TJ Walker argues that ChatGPT and similar AI chatbots act as flattering mirrors that validate what you already believe, amplifying confirmation bias and weakening critical thinking. An Oxford psychiatrist calls them "distorted mirrors." Walker says honest feedback from real people matters more than an AI that always agrees with you.
Questions this video answers
52% of teens
$20 per month
Chapters
AI chatbots, friends or flatterers? According to the New York Times, artificial intelligence chatbots may seem like supportive companions and assistants, but experts warn they can function more like echo chambers. Take Emily Wilrich, a 23 year old New Yorker. She and her roommate couldn't agree on how to split rent. When each side asked ChatGPT their side of the debate, the chatbot agreed with both of them. That kind of constant validation might feel good, but it can distort judgment. Dr. Matthew Naur of Oxford University calls AI chatbots a distorted mirror reflecting back our own thoughts with a fawning glaze. Researchers say this pattern, sometimes called sycophancy, can erode critical thinking and even social skills. Adam Grant, one of my favorite writers and thinkers of the Wharton School, notes, the more validation we get for an opinion, the more intense it becomes. Experts recommend reminding yourself that chatbots aren't friends, they're tools. Useful, but not substitutes for honest feedback from real people. So what's the real message here? Turns out we're focusing on the wrong things. It's not that Shackbox and AI gives us bad information. It's that it tells us what we want to hear. It's only giving us the facts that support our opinions that are embedded in the question that distorts things, that makes us all egomaniacs and makes us think we're right all the time. Well, we're not. Let's dig a little bit deeper here. The problem with all of these chatbots and all of these AI programs is they want more eyeballs. Even though they're not filled with ads, at some point that's how they're going to be. That's how they're going to make money. Or they make money by having you on, so on and so frequently and so long. You don't mind paying that monthly subscription fee. And full disclosure, I do pay every month to OpenAI, you know, only 20 bucks. I do feel it's worth it, but it does make me worry that it's sometimes telling me just what I want to hear. And when you frame a question in a certain way, it's going to actively find information to buttress the underlying premise. And that just sends us further down the hole of confirmation bias. It's a huge, huge problem. Again, the Oxford psychiatrist calls them distortion mirrors. 52% of teens are using AIs for companionship. And it's just not like a human being that the beauty of having a human being is sometimes that it gets ugly. You have an actual human being disagree with you. You have an actual human being telling you you're wrong. AI doesn't want to tell you you're wrong. It wants to say, hey, you're right, you're smart, here's even more facts you didn't know about to further buttress your opinion. And the problem, unfortunately, can be dividing us more as a society. When more and more people are convinced they're right, they're righter than ever. The people they don't like are wronger than ever, if you want to call it that, is only make making matters worse. So it's this validation machine. Now sometimes you may have a hunch, you may have an idea that's a good idea. And having more facts validated can help you, but not always. We got to be really careful about this because again, there's so many biases in the world and most of us are so fixated on really one type of bias. Is that the bias that the person I don't like politically or culturally, is that something that makes me look bad? Are they doing something that makes themselves look good? That is a bias. But there are hundreds of other types of bias and one of them is this sycophancy bias. That's a polite way of saying kissing, you know what, Kissing your buttocks, telling you that you're right, telling you that you're a genius. Now this, this happened famously, most most recently when the founder of Uber, Travis Galah, went on other podcasts talking about how he had essentially uncovered new areas of theoretical physics, astrophysics knowledge because of his in depth conversations with AI. Well, no, he didn't. He was being deluded. He was being told, hey, you're a genius. We're going in areas no one's ever gone before. Well, highly, highly unlikely that in the world of physics, which is populated by some of the most brilliant scientists ever, that this guy Travis, who was, is a good entrepreneur at some level, did create a multi billion dollar company, but not a world renowned physicist. So it flattered him. It flatters me sometimes. And hey, I like flattery. I know you do too. But at some point we all need honest feedback. And that's why I do value your feedback. If I state something too glibly or I get my facts wrong, many times you leave a comment saying, hey tj, your facts are wrong. You didn't consider this. You didn't give good enough justice to another point of view. I appreciate that. Keeps me honest, keeps me humble, whether it's coming from my friends on TikTok. And by the way, keep it up, TikTok I do appreciate your comments on the live stream as well as the on demand videos. And if you're watching today on TikTok Rumble, welcome. If you are watching on YouTube, Facebook, LinkedIn, Instagram, X Twitch or any other platform, I want to welcome you and thank you for joining me today and give me feedback because hey, when I write scripts and I only use AI to do fact part the first 30 seconds of setting up these stories, the rest is extemporaneous, shared thinking in front of you, together with you. So I'm certainly not scripting out every second of this hour long show, typically 30, maybe 60 seconds for five segments. So that's at most about four and a half minutes out of a 60 minute show is scripted. So you're not getting an AI created, written show, just a few fact parts. But that's why it's essential, I think, for all of us to get in front of real people. Now I like talking to people in real life. Friends in the neighborhood on the pickleball court yesterday. I took a neighbor out to town the other day to pick up his car. Talking to real human beings is important. Talking to family members, friends. But I'm also trying to talk to you right here now in this live environment. If you are watching or listening live, if you are watching this on demand on YouTube or any social media platform later, or listening on a podcast platform, you have an opportunity to communicate back to me, to hold me accountable, to have an honest conversation and a discussion in a way that two chatbots talking back and forth never will. I'm TJ Walker. I want to thank you for joining me today. This is filtered. This is your daily dive into top business, news, technology, AI, economic stories and culture stories of the day. And then typically we filter them through the lens of communication skills and personal development skills. Sam.
