Meredith's Husband 0:33

All right, going to talk today about how you can switch from one AI model to another, like from ChatGPT to cloud, or vice versa, or from Gemini to Cloud, or any if you're switching between any of these, or if you're thinking or you want to switch between any of these. And the reason that you might want to do this, this has been in the news, lots of people, there's been a big kind of shift very recently away from ChatGPT. In fact, just in the last, I think, two or three days, the Claude mobile app had more downloads than the ChatGPT mobile app for the first time ever. So there is a noticeable shift away from ChatGPT for a number of reasons. Like I said, uh, number one, they've been in the news. I don't know if you've heard the military conversations going on, but uh most of the models, including Claude, Google Gemini, and Perplexity, have basically all told the Pentagon to go fuck themselves because the Pentagon has said to each of these models, well, if you want to keep your government contracts, then you're gonna have to let us, the Department of Defense, do whatever we want. And that includes creating weapons, and that includes creating systems for mass surveillance. And three of the most popular models, uh, Claude, Google, and Perplexity, like I said, all said, no, we're not doing that. And so now they are at risk of losing their government contracts. And all these models have some government contracts. They do lots of different things. Some of them are military, a lot of them are not military-based. But what the government has said is that if you don't let us do whatever we want with your technology, we're going to cancel all your contracts. And as of today, only Chat GPT, only OpenAI has gone ahead and said, okay, yeah, you can do whatever you want. I mean, publicly they said no, that's against our policies, but then immediately turned right around and just basically did allow that. Okay, so that's one reason. Plus, a lot of people are just kind of sick of the, I'll say, the personality of ChatGPT and how it does things over and over and you can't get it to stop. I've heard a lot of frustrations from a lot of different people about those sorts of things. That's another reason. And then something else, also along the lines of ethics, a few episodes ago, uh Meredith and I talked about the differences, the strengths and weaknesses between specifically Claude and ChatGPT, but really all four of them. And I followed that episode up with an example. There was a particular design, Meredith had a design request, and she used Claude, and she also used ChatGPT for the same request. And the resulting output was there was a night and day difference. And ChatGPT did a much, much, much better job. I described it, I think, as it reminded me of the difference between hiring a professional illustrator and giving something to like a second grader and asking them to do the same thing. The difference was that significant. Now, after that, I got the question of like, well, how is that? I mean, these technologies are supposedly kind of similar. Each company kind of has the same abilities. How is it that that in this particular instance, one of them is so good and another one is so bad with the same request? I got the question, just why is that? And the answer is the training. It's all it all comes down to what these companies give to their models to train it. So in this example, we can imagine OpenAI gives to their model ChatGPT a whole bunch of artist work. And they are basically teaching their model, hey, this is how you draw art, this is how you do this stuff. And so ChatGPT learns to basically be an artist by looking at lots and lots and lots of artists' work. Okay, the the problem there is that those artists didn't get paid. Okay. Their work was used to essentially create a model that is going to replace them. Okay. And that's just not very cool. There were a lot of lawsuits, and there still are a lot of lawsuits, about how ChatGPT goes out and takes data and uses it and then takes business away from wherever they got that data. There was lawsuits between the New York Times that were very well known. Essentially, ChatGPT had access to New York Times articles, which are normally behind a paywall, and it just made them public. Okay, well, New York Times wasn't very happy about that, so they sued. It's that sort of thing. Another one that was in the news was the Scarlett Johansson issue, where ChatGPT developed a voice and they modeled it after Scarlett Johansson's voice. Scarlett Johansson said, no, I don't want you doing that. ChatGPT said, okay, but then they went ahead and did it anyway. So essentially what they did here is along the same lines. They basically, like I said, looked at a bunch of artists' work, taught ChatGPT, this is how you do art, and now it can do art. On the other hand, it seems like Anthropic, with their Claude model, did not do that. Okay, so in that episode, I was saying, oh, look at how much better ChatGPT is at creating designs, but there is absolutely a flip side to that in that yes, ChatGPT is better because they basically took artists' IP, their intellectual property. Okay, so those are the reasons why you might want to switch, at least some of them. Right now, there are probably others and there will be more. In fact, I also mentioned there were lawsuits against OpenAI and ChatGPT from parents whose teenagers committed suicide and ChatGPT helped them to do it. For me, that was one of the that was one of the final straws. I was a Chat GPT user for many months, practically a full year. And when those stories started hitting the news, I just felt, you know, I don't want to, I don't want to pay to support a company that seems this irresponsible. So that's when I started thinking about and eventually made the switch to Claude. Now, there is also, from what people tell me, a general pushback against AI and people saying, you know what, I just I don't like this AI. I don't like where it's going. I don't like what these companies are doing. I'm not going to use it. I'm not going to be part of it. Now, I get that. I really do. But ultimately, you're you're only going to be hurting yourself in the long run because we are well past the point of returning on this. Okay. The genie is out of the bottle. There's no putting it back at this point. And even if every single person around the world, as individuals, said, you know what, I'm not going to pay to use these, it's still too late. Like I said earlier, governments are using this, big companies are using this, so it's too late. I'm sorry to say that, but I say that for a reason because at this point, it's kind of like learning to read, okay, for your future. You remember, I don't know what grade we learned to read, but let's say it's first grade. If you can think back to yourself in first grade, and let's say that you were, I guess, more mature than most first graders, and you had the opinion, you know what? I don't like reading. I don't want to learn how to read. I'm not gonna, I'm not gonna want to read anyway in my lifetime. I'm just, I just want people to tell me, I'm gonna watch videos and whatever, and hey, AI is gonna tell me the stuff I need to know. So I don't need to read. You know what? You could probably get away with, you know, elementary school. You might even get through middle school without learning how to read. But ultimately, at some point, you're gonna be way behind. And then you're gonna have a lot of trouble. It's well, it's gonna be very hard to catch up at that point. So I don't want you to be in that position. It reminds me of what Sam Altman said recently about AI taking people's jobs because there's also this concern that AI is going to take everybody's jobs. And at least at first, that's not true. And what Sam Altman said is AI is not going to take your jobs. Somebody who knows how to use AI will. So learning how to use AI is like learning to read. You're future-proofing, or at least you are preparing yourself by giving yourself a skill that is going to be required in order for you to fit in and earn money, make a living, do something professional. Okay. But the challenge is if you have been using any AI model for any length of time, you've probably given it a lot of information about yourself, how you like to do things, the your workflows, etc. I know for me that was a huge challenge. Like I said, I had been using ChatGPT for pretty much a year before I started thinking about trying another model. And at that point, it seemed totally overwhelming because I had spent so much time with ChatGPT. I had given it so much information about myself and the way I like to do things that moving to another model seemed like it was require starting over. And at that time, there really was no method to do that, at least that I knew about. What I ended up doing is I I started up a new Claude account and I asked Claude, hey, I use ChatGPT and I use it for all these different things. I want to try using you, using Claude to do these things. What is the best way to bring my information over so that we can begin work without losing a lot of ground? And that's how I did it. Okay, was it easy? It was not a something I did once and was done. It was a little bit of a process. Now, since then, more methods, more ways to do this have come up. In fact, I was going to do an episode about this a few weeks ago, and literally like the day after I planned that episode, an entirely new method of migrating from one model to another came out, like literally the next day. And since then, I think a couple more have come out. So at the risk of giving you advice that will be outdated in a couple days, I'm going to try to summarize the methods that are out there now and suggest to you a method that I it will be as future-proof as I can imagine. Okay, so there are a number of ways you can do this. If you want to move from one model to another, whether it's chat GP to cloud or vice versa, or between any of them, you can simply, one way to do this is just copy and paste a chat, an entire chat from one to the other, and then resume that chat in the new model. Okay, that's probably the simplest, quickest way to do that. Of course, there are a lot of limitations. You're just going to be working with one chat. It's not going to have all of your historical data, or if you work with projects, it's not going to have the rules for those projects. There's just not going to be a whole lot of understanding about the way you work when you migrate to a new tool with that method. Now, the advantage of that, of using that method is it's, like I said, it's very quick. It's probably the simplest way to do it. And if you have a chat ongoing with one model and it's relatively simple and it's well contained, meaning it's not part of a uh of a project and it's not related to other chats in that project. It's just relatively simple. That might be a good way to just sort of test out how a new model or how well you like one model versus another model by looking at a single chat like that. Might not be bad. Now, another way you you can do this, and this is a little more advanced, is you can, if you're using projects, look at whatever rules you have entered into a project and then try to just copy that project. Just take whatever instructions you put into that project in, say, Chat GPT, and copy and paste them and do the same thing in cloud. Okay, that will be a little more information to work with. Now that is still not going to bring over all of your uh say chat history, okay, which I think would be a pretty big uh disadvantage. So another way to do this, which has one of these methods that have uh come up fairly recently, is you can export your data from one of these tools to another. You're when it says export, I don't know exactly what is included in that export, but supposedly it's everything about your account. So this would, to me, sounds like be the most comprehensive way to do this. The downside, however, is that there's gonna be a lot of data, okay? And we don't necessarily need to move over all the data. So, like in my case, I had been working with Chat GPT for a year. I didn't want to try to copy and paste and bring over an entire year's worth of information, especially because I know that some of the some of my chats in there uh went in the wrong direction. I know that they were mistakes. I it's things that I would not want to carry forward, basically. So it's things I didn't want to move over. Okay, so that seemed, when this feature became available, seemed like a pretty good method, but also seemed to me to be a little bit overkill. And because of that had some potential downsides of its own. Now, just in the last few days, another feature has come out. I got an email in my inbox about this just a few days ago. Claude just launched a new memory import. And this is slightly different because you are importing not everything, just the memory from another AI model. So just the things that you would want ChatGPT, for example, to remember. Remember earlier I said, you know, exporting your entire account is going to export all the things that you might not want it to remember. Exporting and importing the memory, supposedly, is just going to be the things that you ask ChatGPT to remember, the things that you do over and over. So, in other words, just the good stuff, basically. And the title of this article that arrived in my inbox is Claude Just Launched Memory Import Parentheses. Now you can cancel ChatGTP faster. Okay, so I will actually link to that article below so you can check it out. But my suggestion here is these new features and new ways to do this are gonna keep coming out. And it also is gonna depend a little bit on your on your situation. What do you use these models for? So I can't tell you exactly which method is going to be best. But if I were doing this again, if I let's say I decide I want to switch from Claude to Gemini in the future, which I might, what I would do would be kind of similar to what I did is I would go to Gemini, I'd start up an account, I would give it some details, I would give it the date, and I would ask for the most current methods to do that. And you want to give it the date because one thing that AI is not great at is dates. So if you ask, and you may have encountered this with ChatGPT or Cloud or any of them, if you ask for advice on how to do something specific, it might give you advice that was accurate six months ago, but no longer accurate. So I would begin this process by saying, hey, it is, and give it the date and say, here's what I want to do, here's what I've been using this other model for, here's what I would like to do, what are the methods? What do you suggest? What are some good ways to get started so that I don't get overwhelmed in this process because it can be overwhelming. When I did this the first time, it was a bit overwhelming. I tried to do everything at once, and then I had to scale back and basically do a little bit at a time. And so for a while, I did have two different accounts I was using. I was using ChatGPT for some things, and I was using Claude for some things. And at first, I have to say, the difference, at least I thought then, wasn't that significant. I did think that Claude was a little bit better, and so I started using it a little bit more and more. And then at some point, when I went back and used ChatGPT for something, I was amazed at how frustrating I found it. It was, it just seemed like it was there was a night and day difference. So now I I honestly can't believe I was using ChatGPT for that long. Although, like I said earlier, some of this is just personal preference. You might find that you like ChatGPT more than the others, or you might find that you like Perplexity or Gemini. Whatever the case, I do encourage you to at least try these out. Try out the different ones. They all have different personalities, and you're likely going to find that you really probably like one more than the others. Okay, I hope this helps.