How to use Claude Desktop tutorial for beginners

Transcript - How to use Claude Desktop tutorial for beginners

So, you finally decided to install Claude Desktop?

Well, welcome to the club.

Now, I don't know why you're installing it -- maybe you don't like the politics of OpenAI, maybe you're not happy with the results you're getting from ChatGPT, maybe you like the privacy implications that Claude Desktop has. As I said, I don't care why; I just want to make sure that you get Claude Desktop installed.

And, I'm not just going to get you installed with Claude Desktop, but I'm going to show you a couple of the limitations that we run into when we use Claude Desktop, or any large language model. But downloading Claude, getting it installed and doing a couple of queries to make sure it's working -- that's exactly what we're going to do next.

So, to get started with Claude Desktop, the first thing you have to do is download Claude Desktop.

Super easy to do. I'm on TheServerSide there; I'm the editor in chief over at that site. I'm just going to type into Google,  'install Claude Desktop, ' and hopefully that takes you to claude.ai/download.

Now, don't get phished -- make sure you're actually at claude.ai/download. It might not even hurt just to type that in on your own.

It's only going to take a second or two to download, so just open up the file once it downloads or just click Open when downloaded. And pretty soon, all of a sudden, Claude Desktop is going to start installing itself.

Now, as I said, it's only going to take a second before the desktop application pops up. Click Accept all Cookies just to get rid of that nag screen, and then click on that beautiful Get Started button.

You can register with your email. I'm just going to register with my Gmail account, so I'll just say  'Continue with Gmail. ' I've got my little potemcam2-at-gmail-dotcom account already logged in in my Chrome browser, so it's just a matter of selecting that account.

Then it says,  'OK, give us some information about yourself. ' It wants your birthday -- not the one you use on the dating sites, but your real birthday. I was born on Guy Fawkes Day in 1972.

I'm going to accept all of the agreements there. Notice you can upgrade to Pro or Max, but I'm going to take the free edition because, well, because I'm cheap.

And boom, all of a sudden, we've got Claude asking us a few questions.  'Do you understand how this works? ' Yes, we do.  'OK, well, tell me what I should refer to you as. ' My name's Cameron. Put your name in there.  'What are you into? ' I like coding and development. Maybe you like investment.

Now, Claude's going to start guessing about what I want to ask, but I'm just going to say,  'I have my own topic. ' Look, I've got something I'm interested in knowing about. It wants to get going with a quick entry. I'm like,  'No, no, no, no. Just, I've got some questions. Specifically, I want to know if you know the difference between a capital and a capitol. So, why don't you tell me what the capital of Canada is? ' That should be an easy question. So, we click the button, and boom, it says the capital is Ottawa.

Now, what if we change from an A to an O. What is the name of Canada's capitol building? Let's see if it can tell the difference. And, oh, it's a large language model that knows all of this stuff. There's nothing it doesn't know. It knows that we call that the Centre Block, the West Block of Canada's Parliament Buildings. It knows everything, like, right?

So, you know ... who has the most NHL goals? We all know that Ovechkin just beat Gretzky, so this is going to say Wayne Gretzky holds the record of the most goals in NHL history?

So, Claude is wrong, and that's because LLMs only know the information up to the date that they were trained. Since Ovechkin beat Gretzky's record just a few days ago, well, it doesn't know about Ovechkin's new goal record.

And it doesn't know anything about you.

I work for TechTarget. If I say,  'Hey, what's TechTarget's preferred NoSQL database? ' It says,  'Hey, I don't know anything about your personal information. I don't know about the company that you work for. I don't know about all of your internal stuff. '

So, my point here is: Claude is amazing. These large language models are amazing. AI is amazing. But, don't be fooled -- there are limitations.

One of the limitations is the fact that, well, these LLMs are time locked. They only know about the information that was provided to them when they were trained.

And then, the other thing is, they really only know public information, or at least the information that was fed into them. So, they don't have any context about you, about your work, about your organization, right? Because that's not public information.

So, there are limitations.

Now, one of the reasons why I did this Claude Desktop tutorial is because I've got another tutorial about how to write MCP connectors -- Model Context Protocol connectors with Python, with Java, with JavaScript. And those components can actually give updated information to your models, so that you can combine the brilliance of Claude Desktop, the brilliance of a large language model, with context-based information like,  'Hey, what's the latest information on all the records in the greatest sports league in the world: the NHL? ' or,  'Hey, I want to know some information about my company and how things work internally, ' and I want to connect my large language model to that.

Now, if you just wanted to get started with Claude Desktop, well, there you go. That's how you get started.

And if you're interested in using some of these Model Context Protocol connectors to really elevate your experience, well, check out some of the next tutorials that I've done on creating Model Context Protocol components with JavaScript, Python and Spring Boot.