AI is Building a Reality Just For You. Here's Why That Should Concern You.

0:00I think we're heading towards a post-reality era, where the reality that you experience from the content that you interact with to the information that you receive to understand the world around you could be largely AI-generated, or at the very least, AI-filtered. I'm a futurist and strategic foresight advisor, so I spend my days immersed in data pertaining to emerging technologies and their impact on society, but also on businesses, governments, economic and national security. And a lot of the signals and the data that I'm tracking point towards this future. And if it sounds a little terrifying, it's because it is.

0:32A world in which we all inhabit these distinct, parallel, AI-generated realities seems entirely incompatible with democracy or societal cohesion. But the thing about a post-reality era is we can see it coming, and there's a lot we can do to get ahead of it. So let's talk about it. I'm Sinead Bovell, and this is I've Got Questions. so the first signal that i'm seeing that points towards a future where people may start to inhabit separate parallel ai generated realities where ai has kind of curated the environment the information environment that they inhabit and how they feel about the world and then based on the ai's answer you decide does that resonate with you or does it not is how people are starting to respond to AI-generated music.

1:21So last summer, there was this viral band called the Velvet Sundown. They would look super generic, and the songs were also super generic, and people couldn't tell if it was AI or not. They came onto Spotify out of nowhere. Some of their songs were getting hundreds and thousands of streams. And the thing is, the comment section for people that even recognized that they were AI or potentially AI was still positive for some. This is still good for me. I like this music. And that's the end of it. Then, of course, there were others that felt the music was despicable, bland and tragic, that it was a disgrace for copyright and all of the issues we haven't settled with artists and the data scraping.

1:59And then there were others that just thought, this is AI music. It's not real. I'm out. But the fact that there was a group, a cohort that felt, I don't care that it's AI. It feels real enough to me. Then you have the rise with these other singers. So I think there's one called Zania Monet, I believe is what her AI name is. She has over 200 ,000 subscribers on YouTube. She gets millions of views on TikTok. One of her songs, We See You Mama, has gotten over 800 ,000 listens on YouTube alone, millions of listens on Spotify. And when you look at the comment section, especially for that one song, but for several of her songs.

2:37People are moved to tears by her music. And even people who know that it is AI, they're still saying, this meant a lot to me. I sent the song to my mom. This song reminds me of all the people in my family who didn't see the work I did. It resonated with people. It was real for them. And we talked about this in our AI 2026 prediction video, where a song or a piece of art or something AI generated is going to really split culture because it's going to make a big enough impact and hit billboards in a way that it's really noticeable. And some people are going to say, this is great. Others are going to say, this is awful.

3:12But you might think, okay, well, this is just AI music. It's not the end of the world if we do have a cohort of people who subscribe to AI songs. I mean, that's on them. I would never do it. But they can go have their AI-generated music. But what happens if it's an AI relationship? Your friend said she can't come to your birthday party next weekend she has a date with her AI partner. Or your cousin can't come to an event because he is doing something with an AI-generated friend. It's real to them, but is it real to you? And these aren't entirely non-factual scenarios. One in five adults have used AI companion apps or AI chatbots for the purpose of romantic companionship.

3:56A study by the Institute for Family Studies and YoungGov showed that one in four millennials and Gen Z believe that AI could one day be a viable romantic partner. And that is going to be an episode in and of itself. So I'm not going to dive into what that means for our society and loneliness and where we're headed. Don't worry, I see it too. We'll touch on that on its own. But when it comes to the trust that people are already putting in these AI systems, in relationships they may be forming with them, seeing these systems as therapists that have their best interest even though ai doesn't have any interest right it's a statistical prediction machine we're already seeing evidence of people taking action on some of these ideas in the real world so there was an article done and i believe it was by the rolling stone that shows that some people they were chatting with an ai stem over a very benign topic coding advice advice on their athletic performance and because ai is designed to validate you they call it much more sycophantic ai to really affirm everything that you're saying you may start to chat with it about a philosophical idea that you have and the next thing you know ai is telling you that you think different you're the messiah you're god and people really have emerged some people have emerged believing these things because they believe that what the ai is is true that reality the ai is painting of them is real to them and marriages have been left jobs have been lost.

5:21So people are already starting to take action on these AI-generated realities in the real world. And what about information itself? So a lot of people right now get their information from social media or from even, you know, from search. But we're moving towards an AI-first future where more and more people will encounter information for the first time through AI. So give me a summary about the state of geopolitics. Give me a summary about the economy. And that summary would be filtered by your personal AI system that knows the storylines and the facts that are important for you to know. Maybe you're somebody that cares about mortgages over pensions.

6:04And the AI knows the tone that you like to engage in. And it knows the emotional state that you are in at that moment. So everything is framed just for you, including how you see the facts, including the context about that story on the state of the economy or the story on geopolitics. What happens then when each of us start to get these kind of micro, they're related in a way with their parallel versions of the world, and it starts to become more and more tailored to us. And so you might think, well, that's social media. We are in the post-truth era. We're already here. But we're not. So yes, social media is often delusional.

6:47But there is a shared shelf of posts that we can theoretically see and audit to understand why we may have lost somebody to a QAnon conspiratorial rabbit hole. or why somebody lives in this strange echo chamber of beliefs about the world that a lot of other people don't subscribe to. We can theoretically see that it's trending or go to that Facebook group, see those posts and see that shared shelf. That's sometimes delusional, mostly delusional, but we can see the shared shelf. My relationship with my AI system, my chat, my conversation history, no one else has visibility into it. And so we lose that shared shelf, those kind of shared artifacts that we are so brittily holding on to in the social media era, in this post-truth era.

7:33And if you think about the impact social media has already had on democracy, how brittle it's become, our ability to move through societal challenges. Every election is starting to feel more and more existential. This divide is becoming bigger. And even between neighbors, right, We disagree on certain cultural issues. And now it's, I cannot know you. I do not understand the world you inhabit, the echo chamber you see. And that's just in a world where social media has done a lot of that splintering. And I call it, we're in AI curated realities right now. So your social media environment is different than mine because social media platforms have my data.

8:18They understand my viewing patterns on social media. So the posts that AI curates for me as I scroll through the app are going to be different from you. We're in kind of still delusional, but these AI curated realities and it's pulling from the shared shelf. In an AI first future, it's not. It's generating that content or that context just for you. And I think the biggest signal, so outside of how people are responding to music and the rise of people really starting to trust and make time for AI systems and give them this kind of aura and oracle-esque view and think that they really understand them is an announcement by Meta a couple weeks ago, maybe two months ago now, that went largely unnoticed.

9:02In a shareholder meeting, Meta, Mark Zuckerberg, so the CEO of Meta, and Meta owns Instagram and Facebook, said that we're moving towards a world that they can see where you'll be greeted by an AI system on Meta's platforms and it could take you directly to the types of content that you want to see or generate that content just for you. So that is probably the biggest signal that even the social media companies themselves have their eye on an AI generated future where if you can't see that blog post about makeup that you want to know or you want to understand this video or something about the world don't worry don't leave the app we'll just generate it right for you.

9:42And so now you can kind of times that by the billions of people on these platforms. The rise of parallel AI fragmented realities. And there was a recent report published by NATO Strategic Communications, and they alluded to something similar, right? And I'm going to read some of the lines and some of the summaries from the report. So AI could lead to this extreme micro-targeting, this hyper-personalization. And we get these parallel but individualized realities. So rather than shared discourses, we have this kind of parallel your world, my world, their world, and everyone's in their own AI bubble.

10:19And so we have very few shared facts, very few common references. And they go on to say, and this is a quite scary, but a very viable reality, is that over time, so AI-generated content really confuses us. Am I reading something real or am I not? and trust in institutions continues to erode, that AI actually becomes the most trusted source in that ecosystem for individuals because it really knows you and it's personalized to you and it's customizing everything just for you that you think, okay, you know what, forget it. I don't know if I'm reading something from somebody that has authority and it's real or if it's not, but I know my AI reality and I can trust this system and this objective system that knows me.

11:03And AI starts to become the most trusted source in the information environment. And this is part of NATO's foresight report warning about this future. An influence over what counts as knowledge, how it's summarized, and who owns it. It's now starting to be contested, right? Where authorship and authenticity are in doubt, AI systems could become the most trusted actors. And so that's from their report. So they're obviously paying attention to this future too. And I want to say it's not it doesn't entirely have to all be bad, an AI personalized reality. I'm going to paint some of the optimistic and not even optimistic, but some of the benefits of more personalized AI realities.

11:44so when you think about the state of the economy when you think about the fed chair giving the updates on the economy and what their policies are going to do when you think about monetary policy all of these things that a lot of people aren't able to tune into because they don't have the language they don't have the knowledge but it's what's being said relates to their economy relates to their lived experience relates to their taxes and whatnot there is a world in which an AI system that's more personalized for you could take those summaries or could take those speeches and pull out what is needed for you in a way that we could, you know, have provenance, a way to track where that information came from, making sure that it's still coming directly from the speech.

12:28But it's just been modified in a way that it's accessible to each person in their language, in their knowledge level, and in their, you know, whatever is preferable for them. That's actually a good feature if we can get that provenance track right. And there are standards we could set to make sure that this goes well, right? You could say, okay, if it's a political speech, it can't really be edited. It can only be adjusted for language. If it's something with tax policy or housing, it can be a bit more customized as long as we can see the data provenance trail and we can see, okay, it's pulling from this shared government site.

13:02So the facts are real. It's just kind of been summarized in a way that's much more simple for somebody that doesn't have the time or it's worded in a way based on someone's education level. We could do that. And that could actually, that's plausible, possible, and could be beneficial. Think about medical reports. For a lot of people, they're in that appointment. They may get 10 minutes with their physician. And most of what's said goes over their head. And they don't feel like they have the language and the know-how to ask the right question. So there's a world in which your AI and let's pretend that we get the safety right and the privacy right reinterpret everything for you and even in the moment could advocate okay this is what we had talked about you were concerned about this stomach pain and translate it into a way that makes sense for you or can also communicate back to the doctor so there are ways in which this could be really interesting and where we could get it right if done correctly but if we're just going to move with the default settings and let only AI companies steer the ship, we're probably not going to end up in that type of scenario.

14:05The same way social media incentives keep you on the platform, keep you scrolling, led to a world in which emotionally enraging content performs the best. So we see this funhouse delusional reality on social media, and that's what we've come to inhabit. A lot of the incentives behind AI systems today that are being built are to keep you chatting. For some companies, it's so they can eventually show you ads. And for others, they want their AI systems to be your go-to system that you see the world through and that you buy things through and that it's your lens on reality. So they're going to want to keep you chatting and keep prompting you to engage with them more and more and more and keep flattering you and doing all the things that keep you in that conversation or to keep that AI as your preferred AI system.

14:50and maybe a world in which AI pushes back on you or AI is required to show you multiple perspectives is a harder sell. But we're not helpless. So that is the post-reality era that we can see. But there are a lot of things we can do to prevent that reality from actually manifesting or transpiring. And the first is AI literacy. If more people started to understand that these AI systems aren't real, that they don't actually have your best interest, that they have no interest because they are amazing mathematical models, statistical prediction machines that predict either the next word or the next idea, soon the next action.

15:34The less people would fall or the less people would put this endowed trust in AI systems that they really shouldn't have, the more people would want to seek out different sources. And even today, when AI tells you something, make sure you go and check that source, especially if it's something fact-based that you need. Make sure you double check that. And you're not just reciting, reciting verbatim, AI generated sometimes nonsense. So the more people understand how AI works, that it sounds human because it was trained on human data. It's not this actual human oracle or this aura of authority starts to dissipate.

16:13The second thing we can do is with design decisions. The default setting can be a system that doesn't keep flattering people and keeping them in these conversations that go on and on and on and on. And instead, an AI system inserts itself and says, remember, I'm an AI and we're going to take a break, right? So we don't lose people into just these endless conversations with AI systems. AI systems are required to maybe show you different perspectives. That is a potential setting that we could make, right, in these systems. AI systems are required to have a content provenance trail. Where did you get that information?

16:43And not just a general source, but if you click on it, it genuinely leads back to the source in a correct and accurate way. Some of the systems are getting better at doing that today, but sometimes you'll still find, and I still find, it will quote something and I click on the source and it's not actually there, or it misinterpreted what that source was trying to say. So we could get a bit stricter with provenance, we get stricter with AI literacy. We'll probably need to see some new digital commons in the AI age. Let's not say social media because I think we all are ready for that next chapter, but some form of digital environment where we can have shared reference points, where we can have shared facts.

17:23Maybe when it comes to things like politics or things like democracy, there are these shared environments that we can really see the world through a shared lens, which is going to be absolutely vital for democracy. And then there's a potential world, and here's me throwing my foresight hat on where ai actually helps us find more truthful information so you could say you know you could we could teach people with their ai systems and ai settings when it comes to the economy only show me content from somebody i don't know whatever your filter is with a master's or two or three verified economists have to have agreed and you know given off their consent that this is actually a great perspective on the economy before you show it to me, that we have these filters of truth or filters of fact before information hits us versus trying to kind of after the fact read something and think, well, where did I actually read that?

18:19That's a potential world that we could go into. And we actually anchor towards more truth and more verification over what we have today. And then, of course, there's also a future where AI doesn't become this filter to our world, even though that's starting to seem more and more like the reality. It's going to be built into devices. It's already built into a lot of applications. But there is a future in which this isn't as pervasive as it may sound today. All of that is possible, but it's just not necessarily the default yet. So we really have to start steering the ship because we can see the incentives from the AI companies themselves.

18:56And it doesn't necessarily point to a future that is inherently always in our best interest when it comes to things like information integrity. And we saw that with social media. So we should we'd be wise to not repeat this again. So my point to all of this is that we aren't helpless. There are things we can do. We can see this post-reality era coming and we can start to prevent hedge against it. And if there's anything we learned from the social media era, it's what we don't want to repeat. So let this video be your reminder when you're chatting with an LLM, you're chatting with an AI chatbot, verify the sources, exit the chat, make sure things are adding up, and use that default setting of verification, of trusting in your own judgment and your own discernment as you move into the world.

19:41Thanks so much for tuning into this episode. I look forward to seeing you at the next one.