Have you ever been frustrated by the fact that AI is so good at some things and so ridiculously bad at what seems to be like a simple task? I know it's frustrating when it can do advanced math and calculus, and people are bragging about the, like, fields metal math that it can do. And yet, it it still can't freaking count the amount of survey responses I just put into it for some reason, let alone calculate what's in those survey responses. If you've ever been frustrated about that, that's what I wanna make this episode about. Not only why it's happening, what's causing it, but what to do about it and how you can leverage it to get the most out of it this year as an AI driven marketer. Welcome back to the show. My name is Dan Sanchez. My friends call me DanChez. And today, I wanna help relieve some of that frustration because we've all been caught into it.

And I wanna talk about the paradox that's at the root that's at the center of this thing that's causing so much frustration when it comes to working with AI that makes it just seem so dumb yet so smart sometimes, and it's a term they're calling Moravec's Paradox. Yes, Moravec's Paradox. So I'm going to jump into a deck real quick to give an illustration of what's going on. And if you're following along in an audio podcast, just know that I'm gonna be spelling out in detail all the visuals that are coming along the way. So Moravec's paradox is a concept to explain why some things, some simple things are actually incredibly advanced for AI and why advanced things are actually incredibly simple for AI. The way we look at it is intelligence is linear, right? It's like this bright shining orb. If you're if you can't see it, I have like this this basic ball of light that kind of dissipates the farther it gets out on the deck in the middle I have the word intelligence. We think about intelligence this way.

So if you can gradually learn how to do math, you might learn how to count first and then work your way to something like advanced math, right? I know big steps for a lot of small baby steps in between those things. And it's it's reasonable to think like, well, if I could do advanced math out on the fringe of the intelligence, right? Because I put it right here on like the the edge of where it's the the light is dissipating, then counting is towards the center, like I should be able to get that. It's it's kinda counterintuitive that AI can be fantastic at advanced math and algebra and calculus, but really struggle with something basic like counting. Yet we find that over and over the place. So if I wanna adjust this graphic, I actually changed it to look more like this blob right here where it looks like an ink splat on a piece of paper. Right? And I have intelligence in the middle again and the one drop of the ink splat that's way out there, you know, advanced math sits on that. And then one of the gaps that are in between the splatters, you know, that the ink isn't covering is the word count. Because that's how it is with AI.

It has all these gaps in its knowledge, and it seems to shoot way out there on these advanced topics, yet struggles to do some of the more basic things. And it does this in multiple ways in lots of different places, like it can make photorealistic images, yet it struggles to, you know, have the right amount of fingers. Like how can it be photorealistic with the face and the hair and the backdrop, and yet that person's got six fingers? How is that possible? Well, I'm not going to break down like the tech and the science. You can go and you like just do a search for more RFX Paradox on YouTube, and you will have some great explainers breaking out the details. I find that knowing the intricacies of why it happens is much less important to marketers than the fact that it happens, and it's happening all over the place. For marketers, what you really have to pay attention to is where it's happening and how it's happening. For example, reasoning was something that AI struggled with just recently, right? Until September came around where OpenAI launched its O1 reasoning model, reasoning was tough. You had to essentially do the reasoning for AI.

It could do a lot of advanced things, but if you wanted it to do a project that included pretty important tasks in between, you'd have to just break it out and ask it to do each task and then guide it through the whole project. But AI is always improving, and that's the thing that you need to realize. And the thing that I'm paying attention to the most as an AI marketer is that these things are improving because now this is this is probably an accurate picture of what AI looked like just six months ago. But today, it probably looks a little bit more like this. Now it's getting fingers right almost all the time. It can actually reason pretty well. They just launched, they went from o one preview all the way up to o three, mini is where they're at right now, and the other ones are probably coming out shortly. And now they can do it can do basic counting.