Model Intelligence, Critical Thinking, and AI Regulation

Welcome to wake up with AI, the podcast where human powered meets AI assisted. Join your hosts, Chris Carillon, Niko Lofakas, and George b Thomas as we dive deep into the world of artificial intelligence. From the latest AI news to cutting edge tools and skill sets, we are here to help business owners, marketers, and everyday individuals unlock their full potential with the power of AI. Let's get started.

Good morning. It is time to wake up with AI. We're here with Niko and George. How are you fellas doing today?

Doing good. Feeling good. Ready for another day.

Trying to see if I can write an entire ebook during this 30 minute episode. We'll see if it's possible. But I'm doing great. Doing great.

Man, you wanna talk about somebody who can be present and get stuff done at the same time. I've never seen anything like Georgie Thomas when it comes to that.

It's the acquired taste and skill probably for for those that know.

Figured out augmentation. Like, he's already figured it out. He doesn't need gadgets and stuff on his head.

Right?

He's just got his head. That's all he needs. Yeah. Oh, what are we talking today, Niko?

Always something new. What's funny about that statement is there was a Bloomberg video that, just, I shouldn't say surfaced or anything, but a Bloomberg video that was put out, a couple of them, from the latest, I think it was Fintech Conference that they were having. And the CFO for OpenAI was there, and they grabbed her for about 10 or 15 minutes asking her about what what goes on behind the scenes. And even she was able to say that, you know, every week, she is amazed at the progress that they're making and what they bring to the table. They have these progress meetings on a weekly basis, and on a weekly basis, she's blown away by what what they're doing.

So, you know, on a on a daily basis, it's it's even harder to keep up with. And, you know, again, I I just wanna point out stuff that I was talking to Claude about. Even though the release of GPT publicly was November of 22, the training for GPT 4 was completed the initial training was completed in August of 22. Everybody's been taking 2 years to catch up to a model that had already been finished before its its predecessor. So, you know, even now, right, everybody is really just weighing themselves against 4, and 4 o is part of the 4, you know, base model.

So it's to say that, like, 4 o is, not an extension or, like, to say that's a different model or something like that, It's not really fair per se. It's the same thing as, like, quad 3 and then 3.5. It's not a different foundational model. It's the same foundational model, just an extension. That is just an aside.

It's just something that was on my mind before we ended up getting on getting on here, and, and I couldn't really Can you shift off of it.

Can you just define real quick? Because I I think this is important because we're gonna keep hearing this word, like, catch up. Right? Like, when you say catching up, like, can you just talk about a few things that what do you mean when you say, like, this model's up here and everybody else is behind here? What does catching up look like?

So let me see if I can find one real quick. But catching up is essentially the phrase I use when it comes to other AI model companies that are trying to do just that. Right? They're trying to get to the same level as OpenAI. They're trying to get to the same level as Anthropic in terms of what the model's intelligence level is.

So there are a plethora of different tests that get run to determine what those levels of intelligence are. And some of them are for math, some for English, some for biology, all the different studies. And those tests help to determine, okay, how far along are we with the level of intelligence of this model? And so SONNET 3.5, the the latest version from Anthropic, is currently the the most intelligent model out there And when tested and this is to say that hold on. I wanna see if Claude can actually list or draw out the model iteration progression to say that, like, a one preview when they talked about it and they talked about, like, how smart it was, what the intelligence capability was.

They were saying that it was smarter than 3 different schools of PhDs. I think it was I wanna say it was biology, history, and either physics or just like raw science. I I can't remember exactly which. And then when it came to mathematics, it was like as good as, I wanna say, the top college level competitors in the US, something like that. So it's definitely getting there.

To say that 35 is smarter than that, it's like okay well basically it's like saying that 35 Sonnet is probably smarter than a handful of PhDs. I'm not sure exactly which types, but definitely able to say at least that. Let's see. The tests let me see. The the tests that are used for AI models are basically stress testing, system integrations.

They do input data testing, different security testing. And then there's also tests for, like, intelligence because there's also, like, the safety aspect of it. And they were just talking to Geoffrey Hinton and asking him, like, hey, how much, you know, what what should be done in terms of we were just talking about this, like, before we got on. What should be done in terms of regulation? Like, how could you possibly, like, slow things down or keep a, like, keep a a hold on what's going on?

And so they they basically talked about the fact that, you know, the only way to really pull pull back on any of this is to control the level of compute, to control how powerful they are capable of making the model. And while I I somewhat agree with that, at the same time, I really don't because of the fact that we're getting to a point where even with the same level of compute, we're able to create better models. And that's just a matter of, you know, the software becoming more efficient. Right? That's how we got here in the 1st place.

We didn't get here because the hardware necessarily I mean, yeah, we had the compute hardware to help, but it was a matter of the software. Right? Even with like, if you strip away the the hardware aspect of it, there is a software aspect. And so we're at a level now where the software has absolutely eclipsed the exponent rate of where hardware is at. Even hardware has managed to break Moore's Law in order to try to keep up with what's going on.

So, yeah, at this point, it's a matter of putting a choke hold on compute. And even when you do that, when you think about, like, where things are headed with where the models are going, it is definitely a case where you're going to 0 sum intelligence. Right? Like, we've already if you think about it, we've 0 summed manual labor to a pretty large degree, especially on an industrial scale. And so I'm not entirely sure what it looks like.

I didn't really think this conversation was gonna go down this road, by the way. Yeah. We're we're headed towards a a society where we're going to have a near zero sum intelligence factor as well. It's actually going to be zero sum because it's going to be very quick that, you know, Ray Kurzweil was saying that we're gonna have artificial general intelligence at least, you know, at at I should say at the latest, 2029. So certainly sooner than that.

Dario was saying we we will probably have it in 2026. I would guesstimate that we will have it early next year. It's just that we, external users, won't see it for a little bit. But internally, I can guarantee you they'll have it by next year. If they had 4 completed with training before 3 was released publicly, you can guarantee that internally, we're sitting here waiting for 5.

I mean, just assume that they already have 7 up and running if we continue the iteration numbers and we don't go with any of the name stuff. So This

is where people need to wake up. Okay? Okay? Because you throw all these numbers out there. None of that matters because, like, it's gonna be smarter than us.

At some point, we need to accept that. And this whole concept of controlling and slowing down makes zero sense because almost all the reasons to do it are, like, manufactured risk. I feel in a lot of ways. And that's where, like, could we instead of putting energy there, like, locking down risk assessment, risk management, all that stuff, like, how much energy could be put towards positive training about why it's okay that we have something smarter than us? Because my wife shared a an Instagram post yesterday that was like this minute and a half clip of some, like, you know, NBC News or 60 minutes interview clip.

Right? Where it might have been right. Like, I don't know. You look familiar. I I hadn't seen the clip prior to that.

But it, like, the headline was like like scary AI, smarter than humans. And the whole minute and a half had this, like, background music that was, like, ominous and crescendoing into, like, like, terrifying, like, trying to create this and these back and forths. I mean, press a video clip. Like, first of all, like it and this is the concern. Right?

It's going to have the desired effect of that clip where it's the whole exchanges. So you're telling me humans will not be the smartest thing on the planet anymore? And the dude's like, yeah. So AI is gonna be smarter than us? And it's like that kind of clip is one of the number one reasons why I'm here right now on this show because we can only combat that with, like, positive, like, this is why it's a good thing.

We shouldn't even be asking that question anymore. Like, why is that a valuable question to ask? Are we the smartest on the planet anymore? Like, why are we asking that question?

There's really only one question. I shouldn't say there's only one, but there's definitely a question that I haven't asked Claude yet, which I find is probably going to be the most interesting, here. I find I found a, scale chart, by the way. So this is basically I it's not the the prettiest scale chart you've ever seen, but shows the timeline at the bottom measured by years and then shows the benchmark scores on the left hand side. And currently we're just above, like I would say we're at like 82 or 83% benchmark in terms of like how close to human general intelligence are we.

We're 83% to the mark. Once it gets a 100, that means literally as smart as the smartest human on the the average smartest human on the planet And then over that obviously equals smarter than the smartest human on the planet. You know, that's assuming again that you step outside the box of, oh, yeah. Like, you know, it's not really smart because it can't do x y z. It can't perform x y z.

If you're if you're taking it out of the context of could you answer questions that it couldn't answer, it's definitely smarter than you.

I feel like this is a good lean into the fact that, like, critical thinking becomes very important in the very near future.

Like, just a so just a quick audio illustration, I guess, or to tell people what they're what others are looking at. You get GPT 4 releasing somewhere around the beginning of 23, just between just before March. Claude the first iteration releases in April of 23. In July of 20 and so at that point, gbt 4 was already nearly 70%. Claude 1 was at just above 50.

Let's call it 54%. In July of 23, Claude 2 comes out, second iteration, so then we've it goes up to about 62%, 61, 62%. All the way to October, let's say December of 23, then GPT 4 Turbo comes out. So we're really just talking about a faster model. Doesn't necessarily it's not more capable necessarily.

It's a little bit smarter but not like insanely smarter. A little bit smarter saying it goes from like 68% to about 74%. Gemini Ultra comes out right about the same time. That flat lines around 70%. So that at least that's to say if we're looking at it this way, Google's first attempt beat OpenAI's 4th attempt.

Again, if you wanna look at model number release, but even still, you're saying it took Gemini up to a year, almost 2 years, yeah, almost a full year after GPT's initial release to catch up to GPT 4. So then Gemini 1.5 comes out. So you look at GPT 4 to turbo which is basically 4 to 4.5 and Ford is at like 68%, GPT 4 turbo is about 74%. Gemini 1 to 1.5 Pro goes from 70 to about 71%. K.

And it took them about 2 months to make that happen. In March, just about like end of March of 24, we see quad 3 Opus. K. And that smacks in around 75%. I'll say 75% to goal.

That's what I like. Shortly after that, April 24, you get GPT 4 turbo upgrade. Literally just a tweak to that puts it at 75%. So or just like 76%, let's say, so it's higher than Quad 3 Opus at that point. Gemini 1.5 Pro gets a tweak, ends up at like 77% and then right around June, GPT 4 0 drops and that hits the 80% mark on the nose.

And then I shouldn't say 35 Sonic new but 35 Sonic dropped just around the beginning of July and that hit right around the 82, let's say 82, 83 percent mark. So from February of 23 to July of 24 has been a gain from 68% to 82%.

These are great stats. And by the way, it shows how fast things are growing and learning and becoming better, which is a beautiful visual. But, Nico, when you think of the individual human, the business owner, the marketers, what does this mean to them in your mind?

What it should mean to marketers and business owners is that greater strategy isn't coming, isn't just coming down the road. A lot of it is already here. It's kind of based on how you leverage these things. I think that Mogadot put it best. He said that currently, whatever his IQ level is, he's able to go home and plug into an additional 10 to 15 IQ by using whatever language model he's using.

That's really how you have to look at this. You for marketers and for business owners, if you're struggling with your go to market strategy, if you're struggling even if you're struggling in your day to day as a marketer, you're struggling with organizational skill, you're struggling with time management, you're struggling with client communication, you're struggling with data high hygiene process, processes of any kind, blueprints of any kind, additional skill sets related to, you know, custom coded actions, related to writing emails, designing emails, ideation. These are all problems that we have on a daily basis that cost us tons and tons of time. So for marketers and business owners, what I have to say to you is the one thing that you could not possibly get back from any other method is time. And it's the most expensive thing and it's the most precious thing that you have.

So if you want more of it, this is how you get there. And if you wanna accelerate what it is that you're doing, this is how you get there. That you can't really like, it's it's just you can't ignore this stuff, and it's only going to make you better at what it is that you do, just like all the other tools that we use.

It's funny because you're making me think about what happened in my own house last night, and I have told the Internet many times that I am more of a listener or a viewer, YouTube, audiobooks than a reader. My wife and I, we are watching TV together, and I got this crazy, like, idea. So I pulled out my phone, and the idea started with, I'm wondering how the superhuman framework would help make better marketers. And it gave me some information. I was reading it.

I said, I wonder how the superhuman framework would make better sales reps. And I kept asking it these, like, things to align with these different humans. What about an HR professional? Right? And so all of a sudden, I generated the idea, the outline for, like, this first ebook.

I'm reading the intro to which then I had to read chapter 1, to which then I had to read chapter 2. And my wife is like, what are you doing? And I'm like, reading. And she's I mean, I literally I had to go get my glasses, bro. And because I'm reading.

Because I was so riveted by the information that was being provided to me based off of an IED. How does this thing make this type of person better? And so I was kind of joking, but wasn't joking of the see if I can generate an ebook during this episode. But here's what I'm going to is, like, it's ideation and it's critical thinking, and it's looking how vectors can shoot and align and then, like, the rate in which you can build. Because here's the thing.

Last night and this morning, Noah was putting some final touches and getting in at a dock that it can go through processes, an ebook that is beyond business as usual, the superhuman blueprint for transformational growth. Now here's another one is that we're building the HR superhuman handbook, solving people and culture challenges with purpose and passion. Now the next one that came to mind that we're doing a ebook on the superhuman leadership playbook, building resilient teams and purpose driven culture. But this is all out of an idea sitting on my couch with some critical thinking. And hey, ladies and gentlemen, That's today's AI skill that pays the bills.

It's all about critical thinking with AI insights helping you along the way. Now we all know AI can analyze massive amounts of data in seconds. But here's the thing. Not all data tells the whole story. When it comes to making decisions, it's crucial to interpret AI insights critically, question assumptions, and ensure those insights align with your goals, the goals that you as a human have.

Imagine you're a marketer running a campaign and AI data shows that one ad is performing better than the other. But before pulling the plug on the second ad, critical thinking kicks in. Why is it underperforming? Could it be targeting the wrong audience? Is there a timing issue?

See, AI gives you the numbers, but it's up to you to dig deeper, question the why, and make a call that's backed by both data and context. Here's where critical thinking with AI Insights really shines. AI can suggest a course of action, but it's your job to assess its relevance and accuracy. I love asking the question, how could I make this better? And then saying do 3, 5, and 7.

Not all of them, but the ones that I want because of the insights that I'm getting. Let's say your AI tool flags a group of customers as high risk based on their purchasing behavior. Critical thinking means you're not just accepting that label. You're validating it, maybe cross checking with other data sources and deciding if or how to act on that insight. By pairing AI insights with your judgment, you're ensuring that your decisions are well rounded and truly aligned with business objectives.

After all, AI is a tool right now, but you're the one making the strategic choices. Let's hit the highlights. Question assumptions. AI can provide insights, but it's up to you to question the data, ensuring it's valid and relevant to the context. Validate accuracy.

Cross check AI recommendations with other data points to confirm accuracy before making decisions, and combine AI with human judgment. Use AI insights as one part of the equation, blending them with your experience and your knowledge for balanced effective choices. Ladies and gentlemen, that's today's AI skills that pay the bills.

As you're describing stuff yesterday and now today, so much doesn't get done or hasn't gotten done. Like, all the stuff that you created and the whole framework in general, like, without any of these tools, maybe it's like a year.

Oh, longer.

Not that because it would, like, physically take you that long, but because you've got other stuff going on. Right? Yeah. Yeah. And that's the kind of stuff.

Like, how much has not been created in people's minds? I hang so often we can focus or assume, you know, ignorance or they don't know how to do it or they're choosing not to do it. Right? When in reality, they know how to take care of customers. They know how to do things the right way.

They just have not been allowed the time or the space or whatever to be human. Right? And and get it done the way that it needs to get done. And that's one of the biggest unlocks here. I'll just say that we need to get real good at kind of balancing this the b to b and business advantages with the b to c.

And, like, the consumers didn't care about Google algorithms. They just enjoyed when it got better. Right? Like, we need to start being able to say something like, okay. Like, instead of worrying about AI being smarter than you, like, next year, your phone's gonna be, like, make your life, like, 50% better.

Yeah.

Right? Like, and that's gonna be cool, and everybody's gonna love that. And it's not gonna matter how much it's gonna cost. Your phone might start costing $3,000. You're still gonna pay for

it. Like, and Easy monthly payments.

That's the kind of stuff. Right? Well, it'll they'll they'll just give it to you.

Apple's gonna be something today. If you bought your new Apple whatever number we're on now 16, I think. 16? Yeah. Okay.

AI features for that are supposed to supposed to drop today. That's what they announced. We'll see if it actually does. Everybody knows how launches go.

Sometimes they go. Sometimes they don't.

So embracing this reality. Like, let's be thankful that we have something smarter than us because we don't always act so smart sometimes. Well Like So

And it's funny, Chris, when I hear you kinda waxing poetic the last couple minutes, it's like, I wish people would just honestly Wake up. And realize, like, you can eradicate the hurdles. You can remove the excuses. The days of, like, oh, I'd like to, but I don't know how to, or I'd like to, but I don't have the time, or I'd like, if you have copilot, an assistant, a second brain, whatever words you want to use here, you can literally build and create the thing that you wanna build and create and bring to the world and go to sleep feeling like I've done something amazing. And then the next morning, wake up with AI.

That's a wrap for this episode of wake up with AI. We hope that you feel a little more inspired, a little more informed, and a whole lot more excited about how AI can augment your life and business. Always remember that this journey is just the beginning and that we are right here with you every step of the way. If you love today's episode, don't forget to subscribe, share, and leave a review. You can also connect with us on social media to stay updated with all things AI.

Until next time. Stay curious, stay empowered, and wake up with AI.