AI Is Making You Dumber (study)

Did you know that AI is making us dumber?

Yeah, there's been a lot of studies on this. It makes people think less. I think Harvard or one of those colleges released something on this a few months ago.

For the sake of this clip, he doesn't know. So, so, so, so this guy, I don't know if I showed you this one, but this is from Derek Thompson. So it's a new newsletter. It's called The End of Thinking. So doomsday predictions about AI taking our jobs in the future shift focus away from where it belongs. In present reality, the age of screen tech has coincided with declines in literacy, reading, and writing. We are deskilling ourselves. So, if you look at this over here, you can see that adults, the numeracy and the literacy is just going straight down like numeracy, right? That math and the literacy has gone down quite a bit here as well. And then you look at teenagers, it's gone off a cliff, right? Science has gone down, reading, maths has gone down quite a bit. And these are average scores on assessments across different domains in high-income countries, teen and adult scores use different scales. So, it went from like 510 up here for math all the way down to— now it's down to like 480 or so. Wow.

That's crazy. But what's interesting with— keep going.

Change in 12th grade NAEP math scores. So 2005-2013 versus 2013-2024, you can see that in the black line, lower performing students are getting even worse. This is a change in 12th grade math scores, right? And so you can see statistically the lowest performing students are getting a lot better. They used to get better Over time, but they've— they're getting worse progressively over time. And then the last one over here is in the US, the so-called National Report Card published by the NAEP recently found that the average reading scores hit a 32-year low, which is troubling as the data series only goes back 32 years.

So every year it's getting lower.

Yeah. Yeah. We're getting dumber and dumber over time. Yeah.

But dude, you remember the talk you had with me maybe 3, 4 months ago? About AI and kids.

Uh-huh.

Because you asked me like, oh, do your kids use devices? I'm like, not really. Sometimes they do watch TV though.

Yeah.

Um, but not often. And then you said, oh, are you gonna give your kids ChatGPT or any of these tools at their age? You know, at a young age? I was like, no. And you said, that's good. And I said, why? And the reason we're not doing it is we just don't want 'em to rely on devices and we want them to use their brains and play outside like normal kids, like the stuff you and I did when we were kids.

Yeah.

And you had a little bit of a different response on why I shouldn't, but you're like, oh, you're already doing it, so you're good. And you were pulling up some studies and you were pulling up some studies that if you don't give kids AI at an early age and you let 'em learn to think, and then eventually you end up giving 'em AI, they'll perform that much better 'cause they know how to use their brain and think. And then you combine it with AI, they're much better off.

And you did put an asterisk.

You're like, look, this study is really early. There's no way they can, you know, have concrete data on this because AI hasn't been around long enough where they haven't tested it out on kids on sample size of having them go for, you know, 10, 15 years versus ones who are given AI right away. But yeah, you broke this down to me months ago.

I think what we're getting at was you need to have a certain level of base intelligence. So, you know, the right— you have experience and you know the right questions to ask because you start with nothing, you're going to end up with nothing. It's just going to be slop. at the end of the day, right? And that's why the people who learn the most, who pick up the most, the people who are most curious, they're the ones that are not going to be replaced by AI anytime soon.

Yeah.

Yeah.

Yeah. Because when I was breaking this down about my kids' school and you were asking, I was like, yeah, they go to like a hardcore school where it's just like multiplication at a really young age and all this kind of other stuff, reading comprehension. They learn Spanish and Mandarin at kindergarten, right? Like it's pretty hardcore.

A good analogy of this, and my— I had an AI call me out on this.

He's so— so like, imagine this.

Let's say your child is like a toy race car driver, right? That's all they know how to do. But imagine if they become a Formula 1 race driver, meaning their brain is like super firing on all cylinders. That's a lot more helpful if you're in a Formula 1 car. But if you put them in a Formula 1 car right now, they wouldn't know what to do with it.

Yeah. Right?

So they have to train up to get to that point. And I think it's actually, it's going to become, I think many parents are going to, just like you are already, are going to limit the AI use for a very long time, maybe even until like the teens.

I can so relate to this because for my son's birthday, we got him— he loves racing and pedaling super fast, driving little race cars, like the remote controls super fast. So we got him a little buggy, a kid's version, but it goes like 15 miles an hour while he's sitting in and driving it.

Mm-hmm.

All right. And we have a big driveway, so I was just like, cool, go at it, have fun. He did not do well with it. Didn't like it. A little bit too scary. And it hit me that even though he likes speed, he's just not ready for it yet. And let him keep playing with the toys and stuff, and eventually he'll be at a place where he'll want to play with those, like the real cars where you can sit in and drive at a fast speed.

Dude, you'll like this. So that's amazing. So he's going to pick it up. But my— remember I told you about the book The Great CEO Within? This is like a CEO, like Matt Moschari is his name, and he coached like Naval and all these like CEOs, right?

Yeah.

So really good book. So I use this note-taking tool called Granola, and Granola like takes notes in all your meetings, all these things, right? I even do it like if I'm in a meeting with you, I'll take notes, right? Now what it's done is it's added a recipe. So there's a, there's a Coach Me, Matt Moshari, right? And so here's what it said at the very end of it. It was so funny. It's like, you know, the thing that pisses you off the most, Eric, is that you're an F1 driver in a Honda Civic right now. And it's like, you're, you're like, you're up there, like your thinking is there, but like you need to let the organization catch up, right? And I was like, Yeah, that's fair. And then I just sent it to a few of my leaders and then I was just like, guys, we need to build, we need to build the Camry. Then we need to build, eventually we need to build the F1 car, right? Because I can't sit in this Camry forever. The poor.

Yeah. And then go for the Ferrari and then the F1 car.

But it's cool how it breaks it up because it literally sits in all your notes and it knows what you're working on. And it literally like is coaching me saying, hey, work, you're working on this. That's good. Okay, here's a leverage here. You should be working on this instead. I'm like, oh, this is cool because you and I, we sit in meetings, you're on the phone a lot. And guess what? It records your phone calls too.

That's cool. Yeah, I'll have to check it out. It's called Granola.

Yeah, I'll share it with you at lunch. It's— you would love this because there's nothing more you love than like feedback driven by data.

Yes. Yeah.

Here's the problem with most marketing agencies. They don't understand what is happening with the world of AI. With my agency, Single Grain, we think about agentic workflows. We think about creating agents for the future. When we think about SEO, we think about creating different products such as internal linking products or content consolidation products. And on the other side of things, we think about building products such as care. We just think about making marketing easier overall. If you want to stay at the cutting edge of marketing and at the same time leverage the power of AI, go to singlegrain.com and we'll see you on the other side.

But speaking of AI and, you know, crawling, then walking before you run, I had a call right before I came down here. So this person, I don't know them. They said I reached out to like their dad or something like that, had a call with them. They're based somewhere on the East Coast. And they kept texting me and I didn't know who they are. So they're like, can you just gimme 3 minutes? So I answered them, talked to 'em for 3 minutes. They're like, look, I'm bringing, you know, a lot of amazing marketers. We're building something really cool. It's going to be AI for marketing on steroids. Imagine like a Hyros combined with, you know, GoHighLevel and a few other products. And he's like, imagine if you're running an ad campaign, we come up with the best creative, the best copy, we adjust your landing page, everything just works and it continually tweaks and it just makes you more money with great tracking and you don't have to do anything and AI does it all. And the first thing I responded with is, that's great. I hope you can build it and I hope it works. Let me know once it's done, you have some solid case studies. And even if it works like 50% of the time, I'm open to trying it out. Just let me know when you have some case studies and it actually works.

Mm-hmm.

Because right now when it comes to AI, At least in the marketing realm, people believe they can automate everything. And there's so many people trying to go after this. Like I just talked to a person yesterday as well. He worked at Facebook, Microsoft for ages, then Facebook, then Snap. And we were laughing because his wife is one of the most well-known lawyers in the United States and she runs like big law firms. I'm talking about like the head honcho. They don't call him CEO, but I think it's called like managing partner.

Yep.

And she's running the US office for law firms that have, you know, like 1,000+ lawyers just in the United States. They're global.

Yeah.

And I was telling him like, dude, you know, you probably made more from your stock than anything else. And his wife makes a killing. And I was like, Facebook, Microsoft stock. I'm like, if you haven't sold, you made more from that than anything else. And he was laughing. He's like, you're totally right. And people like just look at me as an employee. But long story short, he had a call with me because He's building a tool for AI that just does people's marketing. I'm like, dude, this is really hard to do. You need to focus on a vertical like home services, dental, you know, plastic surgery, whatever it may be. Get as much work as you can for the mundane, boring tasks, and it's more cookie cutter when you're focusing on industry and then replicate it to multiple industries.

Dude, I had a conversation this morning, so, you know, I like to publish the nerdy stuff I'm doing with the agents and agentic workflows, right?

Yeah.

So guy reached out to me, global CMO, and then he's like, hey, do you guys do this for your clients, right? I'm like, we've been piloting for our clients right now. This is the most interesting part. So I was like, okay, your marketing budget, not a big marketing budget, right? But the conversation shifts, 'cause I was like, wait a minute, like if what we build for you replaces like 4 headcount or 8 headcount, he's like, we're much more willing to have that conversation. He's like, we're like owned by private equity and our CEO used to be a CFO. I was like, oh, you guys are very numbers driven, right? So the conversation now is, longer just about the marketing budget, it becomes the headcount budget, right?

Yeah.

And so I think that's going to continue to shift in this way. And most people, I'm sure you guys are having those conversations too. But it's just like, dude, ultimately what it comes down to is not like, oh, this new solution can do all these things. It's what problems are you solving for me? And have you proven you can solve it? So I, yesterday, I don't know if I introduced you to this guy yet, but I met this guy that runs a thing called Stan, stan.store. Have you heard of it?

I think Gary Vee is an equity holder or something, or investor.

I hosted a dinner like 2 weeks ago and the founder John was at the dinner, right? And we've been talking and he posted a video yesterday with him pitching Gary Vee. And so here's the thing. Gary Vee is not going to take a pitch unless you have traction. So his traction was already $30 million, right?

Yeah.

So then Gary Vee got it and they met like he filmed, like he vlogged like the whole thing. And but that's what you're talking about. Like you need to see traction. Like everyone's saying they can build all this stuff right now and a lot of it seems like a lot of slop, right?

Yeah.

So you need to have some case studies, you need to prove that you have some traction, and then you'll be above like 99% of people.

And I look at marketing and AI similar to development and AI. So if you look at what Cursor's done for the whole development space, it hasn't replaced developers. You know, people talk about Lovable, great product, but I don't know anyone who's building a full-blown product with Lovable without any engineers, and then you can go and just market it and make a ton of cash. I'm not saying we won't get closer to that over time, but I think We're really far away from it.

Yeah.

And you're still going to need humans. Yeah. And Cursor just makes engineers more efficient. So I don't know what we pay an engineer, but let's just use simple numbers like $100 grand a year. And if your engineers become 40% more efficient, let's call it $40,000. Yeah. And Cursor costs— I don't know how much it costs, but let's go with $5 grand. It's a really simple calculus. Yeah. And we pay for Cursor for every single one of our engineers. I don't know what it costs, but I know we save more.

Yeah.

And I think that's where marketing is going to go. I don't see people building something in AI that just replaces everything, because if you do what everyone else is doing, marketing is not that effective.

Yeah, you have no alpha. That's what it is, right? If you're constantly at the frontier and then everyone's just copying you, like you need to be ahead, right? Otherwise the alpha goes down. But by the way, you get this. So because we hadn't recorded for 2 weeks, this is the first time we haven't recorded in 2 weeks in a while. It's a nice break. But so that AI founders dinner, so we were talking and then one person at the dinner was like, dude, with 2 engineers now, we have built what used to take us, you know, 6 months to do, but used to take 30 engineers to do. 30 to 2, right? And someone else said, we had 50 people and then now it's like 5 people doing the work, right? And these are all engineering teams. And so that is—

I haven't seen that big of a change for most well-run companies, like 30 to 2. I'm assuming they're really inefficient when they had 30 engineers if they only need 2 now, like something's wrong with how they're structured before.

Yeah, I don't know about that. One I'm pretty close to, I'll just— I'll leave it at that. And his engineers were pretty good. He's just saying that with all the stuff that Not even just Cursor with everything included, it's down to like 2 can do the work of 30. And so that's— I think we're going to— by the way, you just mentioned that lovable, you think we're pretty far off. How far off do you think we are from like you and me being able to deploy something and put it out to production?

That I don't think we're far off. I think we're there today. Okay. I don't think we're there from a standpoint where you can produce something from a lovable, monetize it.

Yeah.

And add in a lot of customization and features because there's too many edge cases for software where you're building a lot of different stuff for different people, and it's just too much customization that you need, and a lovable is not perfect for that. I think a lovable is great for you to get something off the ground.

Yeah.

I think you're going to still need engineers. We haven't seen too many, like, I know a lot of people who are in the Bay that have raised a ton of money for their startups, and if you look at even the Curses of the world, look how big their engineering teams are. I haven't seen the 30 to 2s in most cases. Typically when we see that, there's something wrong with how they're running a 30-person team. But I have seen the, you know, 6 engineers go down to 2, or I've seen the, you know, 30 engineering teams go down to 10.

Still huge. 3x.

Yeah, still huge.

Right.

And in many cases we're seeing like, I would say the average, and this is just a ballpark number, I would say people are getting like a 30 to 50% reduction.

Yeah, I think it just, it depends also if you're like an early stage startup, it's a different story. But I think, look, What I'll say with Replit now, Replit's gotten a lot better and it's got like an autonomous coder, right? So I think probably a year or two, I think someone like you and I will be able to push things out more. But to your point, there's still going to be a lot of bugs and edge cases that we don't know how to fix. And that's where we need developers to solve the last 20%, the last mile. And I think there's going to be a lot of shops for solving that last mile.

Dude, when I was talking to my buddy Ash, he's the one who's building—

From Pokémon?

No, Ash. Ash is the guy who's building the marketing automation or using AI to market different verticals, right? And when I was talking to Ash and we were talking about, you know, a popular, I think it was VO3.

Yeah.

And I think he was asking for something with a white background and he's like, they couldn't get the damn white background right. And the point I'm getting at is like when you use AI, like we're coding up the new Answer the Public and we use Cursor, but there's just so many, there's still bugs that the AI just doesn't fix. And they got to get some of these basic things down. And these nuances may seem like, oh, this is easy. They'll fix it in a month or 2 months. It's not that easy for them to fix these things.

Yeah. And the problem with the bugs is if you don't catch them, they compound in your product and then you have to come back and do it. That's, that's, that's way more work.

So dude, when we were using Cursor to help our engineers code up Answer the Public, we're releasing a new version of a wheel because the way we report data is in a, in a wheel format and it kept just breaking everything and it couldn't get it right and it just kept making it worse and worse. And we, we eventually got it to be a lot less buggy, but it was done by a human.

Yeah.

Again, I don't think— I'm not saying that AI is not valuable. I'm not saying humans are more valuable. I'm saying I think the winning combination is A mixture of both.

Yep. All right, so that's it for today, guys, and we'll see you tomorrow.