Select text to annotate, Click play in YouTube to begin
Anyone who thinks that AI tools like ChatGPT are not that useful simply haven't seen what it's capable of. So in this video, I'm going to go over a whole bunch of real life examples that I personally have used ChatGPT for. And if you've been assuming, for example, that ChatGPT is only useful for "creative writing," which I actually don't think it is, well, none of these examples will involve any kind of creative writing. I think you'll be surprised at the wide variety of stuff
that you may have just never considered using it for. I will point out that all of these examples use GPT 4 specifically, so you can use this for free through Microsoft Copilot, either through the Bing website, it used to be called Bing Chat, now it's Copilot, or also through the Copilot feature in Windows. Though apparently right now, only the creative and precise modes use GPT 4, but Balanced
uses GPT 3.5. Or you could use it through the paid ChatGPT Plus, which is what I use, and you get some more features, notably a lot more text that you can feed it at one time. Importantly though, don't even bother using ChatGPT 3.5. If you've ever used the free version of ChatGPT, you were using GPT 3.5, which is absolutely brain-dead compared to GPT 4. I never use GPT 3.5. And 99%
of the people who say how dumb and useless ChatGPT is, were using the free garbage version. One more quick thing I want to address before we get to the examples though, is for all the people who say, "Well, how can you even trust ChatGPT to be right?" And for that, I actually agree, but that's why I usually use ChatGPT for situations where the information is hard to get initially, but once you have it, it's easy to verify. And you'll probably see what I mean in a lot of these
examples. You'd realize sooner or later that it was wrong when you go to actually use its answers, and you could ask it to correct itself, stuff like that. Alright, now the first example is asking the AI what a certain command does that you may have come across online. People say, "This command will solve your issue," but you're not really sure what all these different parameters are actually doing. You want to make sure you know first. So you can simply paste it in and say, "What does this command do?" And it tells you each of the parameters, like it specifies the
client mode with -c, the -e option is enhanced reporting, all down the line so you can figure it out. And if you really want it to be sure, or if it's a more obscure program that you're running, you could actually just copy and paste all the documentation of like the help command or something that it returns, and it'll just look up all the commands for you so you don't have to search through yourself. Next up, example number two is kind of the reverse, where you want to do
some command, but you don't know what to type in. So you can simply ask it something like, "What FFmpeg command would I use to extract the audio track from a video as mp3?" And then it will spit out a big command, tell you what each thing does in this case, and you'll be able to easily tell whether or not it's correct right away by when you go to use it. But if it's something simple like this, generally, it'll probably be pretty accurate. Another really good use case,
I think, is if you have a bunch of data or text that is not really formatted well, if at all, and telling it to format it. So for example, here I have a bunch of random data with people and their email address, phone number and stuff, but notice that it's not consistently formatted at all. Some lines, it says like "email:", other lines, it says "email address", and sometimes the phone number
is preceded by a number sign. In other cases, it literally says "phone". So it's not like you could probably do this programmatically. You'd have to kind of manually go through and pick out all the information. But I simply pasted this in and said, "Format this data nicely," and it went and did all that for me, put it into a chart. Though again, I want to point out that's just an example of the capabilities. I wouldn't necessarily trust it with like mission critical data. Again, like if you
have a whole bunch of data and you aren't going to be able to easily go through and verify all the data, but you yourself could decide what uses are tolerable to have a potential error in there. Here's an actual example of something I used one time. I had a whole big error that was spit out by a program I was working on, and I asked it to simply format this error nicer so it's easier to read. And it did just that. It put everything in bold and specified what different parts of
the error were going on. And then it also took the details that were kind of all in one jumble, made it much easier to read. Another somewhat related use is if you have a specific format you want to take some data and put it into, like here I have an example table of data in like a plain text table, and I want it to be in CSV format, comma separated, so I can put it into Excel,
for example. So I asked it to do that and it did. Next up is what I think is one of the most useful things about ChatGPT. And that is to use it kind of like a reverse dictionary where you don't know the name of something, but you know how to describe it and you want to know what it's called. For example, at one point I was looking for this object, which I was not sure what to call it, maybe a spray bottle, but obviously that's going to bring up different types of spray bottles. So
I simply asked it, "what's the name of the kind of spray bottle that you squeeze and it sprays a stream instead of like a mist, like you'd see in a chemistry lab." And immediately it says that is a wash bottle and that is correct. So again, it's one of those things where it's very easy to verify if it's correct. You simply Google it and see if wash bottles come up. So I especially like using it for that. Here's another example. I saw this picture of a weird fire extinguisher that didn't have the valve and stuff on top. I had never seen it before. So I literally uploaded
a picture to ChatGPT with it and said, what's this weird looking fire extinguisher with the flat top. And it told me that it's a cartridge operated fire extinguisher, which is correct. I had never heard of that before. Or you could do the same thing, but describe it with text. This is a separate chat. I said, "there's some fire extinguishers that don't have a handle and valve on the top. What are they called?" And at first it was a little bit inaccurate. Actually, it said it was a stored pressure fire extinguisher, which is more of a general term, but then it did actually say,
"however, the specific type you might be referring to is the cartridge operated fire extinguisher." So it did still get it right. Another thing that I think AI is incredibly useful for is when you want to look up stuff that's hard to do a Google search for. For example, I wanted to know the fluffiest type of towel out there. Do you know how hard it would be to do a Google search to find that out? You'd come across all these random, poorly written articles, probably written by AI themselves that
are not even accurate. They're just spewed out onto the web, hoping that enough people will click it to get ad revenue, and none of it will be accurate. So instead I asked ChatGPT, "if I wanted to find the fluffiest and most absorbent bath towels, what kind would I need to search for? Are there any different classes or categories?" And right off the bat it's like, "go for cotton, particularly Egyptian or Pima cotton." And it said why, and also mentioned Turkish cotton.
It also talks about grams per square meter, but I was mostly looking for those different types of cotton. And then I asked it to elaborate and compare on these three different types of cottons so I could know more about them, which one I was actually looking for. And it gave a whole list. And then I followed it up with, "are there any brands that make towels that are considered top of the line by experts? And not just random blogs who do rankings." I literally put that in there,
because again, if you were to search "best fluffy towel," you would never find the actual best one. And in this case, it gave a bunch of results that I looked up and they did seem like they were pretty expensive, high level brands. So that's another thing I can point out. It is actually good if you ask it for like the top highest end brands in a certain category, it's usually pretty
unbiased with stuff like that. And one of my favorite examples is if you're trying to look for a specific episode of a TV show, but you can only remember like something vague about the scene, you can type that in. And here's a challenging example I gave it. I said, "What's the episode of Star Trek, the next generation where nobody on the ship can see them." I was purposefully vague here. I said, nobody on "the ship" can see "them", like not being specific at all. And it actually got it
right. It was the episode, The Next Phase. And it even says the 24th episode of the fifth season. And yes, it did get the season and episode numbers correct, I did check that. I do want to point out though, before I continue that for a lot of these examples, I'm not saying that you might find these uses useful for you. They're very specific to something I was doing, but more so, I just want to show you some random examples to kind of open your mind about what you might be able to use it for
in your situations. Another use that you may never thought of, not many people might need to do this, but you can ask it to translate a specific word into many languages at once. And more importantly, within a specific context. For example, here's where I ask it to translate the word "file" into a bunch of languages, specifically with the meaning of a computer file. And if you just type that into a random translator, you might not get that meaning out of it. You'd have to like search
through the different results to see which one it says is the correct context. And that would take forever with a bunch of languages. With this, it's all at once. So you can tell me if in your native language, it was correct with these words. I don't necessarily think I would use this for anything mission critical, but it might be easier to get it as a good start. So in this next example, I was trying to install something called a Python "wheel". You don't really need to know what this
is, just that you have to install a specific type in this situation. And in this case, I happened to receive an error that was saying, "Only wheels targeting CP310 Linux OS are currently supported." I was not even sure what that means. There was a bunch of files in the list that I was looking through that said that. So I simply copy and pasted all the options from the download page and asked which one I should use. And it gave me a result based on the message,
"Here's the compatible one that you should use." And it ended up working. Next, another use is if you do any coding, you could ask it to add comments to your code to make it easier to understand for other people. And then of course, it's easy enough to just simply go through it and see if the code comments are accurate. And you can also tell it to elaborate more and add more comments until you're satisfied with how detailed it is. Next, another example I've used
it for is if I was trying to debug some code in PowerShell or something, and it wasn't something I could really use with debug tools. Well, I could ask it to add in a whole bunch of print statements throughout the code saying the values of different variables at different parts and what's going on as it goes through the code. So it makes it a lot easier to debug and I wouldn't have to go through and add the print statements myself. Here's an example that's probably more practical
for most people. I was looking at the various subscription plans for another AI website, and it offers a subscription with credits and you can buy additional credits that vary depending on the plan level, a whole bunch of nonsense. So I gave it all the information. I said that for $12 a month, you get 625 credits and you can buy additional ones for a cent each and that each video costs 20 credits. And I did do a little bit of math to give it some more context to help out. And then I said,
"However, there is an unlimited plan for $76 a month. So how many videos per month would be worth it to have the unlimited plan?" In this case, I had it include the Wolframe Alpha plugin, which is like a math thing, because I didn't trust it to do math by itself. Never ask a large language model to do math at this point in time, it's terrible at it. But in this case, it basically has the other service do the math, and it actually solved it for me. And it said that the break-even
point would be 351 videos, and then it would be worth it to have the unlimited plan. Here's another kind of technical example. I was working on some filter rules in my router, and I wanted to know what IP ranges this particular IP address was falling into. These are all Google IP addresses, and this was from a list they had published. Anyway, in this case, the ChatGPT used the "code interpreter" feature where it literally used Python code and used a package called IP Address
to check each one. So it didn't just use its knowledge to guess which ones were in there. It literally tested it with code. So I was confident that it was correct in that case. So if it is something that can be done programmatically and you don't want it to just trust its own knowledge, it can do that too. It can literally say, "Alright, I'll write up some code." So yeah, those are just a bunch of very random admittedly use cases, but it just goes to show the wide variety
of stuff you can use ChatGPT for. I know when ChatGPT originally came out, I kind of didn't even think to use it for a lot of stuff, but now it's almost like one of the first things I consider if I'm trying to solve a problem, "could ChatGPT do it?" Or are there parts of it that could be done by it? And it saves me a bunch of time. I use it pretty much every day. And just remember that it's getting better and better. I would bet in a few years, sooner than you think, you won't even
have to worry about it hallucinating incorrect data. You can probably be able to trust it a lot more in the future. Again, these days, maybe you don't want to trust it with mission critical data, I wouldn't. But for a lot of stuff, it works perfectly fine. Let me know what you think. Are there any uses that you think are very interesting that I did not mention and others might find useful? Let us know down in the comments. If you want to keep watching the next video I'd recommend is where I talk about a bunch of cool, useful free programs that you might want to check out.
I'll put that link right there. Thanks so much for watching and I'll see you in the next one.
End of transcript