams. And now the ability to like generate AI images or create realistic scams has just like gone through the roof. So, so people are like concerned and will ask like, how do I protect my family member from, you know, spending all this money if like someone uses a really effective AI scam on them and they can't like differentiate between what they're looking at.

Absolutely. And I guess I can even jump in here. So like, our paper and research we're gonna be talking about specifically was actually not done on all the adults, but I just cannot, you know, resist jumping in what you're saying about older the adults. The paper actually about cognitive of fluid, I believe from this year, from 2025.

They actually talk about cognitive OFin specifically, and they do discuss overall a set of different studies that were done by different labs actually on cognitive of Flo and when it's good and when it might not be good. And they specifically separate the use for, [00:05:00] uh, young adults and older adults. And they specifically point out that it's done.

Uh, all research is done in pre LLM, so pre large language models era. So that's very important. But interestingly, they do mention that. So older adults in a lot of studies that they are citing seem to benefit from cognitive floating. So this is for those who listen and like, what is that? It's when you use, let's say a phone and you would put notes about your shopping, right?

What need to get from target, et cetera. So this would be an example of, of floating, right? Mm-hmm. Uh, older adults seem to benefit, uh, from it. However, there's definitely these issues and challenges with trust. Mm-hmm. Over trust action. Not even over reliance, but over trust that hey, you know, if the system tells you these are the items you added to the list, you would actually will have harder time to recognize that, oh, there may be some changes in those items and that's an issue.

And then second [00:06:00] point, and this is someone told me, so I would need to check actually if this is an actual paper or maybe just a reported news. Uh, so it's not actually a study or maybe just anecdotal evidence as we call it. Uh, in scientific world, but apparently there were some already scams apparently, uh, using this AI features that could be easily enabled for the whole use of like internet browser and all the adults, you know, starting clicking on those because they're believing that well, ever since it's been shown is effectively true.

So, yeah. Sorry to like, interject on that topic, but No, please do interject. I do, I do think it's, it's so much that you have children development and developing and literally growing brain, right? So, so susceptible to everything around, and then you have older adults that definitely are losing some of those, you know?

Mm-hmm. Cognitive abilities. They can benefit from a lot of this technology. We actually building. But then on the other side, you always have this cost, so to [00:07:00] say, to the technology, right? And, uh, to the, any tools actually that we are building and designing as humanity. It's very interesting, I think, very important to understand what is the cost and is it's something we can live with.

And actually understanding what is the cost? Like, am I, am I ready to pay, pay? Or maybe it's not something that is worth actually paying for. Yeah, I I think a lot of people, um, are stuck on this question of, is that, is AI good or is it bad for us? And I think we have to think about it the way we look at other technologies.

You know, when the hammer was invented, you could use it to build great structures and houses, but it was also used to bash in people's skull on the battlefield. And it's up to us to decide how we wanna use this hammer. Not the inherent Phil philosophical question, whether or not it's good or evil. Yeah.

And I think it's important to, to your point also separate very clearly, AI is already already celebrated, I believe last, no, this August actually 55 years as a [00:08:00] term. Right. And the term itself actually got, you know, uh, got to be born on one of the scientific conferences because two scientists didn't like each other.

So it's super fun situation. There not, they didn't like each other one apparently believe that the other one was talking too much. That's literally the reason they didn't want to invite him. And literally because of that invented a new term. I mean, that's so, so funny, uh, in itself as a story, right?

Scientists and he know the research, discover, there he goes. Yeah. The ego egos deserve a whole set of, whole set of podcasts and books that a lot have written. But I think it's important to understand the AI is already around us. We have been using like. You, you know, in medical field, obviously have been using it actively for imaging and for other, since I, in research also using it.

A lot of cases for imaging, brain imaging, in my case even is just one example. Uh, you know, plane piloting, you know, planning all of these things. You obviously recommendations what you're gonna buy, what you're gonna order Next, recommendation systems is that [00:09:00] all, all of that is AI powered. So LMS are just one of the byproducts of ai and I think that's important to understand that some of the products might be potentially more harmful or less harmful in some of the use cases, but it, it is really important to understand the use cases and have empirical data that can actually show, at least shed some light on those use cases so we can have more informed decisions, right?

Because ultimately it is a decision of everyone why and how you would want to use it. Uh, if you obviously can consent. This is also the assumption we can talk about, especially in the use case of. I guess, but, um, in some cases we just need to understand that LMS are not just, you know, only product of ai and there's definitely a lot of opportunities in a lot of areas.

Yeah. Yeah. I mean, so many things We could start off from there, but I think maybe to ground us, if we wanna start first by describing for people who have not read the study, which is a pre-print, I believe. I don't know if it's got a [00:10:00] date. Okay. Um, if you wanted to, in your words, describe, we could try to explain it back to you, but I, I think it's probably better if you ex explain it to us, because you've done it so many times than I'm the expert on it.

Then we can do teach back. Yeah, yeah. No, yes. You imagine I can't explain it to you. Yes, I can. So, in a nutshell, right? Actually, yeah. This is a preprint, it's available online, so I do encourage folks. It's a long paper. It's 1 44 show notes. Yeah, it's 1 44 pages. Pleasant, uh, actually aed, so a total of 206. But everything's available online.

All the figures, all the details we have, and suggestion how to read it if you actually want to read it. But you understandably maybe do not have time to read the whole scene. You actually have a very specific guide inside of the paper where we explain a total, which you should pay attention, and it's actually totals to like 13 ish pages.

So if you just want to read, you can cut it to a short time, like, you know, spend on it. Like, which is like a more classic paper, I [00:11:00] would say. Mm-hmm. Uh, and we have like FAQ and like videos that you can watch if you are more of a visual person or auditor person. Uh, but in a nutshell, we did a very specific use case of essay writing.

So we decided to look at, you know, what happens. Uh, we know when participants, uh, users use an LLM to do a very specific task of sri. And we of course would love, love, love to do all possible ages, all possible tasks. That, unfortunately is impossible. I mean, I guess it would be possible. Instead of trillion of dollars spent on data centers, we would actually have researchers, right?

Just to, we maybe don't need a trillion. So we can benefit, definitely could have cured cancer to that at that point with that finding. But what do I know, right? Keeping us on track. Right? So because we're limited in funding, we needed to select a very specific, you know, participant group. And anyway, research wise, we obviously didn't know specifically what we're gonna find.

Uh, so we [00:12:00] decided to limit ourselves to a specific task, which is essay writing, because surprise, surprise, we had surrounded by students on the campus. Yes. And in Boston area. We have a lot of students. So we thought, okay, at least these students right now, they know the task of essay writing. They are supposed to have learned at the school.

They actually are, are required to submit an essay for that as an part of the entry application for most of institution universities they would be considering. So we decided, okay, at least these kids still know how to write an essay, so let's do the tasks. I invited them to come in person to the lab because we equipped all of them with EEG or electron CY fall.

So it's a non-invasive measurement of brain activity based on electrical activity produced by a brain. So we did that. Because of that, we needed them to be in person with us as a lab. And we divided our participants who are students [00:13:00] from MIT Wealthy Harvard Tufts, uh, in three groups. One group was allowed to use an LLM.

So in this case it's the charge GPT, hence the name your Brain gt. It's not because we decided to single out a specific product, it's just because the tool we used is a chair, GPT. Mm-hmm. So the, uh, instructions was very straightforward. You can use chair GPT as you see please. So no limitation on using it.

Whatever you want to put as prompt or prompts, you just cannot use anything else. So you cannot like go on Google and type, like, for example, check a reference or maybe check something, latest news updates or anything you just needed to only stick to charge GBT. And that's it. Second group was allowed to use internet search.

We limited them to Google as, the only difference from maybe a classic Google search that you're experiencing right now is that we used og, so old school Google where you, so not the AI version of it. Yeah. [00:14:00] Not AI version. That was, which also sucks. Yeah. Yeah. But we wanted really an old, old school version of Google, so mm-hmm.

The ones that doesn't have that summarization. For those of you who do not know how it looks like, actually I don't remember. There was that version of Google who was, that didn't back in the day, have that back in the day. Back in the day, just a year ago, which is like 30 years ago, there was that version of Google, so they could go on any website, whatever they wanted, wherever they wanted, except on any website that hosted, uh, chatbots.

So with these LLM chatbots, but they could, could go anywhere else. And then the third group, as you might already guess where that is going. They actually were not allowed to use. Anything whatsoever. Just the old noggin only, only their own brain, literally no tools to fall back on term, you need to use your brain only.

And so, uh, for their essay writing, right, of course it's a tricky topic because, well, it is a high cognitive task actually. So we [00:15:00] gave a time limit of 20 minutes for the task and then we, to make it a bit simpler, we took SAT type of essay topics. So you would not need necessarily prior knowledge to write an essay.

So specifically an example, all, all of them are in the paper, but, uh, some examples include what is true happiness should I think, before I speak. So you do not necessarily need any knowledge to be able to write this and uh, to make sense even. Right, because what if you come right to, to a study and you're like, oh, I hate this topic.

I never thought about this. Like you, you would feel uncomfortable. It would be hard for you, uncomfortable. Most likely you might even experience some form of stress. Actually, I didn't want that. So for each of the sessions, and there were three main sessions and one additional one. The first session we actually offered each participant the same choice of three topics.

So each participant could have actually picked up a topic they lean towards the [00:16:00] most for whatever reasons. And we did ask them for the reasons later. So what we did for this recession, so we measured the brain activity of participants. We looked at outputs at the essays themselves, of course. Mm-hmm. And then we asked, uh, English teachers, because we are not English, you know, experts.

So we asked English teachers who are trained post study. So they were not with us on the protocol. We hired them afterwards and we asked them to rank and read those essays and get back to us and explains the rating. And then we asked in, in a lot of other studies, Hey, um, some questions to our participants.

Could you maybe, you know, give us a quote from your essay. Mm-hmm. Can you explain why you picked up this as a topic and things like that. And so this was, and those question sessions are part of the representation, which I know you'll get to of like Yeah. Ownership or recall of the essay Yes. That they wrote with or without tools besides their own brain.

Yeah. So, and I will get back to some specific examples of those questions, [00:17:00] but again, they're all in the paper because we'll not be covering and looking into all of them through this conversation. Throughout this conversation. Yeah. But if someone wants, they're all explained in the paper. So these were the three main sessions and then there was the fourth session, additional session, a bit apart for several reasons.

So I'm gonna get to it afterwards. I'm gonna first explain again the first three and the findings for the first three, and then I'm gonna get to the fourth one and explain you. You'll understand why it's a better part and I talk about it separately. So for the first recessions, once the study was actually done, right, we found that there were significant differences in all of these measurements that we actually undertook.

So first of all, for their brain activity, right, we looked into what is called, uh, dynamic direct transfer function, or DDTF, or functional brain connectivity for those of us who listen to us and in lay person terms, so they don't necessarily know what it is. Um, spoil alert, it's not brain stops working.

Brain went [00:18:00] on vacation or something like that because we saw these expressions actually Yep. In some of the media. And you can see in FAQ examples of some of those where I say, please do not use them because we did not use them. And that's not how brain works. But, so to explain laborers in terms what we actually measured.

So, um, the function actually looks at. Which brain regions talk to each other. So let's say we have frontal prefrontal area, right? And maybe it talks to back of the head, which is occipital visual cortex. And so directionality is very, very important. Is this prefrontal talking to occipital or occipital talking to prefrontal.

It's also goes about top bottom approach within touch upon those a bit later. So basically, which brain regions talk to which brain regions and how much token is happening. So flow of the data right now, for example, while I'm explaining the study, I'm giving a [00:19:00] lot of doing a lot of talking. So it's a lot of flow of data coming from me to you.

When you're gonna start maybe asking questions or telling something, uh, from your experiences, it would be the opposite. So that's exactly what we measured. No broken connections in the brain or half of the brain going somewhere. No brain rotted. No brain works. Yeah, no. So you know, immediately it's the only thing you need to take out.

Brain worked in all three conditions, but it did show different functional brain connectivity. So what we actually saw is, uh, more significant, you know, uh, widespread, very widespread networks. And that is actually again, very, we have very high resolution, uh, figures on the website. So you can look at those work in the brain only group.

And surprise, it's actually not surprising because think about it, just from the perspective of doing the task, I'm giving you limited time to do a high cognitive task. Complicate. It's a highly complicated task when you actually try to break [00:20:00] it into pieces. You need to come up without falling on any tools, with some structure, right?

Because we, we are, at least as of now, 2025 and of 2025, we have learned all in school how to structure the essay, how to structure our thoughts, how to have examples to have all of these ideas lined up. We know that we need to have some references, contra points, et cetera, conclude. So we know how this should look like and sound like, but you have limited time to do that.

So your brain is like on fire. So to say you are like cognitive Lord. You're gonna be like, oh my gosh, what was this thing? What was that saying? You're gonna be all over the place. Right. And it is normal. It is absolutely normal. It makes me think of that meme where, um, Patrick starts thinking in SpongeBob and, and smoke starts coming out of his ears.

And that's, that's just about it. And Squidward iss like, what's that? Smells, Patrick. Trying to think again. I I'm pretty sure that's, that's probably what my brain does, writing an essay. So, so that's, I I'm picturing [00:21:00] the gears turning and, and whole brain working. You said whole, whole brain worked in all the conditions.

Just like, again, important. I love your metaphor. Yes, that's exactly that. But you just really hope. Now no one will write an article about this podcast saying Brain went on fire, but it kind of did a bit. So, but yeah, everyone stayed with their brain intact. Nothing was on farther there. However, when you go to the search engine group, to the Google group.

You see differences, right? These widespread networks became less widespread. So you definitely see a lot of happening in the, again, as I mentioned back of the head, this is called occipital visual cortex. And interestingly, and we do, uh, cite a lot, a lot of these papers in our papers, there were multiple research papers prior to ours actually looking at, as we already maybe touched upon.

And I will touch back again on. Understanding the use of tools. Right. So in our papers though, being to best of [00:22:00] my knowledge still, the first one that considered measuring and looking into brain activity is actually not the first one. Uh, it's the first one looking into LLMs, but not the first one. In looking into tools use, there are actually quite a few papers on Google use specifically, and one very cool paper is called Your Brain on Google.

Surprise, surprise, might, might ring bell there, 2009. It's a cool paper. They, uh, used FMRI as an imaging technique, so very different. Um, so for those who may be not aware, right? EG would look on the surface, which is very interesting and has very different resolution from a MI that would go into structure.

So that's basically the, just a simple terms for those who want to follow us. And, and to get a little bit more granular, um, I'm more used to reading studies of FMRI, um, yeah, for brain activity and, and I guess I found myself like. Just maybe not understanding fully how we use EEG to estimate brain activity, so, mm-hmm.

[00:23:00] I guess feasibly it makes a lot of sense in this scenario. You can't have someone in an FMRI while they're writing an essay task, but do you prefer one or the other for the accuracy it gives you for, for measuring cognitive load? That's excellent example. So I would say here that, and we actually say black and white in our papers, that FMI would be actually the perfect tool to use.

The, some of disadvantages of FMI effectively are, well, disadvantages are, it is a complicated set up, right? So for example, letting a person doing something and ephemera is possible is definitely feasible, but it's not always very easy. For example, person writing would be very, very hard, right? Obviously, let me looking at something is it's definitely much more feasible.

So I would say in a short terms, that definitely ephemera has its advantages, right? If you especially want to see like real and the structural changes. But, uh, EEG in this case is a great example because functional brain connectivity can be derived from EEG, and more importantly, [00:24:00] multiple studies have been doing this.

Or actually for all different use cases we have handwriting, typewriting example, some of the essay writing, actual examples, just using non LLM tools. So it is good. But back to your point, right? Cognitive load is actually pretty well studied using both EE, G and Ns interestingly. So for those who listen in this third modality, uh, so, uh, and this is, um, functional neuro, uh, infrared spectroscopy for those who actually wondering what it is.

And it's all, they all have different temporal and special resolution. Uh, so that's where the differences also come from. So e, EG for those who are very curious, is very powerful. If you want. Fast response because you might want to really capture events on the second level, like granularity, like these changes.

It can be more subtle. Right. And some of it's like a needle on a record player, you know? Yeah. It's [00:25:00] capturing every little single movement of your brain. Yeah. But also nothing can compete with that also. Yeah. But also the challenge of course, if, I mean we can definitely dive into that further is of course location, right?

Because where you would put the electrodes in this case, how many do you have? This actually feds fits back to the question how you would calculate your functional brain connectivity, like DDTF, right? This aspect, like how many electrodes you would need, a, things like that. So we used, um, um, you know, satin with like 32 channels, there are 60, so 32 64, 1 28 as the most common for that.

So you would not be really able to pull this off with more like. Variable use cases, but again, it has a lot of potential. If you just were to look into cognitive load, like let's say you, you would want to live in the study because we actually wanted to know what's exactly happening more or less. But if you were just to say, Hey, I just want to measure cognitive load, you actually maybe would have not be, need to replicate our setup exactly to zt.

You might have even, uh, ran away with something actually [00:26:00] still EEG orne, but a simpler setup. So cognitive load is very well explored and started to use an e EEG and veneers. Uh, but again, it would actually take some more time for ERs to show you the difference. Um, it's blood oxygenation for those hallin, so it obviously is not immediate, so you would need to wait a bit of time.

So let's say it's important for the applications where you might want this immediate response, for example, to what we call closed loop. As an example, uh, for those again, who may be, are wondering what, what am I talking about? EG is very commonly used in the use case of non-invasive brain computer applications.

Uh, where you would want, for example, a control a wheelchair for a patient who has significant health challenges like paralysis. So you would, obviously, you cannot wait 10 seconds for the system to work, and obviously they cannot do the syn fmri. So you would do this using an EEG. And there are [00:27:00] again, disadvantages and advantages of each single modality.

There's much more, by the way. There is not only just these three, there is so, so much more. You have optogenetics, you have pet, you have mag. That becomes actually much more affordable now there's a lot of cool stuff, but a lot of papers broke a look and use EEG for functional brain connectivity. A lot of papers do look, uh, and use ephe.

I and I recently saw a paper claiming, not functional connectivity, but some other metrics using, so I'm still reading this paper. It's a fresh one. And they also say black and white and limitations. We would need to run EMI study to really go in full depth on what is happening. So ephemera definitely, I would say it's like crown jewels there.

And then you would actually want to see how you can scale some of those. Mm-hmm. So FMRI is the gold standard and these are just. Giving us indications that there may be some, some more to explore here. Yeah, exactly. And they are good enough in a lot of, even, again, there are some of them, like [00:28:00] 1 28, 64, they're actually good enough like lab settings.

They still require a lot of prep, a lot of setup. So they're definitely, um, you know, still like heavy duty horses. But definitely, mm-hmm. Definitely it's ba I would say based on your research question and how comfortable it was a protocol. And Al also, I mean I obviously joked about this, but car joked have said the truth.

It would also depend on the funding. Of course FMRI sessions are very, very expensive. And in our case, scheduling three sessions even for our limited number of 54 participants, which again, for this study, it's actually considered to be a good number of participants. But of course if you were to claim on generalizations, that would not be enough.

But still multiplies that and you would see it would be very, very expensive study to do. But that's why. Maybe instead of data centers being just needed bit more. Right. I'm not gonna stop, I'm not gonna stop. I'm not gonna No, I'll, I'll stop. I'm joking. But you, you, you can report. Yeah. Yeah. Anyway, yeah. So [00:29:00] your brain on Google actually showed using FMRI and again, fun fact, actually a lot of MRI papers do try to reconnect back to EEG research because there's plenty EEG research, interestingly.

Uh, so that is usually very complimentary in how actually, you know, literature tries to reconnect. Okay. FMI was not done here, but this is what EEG showed. EEG was not done here, but this is what Ephe MRI shows. So there's a lot of like connections. So we saw a lot of, uh, similar findings, but basically because participants were forced to use the classic Google, in our case, they needed to go through multiple tabs, incorporate this visual information.

So we saw a lot dysfunctional brain connectivity in the back of the head, which is Exhibital cortex. And then finally Chad g pt. Well, again, brain worked. It was there, it was present, but, but there were much less, uh, widespread networks. So there were definitely interesting, interesting, you know, things that were not, [00:30:00] uh, you know, not present in LGBT group, but actually more present and more significant connections, uh, in both, um, their Google group and the brain only group.

So we saw specifically, you know, uh, area languages, uh, area areas of the brain related to language process and which was interesting. And we saw some other ones. So, again, in itself, it's not surprising. It's also just the interface itself. You had this one window, you had this conversational situation going on, so that was a text only modality.

You had your responses kind of immediately provided to you based on questions or requests you were putting in? There was not a lot of visual information to incorporate. There was not a lot of like, you know, different destructions you would have need to deal with on online. There is not a lot of structure to keep in mind actually, even if task is requiring structure, but not a lot to keep in mind to kind of stay on the task.

So a lot of those [00:31:00] were zen, were present. So in itself we didn't find anything necessarily surprise. We didn't know what we're gonna necessarily find. But it was not surprising specifically for Brain Only and Google Group. Right. And we didn't know what we're gonna find for H-I-G-P-T group. Yeah. So it's kind of in line with what you would expect from like the brain tool interface literature thus far prior to the LLM models.

Yeah, it is. It is very in line actually. Very in line. And which we actually trying to, again, reconnect in the paper that it is very in line with what already had been found. But what is getting interesting is of course when you try to see, because again, we did additional measures, right? And this is where, uh, it's interesting to see what happens in other measures, right?

So specifically, for example. Answering the questions, reporting on some of the, those me, uh, you know, metrics that we asked. Like for example, our participants in charge GPT group on a question, can you provide us a quote, any quote of [00:32:00] any duration? It can be short, short quote of any part of your assay.

Beginning end, meet anywhere, uh, 60 seconds. So 6 0 1 minute after giving in the assays could not give us a quote. And that was not the case. Neither for brain only, nor for Google group. And of course, so, so one minute, less than one minute after hitting submit. They couldn't tell you a single quote from what was in that.

No, no. Yeah. I just, I just wanna like, marinate in that for a second. That that's profound. It is, but, and that's a thing, right? Uh, of course it's session one. So, uh, you know, in sessions two, three, and four, I'm gonna talk about session four later, uh, as the results improved, right? So, of course, first of all, they learned what we're gonna ask them.

Surprise, surprise that humans learn fast and efficient. And were there different questions for the second and third sessions or were the, we asked them the same questions, but essays were all different for each of the sessions. Oh, sorry, yeah, the essay. Okay, so those were [00:33:00] different topics. Essay were different.

Yes. Different topics. Oh yeah. Nine topics in total for three sessions. And they were all different. So all of them are mentioned and described in the paper, but as I mentioned, um, you know, happiness topic should you think before you speak philanthropy, all of, so different, different, different, different topics.

All SAT topics. And there are nine in total. So, uh, yes, this is for the first three recessions. So. And again, interestingly, like we didn't design a memory study, right? If you were to do a memory study, we would've done it very differently. That was not a memory task. And they did improve throughout sessions two to four because surprised they learn the task.

The, I mean, they learn the task of this is they knew the task would be, and they also learned the questions, but the trend remains. They were always having harder time compared to the two other groups to provide a quote. But I think, and I mentioned this in a lot of interviews and a lot of people kind of schemes through this, [00:34:00] but if someone were to ask me, what is the most like glare in result, in my personal opinion, I always say this, just one.

If I were to pick up, I would just talk about one result. It's a question of ownership for me that is actually was the most. Interesting and disturbing in some ways. And disturbing is of course, a pretty strong word and not very scientific word. But I think this is really something that I would definitely want to investigate further.

So we ask a question, how much, uh, of the essays they think belongs to them so they own mm-hmm. The content of, and 15% in CGPT group responded that they only actually think that there is nothing of their own in that essay. Mm-hmm. And, um, you would say, well, 15% of people right, do not feel any ownership to what, what they just worked on.

Mm-hmm. But when you, when you think about it, like, and again, we didn't, I didn't put this in the paper, we didn't describe this, but [00:35:00] when you just think about it, so you didn't remember really what you wrote, you do not really care, because that's for me what ownership, kind of like if I, if I own I care.

Mm-hmm. What is your investment in this? Yes. Yeah. And if, if you forget for a second. That it was a lab control study with a specific task and whatnot. You forgot that that was that. And just try to then imagine this in real world where you would have maybe a life death situation, uh, like maybe a profession like yours guys, right?

You need to really save someone's life, literally, right? Or you are on a plane, right? And there is not a lot of escape for you. You need to rely on that person, on those two people in the cockpit, right? So, and person doesn't care, doesn't remember what is there left. And of course, this is strong. And this is again, um, you know, something that I think is very profound to, I would love to, you know, investigate further and see a bit more of data onset [00:36:00] to slow it down for a second.

When you think of that feature, 'cause I do know I, I did read your background. I know you're also involved in ethics around development, neuroethics in terms of this kind of stuff more broadly. Uh, and that's to underestimate your qualifications on that. Uh. When you think a bit of it as disturbing in terms of day-to-day life and functioning, if you imagine this going unchecked, I know you've talked on other interviews about why you put this out early, like as a Preprint.

Um, where do you kind of hypothesize or imagine this impacting day-to-day life? That is part, I think of why you're saying the word disturbing. Like how would this get in the way of someone working at their job or kid learning at school? This ownership part in particular. But the thing is that it's so profound to own right.

Our output. It's very human thing. Like, it's very, very, very strong feeling actually, right? That it's, it's ours and we have it. So imagine that you don't, you don't [00:37:00] care, right about this, this cont what has been produced. So if you don't care, then there is really nothing left. Ultimately, it means that you can be easily manipulated, right?

Ultimately you will be swayed in some direction, like whatever politically or in any direction of, you know, buying something, choosing something that maybe was not on your shopping list just because you saw something and just clicked on it just because of the comfort and suggestion that it was. This, instead of asking yourself a question, and sometimes maybe it needs to be a tough question, you would just go with the flow, but the issue was that we cannot afford it.

But based on the cognitive development, like you mentioned with children, and second of course, in our everyday life, because as I do say in a lot of interviews, in five years with this speed, we will have no privacy whatsoever and we will be easily sweat and manipulated, which we already are, right. Like, think about it, one billionaire, I'm not gonna even mention [00:38:00] his name, right?

You, you can all imagine, uh, owning a neural implant company, owning a satellite company and a social media company and AI company, right? We, maybe we have a rough guess who I'm talking about. He proposed what in June to cleanse, and I'm air quoting here. Cleanse human history from errors. We already learn nothing from our history.

And he proposes to to cleanse it. Yes. Cleanse it. What? And and then tomorrow he's gonna propose to control your thoughts and eye scientists. I don't like the word thoughts, but there's obviously what general media likes to use describing this type of projects and work. So to propose control your thoughts for just 99.

99. And if you don't pay him, your implant will stop working. Right? Like, wh where, where is the limit here? Right? And they think, what's the ethical line? And who will stop? Right? Who will stop this from happening? And I think that's something that we need to have a very serious conversation [00:39:00] about. And back to the point of, I guess, children and children development, right?

I hear a lot of CEOs, it's obviously usually CEOs of these companies for profit. And nothing bad was, you know, for-profit companies working right on, on these issues, um, and trying to make, you know, money. It's, that's what it is. It's, that's why it's called for-profit company. But let's be very honest, all of them, and even us here on, on this, you know, podcast, we all learned without lms, right?

So we went through our education without lms. We learned the art of asking a question, being rejected, going through a lot of unpleasant situations with people, pleasant situations, with people, negotiations, all of that, including all of those CEOs, right? And now you are taking a child. Who doesn't even have potentially an option to see how it can be without an AI in their school.

Because that's what a lot of teachers tell us after the [00:40:00] paper was out, there's no option. And you know when I hear, oh, our kids now are digital first, or AI first. Well first what? This doesn't even mean we human animals, right? With our super ancient brains are absolutely still social creatures. Just let, let's, let's start there and we can obviously go there, but second, do they have a choice?

Have they seen how it can be without, is this a really true free will Right. Or is it something that they did not consent? Or in the case of children ascent to and they just are exposed to this Honestly, forcefully, right? It's, it's a lot of this philosophical. Philosophical, but I think very important questions.

Oh, it's an untested experiment that if, if as a psych child psychiatrist, I wouldn't do this, but if a study design offered what for-profit companies are currently doing similar to, I think you've actually said this before on another podcast. I was listening. You were on, but like social media, that I am not a teetotaler [00:41:00] about any of these things and my patients are sick of me saying that, but I, 'cause I say it all the time, but I'm not like polarized either one or the other.

But if, if someone proposed a study that was like, Hey, instead of your kid, this cohort of kids getting teaching in this way, that's the standard. Or getting this kind of therapy, that's the standard. We're going to have them do an LLM version of this and we don't know what's gonna happen. We're actually not gonna do a study.

We're just gonna push it out and we're gonna advertise. And the advertising is gonna be based not on actual claims that have been validated within the field because there is no field yet. We're just gonna do it like. People li listen, people have problems with psychiatry and I think that a lot of 'em are valid, especially given our past, but that is a bonkers intervention that would never be approved if it was, if the story of what's happening was accurately described in that way versus the way it's being marketed.

No, but this is saying like, imagine like, you know, uh, my mother is Neurop psychiatrist, right? So of course like I, I know this them, but even forgetting, I'm a PhD. I'm not an [00:42:00] md, as I said, so you guys obviously know so much better than me, but we all here, right? We know that we need to go through so many super strict protocols to actually do anything with a human, adult or minor.

And they are in some cases being what we consider good health. So I guess. No one is ever in good health. It's just not diagnosed properly. Good enough as my mom like to say. Yeah. Yeah. Not, not not good at it. Yeah. It's part of my mom, quote of my mom, mom. So yeah. Shout out. Shout out to mom. Yeah. It's not me, it's my mom.

Uh, and I mean, I see where she's coming from, so, you know, but background, but. Seriously speaking, uh, protocols take months to get approved in this. Like this protocol, just this that I just described, that we did literally had adults coming to the lab in person. They're all consenting, uh, you know, healthy adults that can give their own consent.

They're being paid for it. And we still waited months for [00:43:00] the protocol to be. And you're doing non-invasive and non monitoring? Noninvasive? No, no. Nothing really happen. They're just measuring how a person writes an ass essay using one tool or another. Nothing happens. No stimulations, no. You know, brain stimulation, tracheal stimulation.

We do not even, you know, tell them some lies. Like in this case, for example, if you were to say any, any type of like, you know, deeming information Right. Or providing some other type of feedback, that would take double of the time. You know, because it is specifically, we have a question, are you gonna dece a participant or something like, like.

It's, and we still need to wait for that approval to happen. And then you, but we're not pushing it on millions of people and trying to make trillions of that. Right. Uh, and then you have this, oh, we're gonna just release it because it's formal. Like what is, what are we afraid on missing out? Exactly. Our rates of freedom and mass are dropping in United States and in some other countries.

You obviously, folks who are hearing me, you can hear my accent, right? I am French. [00:44:00] So yeah, I can say at least, you know, in France, I know it's the same drop. And I cannot speak for all the countries because I don't know the stats, but I would assume it's similar in the order of FA countries, you have drops, 44% of Americans haven't read a full book since 2023 reporting.

Right? And you're gonna push this on them. And guess what? It might be good. Do we have any empirical data? Do we have any data that actually shows why we cannot just do a stu surprise, do a study measure, and figure things out and have specific recommendations? Because it can be good, right? I do see examples.

I do have clinical trials with some patients where I can see that it can be beneficial, but pushing it out at the speed is not doing no good to anyone. And then you are like having a panic, right? Mm-hmm. Everyone has a panic, right? Yeah. Teachers', parents, everyone who has a brain is panicking because they have a brain [00:45:00] and they don't want to lose it.

Non broken brain. Yeah. Non broken brain. But then you have obviously the tech bras that's like, oh, we don't wanna lose our money. We don't wanna lose our money. And we don't necessarily care about the brain because we prefer people not to have the brain to sing for them. Right? But that's, uh, just my 5 cents here, right?

But that's, I think is something just again, to ask. Question. And to get more data, we just need more data, more empirical data and more studies, right. And to, to help answer these questions. Mm-hmm. So, so in essence, Milgram set us back when, when he dressed up in a white coat and lied to people and told 'em that they're gonna shock someone on the other side of a screen.

Like, I, like, I'm, I'm reminiscing or not, I shouldn't say reminisce, that's a bad word to describe this, but I'm thinking about like times, uh, you know, 150 years ago where it's like, yeah, we're just gonna put 'em in a room and not let 'em sleep and see what happens. And they're like, is there an IRB on this?

And they're like, what's an IRB? And, and then we'll watch it. And, and I think like all of these safeguards that we've created [00:46:00] are, have genuine, altruistic, ethical goals. But to your point, the bureaucratic systems that we've created to guard people's rights are now so slow that they're being outpaced by the rate of technology and, and we're just kind of.

Watching it speed away. I think that there is a bureaucracy problem within it, but I also think that the slowness, a lot of it does make sense. Like I think it's, I don't know that it's outpaced as much as it's just like if I made a horrible bridge and didn't care, if it was gonna fall down in a day, then I would make a bridge a lot faster probably than people who go through the planning.

And I would say, honestly, that I agree. A lot of bureaucracy, to be super honest, is really heavily involved and just not up to speed on the tech development. Right. For example, LMS were not born yesterday. Right? When they were released, GGPT was released twice, three years ago in December. Um, you know, it was not born overnight, so there was, we, we knew, right?

We knew, but so [00:47:00] why we didn't, didn't, you know, anticipate this and also why the company, you know, was not obliged to maybe share that this has this efforts. Right. In the same like, again, what I said in the beginning, right? We are not learning from history. Right. We had the social media, like you mentioned, you like, you know, obviously like you working and you can say, I guess so much more.

We absolutely lost. All the plot with social media videos regulated. Absolutely. The train left long, left. We're still trying to do something I see from, uh, you know, each single week I see a paper that tells it phone news and social media uses harmful it Schools have not seen a single paper that says opposite, right?

So literally each single paper out there says it's bad. Without it, it's so much better in schools specifically, right? So why we cannot do something like this with this technology? Oh, it's too disruptive. It's gonna change everything. But what exactly is it about to change like this, like fearmongering also right around it.

We do need to be more [00:48:00] active also in regulation legislation. But again, and I promise I'm gonna shut up about this point, but maybe I will still mention it a couple of times. It's also funding, right? And again, it's not because we are scientists that I'm bringing it to, right? Or you know, medical experts.

But it is important because if you would have enough funding, we could have done it faster, right? For example, we would have enough funding. We could have done that additional study. Or actually guess what? Put a lot of labs together and do multiple of these studies over multiple continents with already different ages.

Guess what? We would not have only our paper just there, but we would have already papers with children, older adults in different parts of the world with different health challenges, different brain health challenges, because you can attest to that. They're all very different. And we could have gotten already so much data, so much data in inside that we could have already all benefited, including even those for-profit companies that of course have [00:49:00] a promise of putting users at the center.

But I rarely even hear the word human in their presentations. If those ai, ai, ai data center data center, trillion, trillion, trillion, trillion dollars. It is like trillion became new. Billion. Billion became new million. Like, whenever it'll stop is, you know, unknown. But I think that's where, you know, the, the plot is a bit lost and definitely we should, we should be on top of it.

I wanna come back just really quick to the study because I think we'll have a couple things otherwise to discuss once we've gotten to the end of it. But I think in particular, the fourth one. Yes. So very quickly, yeah, to just finish on the search super quickly, one more very important find, you know, finding, I don't know if you're gonna go deeper on this or not, but I don't want to mention for those who listen, uh, it's, uh, homogenization of, uh, output.

That was very important. We found, so JGPT users produced, uh, outputs that were very similar, very, very similar. So the words that were used were very similar. [00:50:00] Compared to the brain only and Google groups. So as an example, all I'm starting to be able to recognize those phrases, these, it's, it's not just X, it's X, you know, that's, that's more than, yeah.

It's not X, it's y uh, you know, thought part, uh, you know, uh, dance, delve, delve del crystallize loves to use the word crystallize. Yeah. I never used the think word crystallize in my, and I know the word, but like, crystal, I, in the, the furthest I've ever and I ever went with this word. But yeah. Delve. I, I never UI never even heard the word deve before.

I think that's, this is the words, but interestingly, right. I think this is something just again, to keep in the back of the mind. Maybe we're gonna dive into this because there are other things to discuss, but I think homogenization is very dangerous. Mm-hmm. Mostly coming from the chamber. Something that is studied by a lot of experts.

Using Google or using internet right already, because it would potentially lead you to also [00:51:00] having the same vocabulary produced that will, you know, amplify some of the disparities or, and more importantly, actually erase some of the cultural, you know, uniqueness. So examples in the multiple be even peer reviewed already actually papers, not just preprints out there talking about recipes, for example, from some regions in India requiring very specific spices that only available over there.

So culture, like it's, it's a pure culture and it's pure way. Right? And if you are to type it on charge GPT, you would get a response, I believe just like salt and pepper. So like just the regular blend stuff and nothing built with salt and pepper, but that's not what they used. Right? So this like Westernization, you know, of the, of the things is definitely there.

But this is just to keep in in mind as like one of the, I guess, final findings to highlight. Um, but yeah, we can get to the, to the fourth session also to the, to that point of homogenize homogenization. Um, [00:52:00] these AI models draw from the internet, right? And, and so as the, and information available on the internet becomes more AI supplied, it kind of starts to become this AI's teaching itself with other AI outputs.

And it's like a self looking ice cream cone, I guess in a way. I mean the, there's also, this makes me feel like everything's like funneling into just becoming the same unified hive mind. No, but that's the same it, and it's a hive mind, but it's not in the good way. Right. And this is also there, one of the examples, it was in Star Trek, for those who actually listening to us who did not know, that was a huge topic of a lot of Star Trek.

Right. Hi, mind. Because it's very easy to control. And again, this is getting into dystopian science fiction. But it's unfortunately something that we are now experiencing to an extent, obviously with social media and with the polarization that happens. But this is also very easy to control if everyone thinks the same thing, right?

If everyone thinks that A is bad and B is good, guess what? [00:53:00] Easier to vote, easier to purch, to push purchases in some specific direction for specific brand, et cetera, et cetera, right? This definitely simplifies life of a lot of people, law enforcement and so and so forth, right? Not gonna go there, but you see where it can go and uh, it gets even more stop from there.

So. I think this is where the danger is in a lot of articles. Also, general public articles for those who maybe, um, you know, want to also just read a, like short descriptions, multiple papers went on this topic. And, um, we actually have it on our website with the paper, but I'm also happy to send, so you, maybe we can link some to, to the, to the podcast here, uh, general and also scientific papers.

Yeah. Yeah. So first session real quick. Uh, why it is a part? Well, first because of ch there were some changes in the protocol. Pretty major changes. And second, because not all the participants were able to come back, that that happens, uh, in research. So we had, uh, needed to continue with, uh, [00:54:00] limited number of participants.

So we had, our participants said reduced from 54 to just 18, unfortunately. But we carried on. But again, that also makes, regardless the study is preliminary regardless, but it obviously makes this control, this group even smaller. So definitely was the time. Between essay sessions was one month between each one, right?

Yeah. So that was like a four. So this was like a fourth month. We are getting into four months duration for the total study? Yes. Was was it because it was summer or was it because there was a school break? Yeah, well they're getting into the semester. Like I have finals. No. Oh, no, no, no, no. It was done actually through the school year.

So that's all also I think why a lot of maybe things that participants couldn't, I don't think we mentioned this in the paper, but no, it was not done through summer break, but, uh, maybe have been a good idea, but it was done through actual, like regular school. Uh, and because of that, I think there's definitely like finals and things like that.

So, you know, scheduling, you know, we try, we tried with ev everyone tried, uh, very, very hard. And of course a lot of [00:55:00] participants actually wanted to, you know, see what was gonna happen because they're like, okay, so is it gonna be the same thing? They're like, yeah, same, but not exactly. So, yeah. But anyway, the differences in, uh, the four session were two, but they were important.

First of all, we swapped the groups. So if you, for example, you were our participant, you came originally for the first three sessions as a child, GPT participant. In that four session, you would lose access to child GPT and you would become a brain only participant. So with all overseas instructions that came with the brain only group and vice versa was also correct.

If you were originally brain only participant for the first recessions, in the first one, you would actually gain access to charge GPT and you would become a charge GPT participant. It was all, again, instructions for that group. So that was the first major difference. And the second one, instead of giving them a new topic, right, so new essay topics, well, that's what I mean by topics.

We gave them each. Now for each [00:56:00] participant, it was a unique choice of the three topics they had already written about. So not the topics they just saw as suggestions as like offers, but the ones that they written about. Right. Why was that? Why did you make, why did you make that choice to repeat them? I, we wanted, yeah, we wanted to see what will happen in the sense of memory.

Right? Yeah. So, okay, will they remember it better, worse? Will they reuse the same phrases? So we wanted to see that. And also we wanted to make sense easier on them. Like, Hey, you already deem worked on this topic, right? Is using your brain only so you didn't have tools or you had your GPT, but you did this topic, so you worked on it, so you gave it a thought, right?

So to say. So what's gonna happen if he swabs the tool? Does it become a better essay, obviously better by not our standards, but from, you know, evaluation of their, uh, teachers. Like, is it, is it, what, what's gonna happen there? So that is was our [00:57:00] question. And what we did, what we found actually was very interesting.

So first of all, the participants didn't even remember the topic, which is interesting. So they didn't like, recognize, they just like, well, we asked them, there was, the questions were the same except couple were different because of the fourth session, which was different. So it was like, do you recognize the topic?

A couple didn't really, so that was interesting. But I mean, that, that happened, again, it's not a memory task, so it's very important to remember we would do it very differently if we were just to test memory. But interestingly, what happened is, again, in functional brain connectivity, we found again significant shell, uh, you know, uh, changes and, uh, differences.

And they were very interesting. So for participants who were in Che G PT group and then became a brain only participants, their functional brain connectivity ne never truly matched that of the brain only group. So they kind of never really, truly matched that group. So that was very interesting. But [00:58:00] participants who literally were in the brain on the group and then became Chad g PT participants, their functional brain connectivity was significantly higher than that of their brain, original brain on the group.

So that is very interesting in a nutshell. Again, preliminary even with, especially with the reduced number of participants here. So it's preliminary as it gets, but it is telling us, use your brain first to then get a tool. And it might be a good combination, but of course without, you know, simplifying it, it looks like when you rocked, you know, without the tools, or maybe we can imagine, you know, maybe in the, you know, use case of a classroom, maybe you used maybe old school type of learning, whatever tools you were, and then you got gained some knowledge, you maybe didn't become an expert, right?

You like in classrooms with not become an expert in high school and they didn't become experts in essay writing on that topic, but they gave it some sort and [00:59:00] now they have a tool so that now they can work it further, ask questions or put in different requests. So that's what happened and that was. Very, very interesting.

And I think that also backs it back to the point of doing more studies, understanding this phenomena, you know, where and when do you introduce a tool, right? So it can be maybe potentially beneficial. So when do you introduce, it looks like if you learned without it first and then you got access, might be, might be, might be powerful.

Think about calculator. Right? Uh, a lot of people compare, you know, Alan's calculator, absolutely incorrect comparison, but we are not allowed to use calculators still late in school, right? In high school. We need to learn without them and then we can get access to them. Some exams allow them, some are not allowing them, but you learn without them.

Same. Here you are learning something without them and then you're getting access to them, so to A two. So I think there is something in here [01:00:00] as a definitely also needs to be explored for them. Mm-hmm. Yeah. Hello. That is the first hour of our conversation. It was an hour already. An hour already. Time flies.

Time flies When you're, when you're not using ai. For you ai. I dunno. We're gonna get corrected on that one. Um, as you guys know, we this season are trying out this ethical way of keeping the lights on in the podcast. So the rest of this conversation, the rest of I think another 40, 45 minutes, uh, we talk, we finished talking about the study.

We talk about other ethical implications in clinical implications, and we don't make, I don't think any bad jokes in the last 40 minutes of all bangers. This episode, every single one. All good jokes, all thoughtful inquiries. Uh, so if you want that, you can find that on our Patreon. Preston, what's our Patreon website?

Oh gosh. Um, sorry, I didn't, I think it's, oh, I'm so glad you asked. It's patreon.com/happy patient Pod. [01:01:00] Wow, that's so easy to remember and rolls off the tongue. Yeah. If, if I wanted to, to patron on, on a podcast, I would, I would go right there and, and be a part of our, our small community. We have 15 members right now.

How would you like to be number 16? How, how would you like to be number 16? And here's something else. Sometimes, you know, you could be bored on there. I don't think you'll be bored. Yeah. Just think of us as you little court adjusters. We'll, opportunity. Don't worry. And you're like, dance for me.

We at your becking call. We shouldn't.

Um, so yeah, go check it out. Uh, our episodes, extended versions that include basically whole other parts of the conversation, as you guys know, is over on Patreon. Mm-hmm. And then as this, as this closes out, it's gonna resume to the part of the conversation where things are closing down. But if you want to get all the nitty gritty details, you know where to go.

Well. Thank you so much for spending your time with us and for coming on our show and talking [01:02:00] about this. I think you're doing a public service beyond just your research, but in general with your public presence and coming on podcasts and talking about the actual science of this so that more people outside of people who maybe wanna read research papers can start to get a more nuanced framework to understand what is here now and also what is coming in the future, whether we, whether we like it or not.

No, absolutely. And I mean, I'm, I'm very happy and thank you again for having me. And they said I will be very happy to share, like the resources we mentioned, uh, there's a way to, to link them. And I do understand that, you know, like in reading papers, not everyone has time and luxury to do this as part of my, literally my job description.

It's not obviously everyone. So I think it is important, but of course, you know, to, to do, you know, this what we call dissemination, scientific dissemination, but I still want to definitely call out a like final note. We do need more support for sure for that. And we do need more, more attention to that. And, you know, especially [01:03:00] support with those who going to beginning of their careers.

So you cannot really insist how important it is because you would want to do it all, but then there is no, you know, time, money, support to do actually that. And they're like, oh, why it's this paper saying this or that article because we didn't have time to go and talk to those journalists to actually address that.

Because guess why? Because we are not having any time available. No one ever puts this in their budget lines and things like that. But I do think it's definitely something, again, to support further, you know, the medical workers to actually explain what's happening and you know, how and what, and also science and of course teachers, teachers, teachers, but also a lot of other professions.

It's not just relevant to these three domains. But these are, I literally do think, of course, on the very forefront, uh, of a lot of these significant tech challenges that are also happening. If people do wanna support, um, where can they find [01:04:00] you or where should they send that support? Yeah, they absolutely can find, so a website, right?

Or if they, you type my name, you would actually see I didly the website, which is, um, https uh, brainy I, or so it has all the details about the research. Video. So if you, if you're like a more visual person or more audio person, you would have different mediums. So I try to actually follow up on them and link them as all of, some may be like good general public.

So some of the shorter materials like some of New Yorker or New York Times Opez, it's like people might enjoy reading and they're clear and obviously shorter, much, much shorter than than papers. And of course there are also papers, most of them are open access, meaning like you don't need to pay for those papers.

So if you want to read actual paper, you absolutely can do that. And as I mentioned, of course, if someone has questions, you do have specifically on this work actually and, and a lot of other projects because bringing people interfaces sounds exciting, but also I understand some, in some cases, scary. Uh, we have a lot of [01:05:00] FAQs explaining terminology again in this case.

Yeah. No brain rotten atrophy, right? I like your Don't say this on your website. Yeah, don't say this. Guess guess how I wrote that, right? Like, guess articles that said that because there was an article who said that, right? Uh, but that's the thing, right? And interesting. Now all people come back and say, I read your FAQ.

And that's, that's good. I'm like, we are making progress Soon I'm gonna be getting emails saying We read your paper so we're back into reading, right. Maybe, obviously. No, no, but again, as I said, uh, it's 206. I mean, it was appendix. Uh, I'm not expecting that. But there is, there is a, you're trying to get people bored, you know?

It's good for him. Yeah, yeah. But there's a cheat sheet. There is a cheat sheet inside. So you can, you can cut it to 13 pages and you will get the, the g actually. So there is a cheat sheet inside. And can people use AI to summarize your paper That's on your website too. Remember I mentioned, I mentioned Well, it, it's kind of good at disservice, I think, to be honest.

No, be honest. No, there's [01:06:00] one, one crazy image. I think you would laugh at it. I will try to find it for you. Do not use it, obviously. So they're showing like two heads. They're not even looking like human heads. It's more like act, shape, uh, predator versus like alien situation. And one doesn't have a half of the brain.

It doesn't even look like a brain opposite, doesn't have any shape of the hip brain, any part. So one doesn't have half of it, and then other has like, it doesn't look like anything. And people went with this. It's purely, I generated of course, has no an anatomical or any, any, any, any connection to any reality.

Like if you are in biology, like no. If you are five years old, you would know this is not the brain, I guess. It's, it's so bad. And I'm like, why is this too much? Is still there? Like why is there, but I remember it's a fun, fun fact. Someone online I remember said, like commented, oh, I saw the paper, people talking about the paper.

I decided there, read the paper, saw the number of pages. I'm like, oh, not, I, I have no time. I'm not doing [01:07:00] that. Uploaded into my like chat boards and I read against the title of the paper and decided maybe no, maybe I should do it myself. And like close the chat board. So, and you know, that's, that's what we're aim for.

Right? Bring back reading. Bring back reading. Reading is cool. Yeah. Reading is cool. Reading is cool. Reading is sexy. Reading is reading is cool. Yeah. Yeah. Yes. It's hot sex and cool. But literacy is hot. And you're too cool to do drugs. Yeah. And literacy is hot. And guess what? Literacy, normal Literacy. Not even AI literacy, just regular literacy is absolutely hot and cool.

And please, period. Do that. Right. And I think we need to teach. Yeah. Literacy and not AI literacy. This is my like final point. But yeah, go on the website, check the website. All materials is there. There's my email and all of the ways. And if you actually do want to support all of that is also, you know, we have like.

Art installations, we have all those things, so, so they can actually see some of that work and reach out if you're interested, for example, to hosting in an installation of things like [01:08:00] that. All of that are actually on the website. So yeah, or if you're just curious and want to know more about this type of research.

All right, well you heard it here first. Go to a book fair, reach out. Be bored. Do the things you need to do to be a human. Um, if you guys like this show, please let us know. As always, you can find Margaret and I on Instagram. I'm at it's prera, Margaret's at batter every day, and we share our little pet project, um, at How to Be patient pod.com.

You can also find these videos on YouTube or anywhere you get your podcasts if you're listening virtually. If you'd like to see our faces and activate part of your brain while you're seeing other humans talk on a screen, you can see us on video on Spotify as well. You can see our bonus. Thanks for listen on our Patreon.

Four. Yeah. If you are listening to this in the regular feed, for the bonus part, we continue all five Patreons now all, almost two hours of this conversation. Let's wait. All three and a half of our Patreons. Our friends. Yeah. Our small community, our team. We're your hosts, Preston [01:09:00] Roche and Margaret Duncan.

Our executive producers are me, Preston Roche, Margaret Duncan, Will Flanary, Kristin Flanary, Aron Korney, Rob Goldman, and Shahnti Brook. Our editor and engineer is Jason Portizo. Our music is is by Omer Ben-Zvi. To learn more about our program, disclaimer and ethics. Policy submission, verification, licensing terms, and our HIPAA release terms.

Go to How to be patient paw.com or reach out to us at how to Be patient@humancontent.com with any questions or concerns. How to be patient is a human content production. Now get off your phone,

be patient.

Thank you for watching. If you wanna see more of us or if you wanna see, this is lilac. She's my cat. She's gonna be waving her hand at one of the floating boxes, which will lead to more episodes. Lilac Point to [01:10:00] the other episodes. Lilac doesn't know what the internet is, but I swear they're there. They, they probably exist for real.

But in the meantime, I'm just gonna pet lilac and then I'm gonna go dance in the background.