n unless your personal version of AI decided, based on what you told it directly or what it knew about you, that it would always be scanning for new products and if it found anything that was right on point for you it would show it to you. But it wouldn't show it to you while you're on task. In other words it wouldn't interrupt you while you're reading something. It would simply know that a good time to talk to you or even maybe just make a little shopping list for you that you don't even see unless you tap on it.

I don't see any way that advertising works as a revenue model and as a business model in ten years. How many would agree with the following statement? In ten years the advertising model will be completely dead. In ten years. All right. I think I can convince you in ten years it'll be dead. How about five? Will anybody go with me to five? Because the thing that'll kill it is likely to be some AI-related architecture. I think five. Yep. Now in five I don't mean every single ad model would be dead. I mean in five you could maybe imagine that X would be ad-model-free while still giving you all the advertisements you might actually say. Let me ask you this. How much would you pay to never see another advertisement again unless you wanted it and you only see it when you want to? How much would you pay to remove just advertising from social media? Because we're getting pretty close to that.

And here's how. If Musk succeeds in making X the place you go and just live because you can make payments there, you can see the news there, you can comment there, basically you could just live there, see all your videos and everything, then suppose he makes it ad-free. Imagine going to an environment that's ad-free and you know playing in your playground all day. How would you feel the first time you turn on cable news and it goes to a commercial in 12 minutes and it stays there for another 12 minutes? It seems like you will be so done with the ad model once you buy out. And I'll tell you I have some experience with this because I've bought myself out at the YouTube advertisements. Meaning that if you pay extra, I forget what they call it but whatever the business model is on YouTube, you pay extra to see no advertisements. The YouTube Premium. Somebody says. The nature of your experience without advertisement is so transformative that it makes YouTube in my opinion the premier entertainment vehicle in the world by ten, probably ten times more entertaining than the next most entertaining thing. This long form, just talking about long form.

So anyway I think Musk might have more of an instinct about advertising being dead than you do. And just speculating that I think he's probably thinking ahead to the death of advertising and that he probably didn't feel that it was an existential threat to X if he just killed it early. In fact it might accelerate the urgency to get past the model which is going to happen anyway. So he might actually come out ahead. Because you could imagine what would happen to the team at X. Imagine you're working at X. You've got some people are trying to work on the non-advertising model and some are working on the advertising model. Do you think when he blew up the advertising model that that changed the energy? Probably. Probably the people figuring out how to make AI work and all the alternatives probably got real busy. Probably also the advertising team got real focused on smaller businesses, the ones that are more likely to be advertising. So I'm not sure that that was the bad thing that you think it was. You know, maybe it was. There's no way to read his mind to know what he was thinking. But I'm going to go on record as saying I think that Musk will probably say something that suggests the advertising model is not forever. Expect to see that.

All right. Climate activists are concerned that AI would be another risk to climate in two different ways. AI might be used to spread climate disinformation. Well my God. Yeah I'd hate for there to be any climate disinformation out there because you know what? The climate information that we have now is so good that we all have the same opinion about the risk. Am I right? Yeah. Because it would be a real problem if suddenly our good climate information turned into unreliable stuff because of AI. I mean that's reason enough to stop AI. Let's slow that down. We don't want any unreliable information about the climate out there. But also the AI models are huge data center users and so they use a ton of electricity. So that's a good point. Or is it?

Here are a few things that AI might tell you about climate models. Do you know how I know that it might tell you this? I asked it. So I asked AI, hey I got a few specific questions about climate models because I would like to know how reliable they are. So here are some things that, there's no opinion in what I give you now, so these are just facts. I asked how many climate models are there? Do you know how many climate models are there? Thousands? Dozens? Hundreds? 25? 60? 13? Now don't you think that if the entire strategy of climate change is based on the models, and it is, don't you think you should know how many models there are? Would it affect your opinion of the reliability of models if you knew how many there were?

Let's take two extremes. Let's say there had never been more than five models and one of them got thrown away over the years because it just didn't work. But imagine if the other four had been doing a pretty good job for like 30 years. That would tell you something important, wouldn't it? If you had three different models and they actually predicted pretty well the actual temperature for 30 years, my God that would be amazing and I would be quite inclined to believe a model like that. Do we have that? No, nothing like that. That would be very, very believable.

All right. Here's another question. Oh so the answer is that the IPCC averages 29 models. Is that the answer to how many models there are? If the most official body takes the average of 29 models, so would it be fair to say there are about 29 models? Does it follow that if they only use 29, well they can't be that much more than 29 because the whole reason that you take an average is because you're not sure which ones are right. So if you had hundreds of them and you're not sure they're right you would take the average of all the hundreds. So therefore if they're only taking an average of 29 doesn't that suggest, since they're not being super picky about which, you know because they don't know which ones are right that's why they take an average, doesn't it suggest that the total number might be close to 29? Because why would they throw out any models? What would be the reasoning for discarding any models at all if they don't know which ones are right? And they don't know which ones are right. That's why you average. Because you don't know which one's right.

Well the answer is it's probably hundreds of models but they average 29. Huh. Huh.

Now here's another question for you. How many models are retired every year? Do you know what that means? It means there used to be a model that they were tracking but it became so bad at tracking over time that they said you know we can't even tweak it. We're just going to throw this away. How many? So out of the number of models that you don't even know how many there are, maybe hundreds, maybe 29 good ones according to the IPCC, how many of them change out every year so that the 29 you're looking at are not exactly the same 29 as five years ago? Is it half of them change out? Because if half of them changed out wouldn't you say to yourself well that doesn't sound very good at all. That sounds like guessing if half of them are replaced every year. Or is it one out of 29 wouldn't be bad. But how about the fact that you don't know the answer to that question? Because if you don't know how many get changed out every year you don't know anything.

Do you understand that if they're tracking 29-ish models but you don't know if they're replacing models as they go, you literally don't know anything. Nothing can be deduced from tracking a changing basket of anything. Imagine if this had been the stock market. They say we're going to track the S&P 500. Oh but wait the S&P 500, the biggest 500 companies, it changes all the time. Things are coming in and coming out. So at the end what did you really check? It wasn't 500 companies because at the end it might be 350 that are left depending on how long you're tracking. So if you're not tracking the same thing you're not tracking anything.

All right. Here's another question. How many variables are in a model? How many variables are in a typical model? If there were three variables I would say to myself, hm, three variables. There's at least a fighting chance that you could wrestle three variables into a useful model because three is not too many. But suppose you had five. You know five is not just two more variables. When you go from three variables to five your possibilities explode. So how many variables do you think are the essential variables? They even have a name for the essential variables because they're infinite variables but they're not all meaningful. How many are the essential ones? Well if you check online you'll find out four. The four variables. So I said oh that's not bad. Really there are just four variables. And then I looked a little deeper. No there are four categories of variables. There are four categories. The total number of variables is closer to 54 essential variables. 54 variables.

Now let me ask you this. What if there were one more variable that they found tomorrow? Because I see stories about this all the time. Oh we found another variable we didn't know about. If you were to find the 55th variable and you added it to the 54, what does your intuition tell you? Does your intuition tell you well it's only one more variable. You've already got 54 variables so if you're only going to add one more how much difference could it make? Am I right? It's only one more variable. You've already got 54. Couldn't make that much difference, right? No. One variable can completely change the outcome from up to down. If you've never done modeling you wouldn't know that.

How many of the 54 variables, if you got any one of them wrong, just one of them, how many of the 54 could completely reverse the direction from the temperature going up to the temperature going down in the future? How many of the 54 if you just had one of them a little bit wrong? How could it change? A lot of them probably. Probably a number of the 54 if you got them a little bit wrong and you iterate that over time because small errors exaggerate over time with models. Probably there are 54 ways that that could go wrong. In other words you'd have to really have a solid lock handle on all 54 variables or else it wouldn't necessarily be anything. It would just be I don't know, could be anything. That would be the only way to analyze 54 variables.

Now let me ask you this. How many models that purport to forecast the future, and of course there are financial models, economic models, so a lot of people have models that try to forecast the future, how many of them have something in the order of 54 variables and a track record of being correct? Now I'm laughing because everybody who's been involved in any kind of forecasting or modeling knows, and I'm going to say as clearly as possible, they know that this isn't real. This is not real. There's nobody tracking 54 variables and maybe hundreds of models but some humans are picking the best 29 out of the hundreds. Because what do you know? How they pick the 29 models out of the hundreds of models? Is it because they're the most accurate? I doubt it. Or is it the ones that support the narrative the best? What do you think? What do you think? Yeah. Use your understanding of everything you've ever seen in the real world and you tell me that there are hundreds of models to choose from. They're only going to choose 29. Do you think they're going to choose 29 models that when you put them together they don't support the narrative? There isn't th