ChatGPT Dude, these are the famous words from South Park Everyone knows this tool now, even my mother used it And I got a lot of comments from you guys saying that not everyone gets the same results But we're not just talking about ChatGPTDude, we're also talking about Bing search, MidJourney, graphics creation, everything that's based on large language models GPT-4 or GPT-3 or BARD or LAMA, all these tools that we can use And of course there are more of them All these tools process natural language and try to give us a solution, an answer that satisfies us Whether it's in the form of a picture or a search engine Or a search engine that can also create pictures Or whether it's just research or text There will be more and more things and I think it's time we learn how to deal with it Fun fact at this point There is a website, no it's not an ad, there is a website for prompt engineers That means people who build you a command to get the best result from ChatGPTD, MidJourney, DOLL-E, StableDiffusion or all the other tools There are some guides on Fiverr on how to become a prompt engineer and how to find one And I thought to myself, I don't think anyone has spent as much time with these tools as I have Because I use them to fact check my videos, well not to check them in fact But to check if my storytelling is correct, if my, well for the longer videos, not for the tutorials This is a tutorial series if you haven't noticed, you can't see my face But for the longer videos I use it to ask small questions and get quick answers As a kind of search engine replacement I use it to program, I use it to generate images that are thumbnails for my videos Or that are built into the videos for editing I use it for research, I actually use it for a lot of things And that's why I thought I can share it with you so you can get good results yourself Today we want to deal with the basics first Because as I said, this is going to be a series and I'm still learning more and more And these models are changing of course I had the pleasure to deal with the technology behind it With AI, with neural networks, with the transformer models, with all these great technical things But this series is actually for everyone I won't do it just for computer scientists, even if there are some topics That are especially useful for computer scientists, like code generation We don't want to limit it to that I think it's such an important topic that a lot of people could start with it You can write the actual start in the comments, it's at 3 minutes 20 A large language model is exactly what your keyboard is It completes your sentences So if I write, the sky is, and I'm actually changing to default chat gpt 3.5 That's the old version, because here I'm limited to 25 entries I don't know how much I can or want to record today The sky is, blue, period, what is that? That's a sentence completion In the end, I could also write something like, and now it's important here This button, I don't know if you've used it before, is for editing a prompt in chat gpt With that you can simply change an existing command and modify it again That's super helpful, because we'll also deal with things like additive prompts That means prompts that refine and improve previous prompts And sometimes even let information from the AI flow in here So, the sky is blue, period, okay The sky is blue, but Everyone knows that the sky isn't always blue The color of the sky can vary depending on weather conditions and time of day, blah blah blah You see what I mean My sentence is being completed And that means you can also give commands to this thing So really, that's what your old keyboard did, but in a much better way That's legit, that's how it works And if I want to make it a lot better I can simply ask questions Is the sky blue? And now it's being completed for auto Yes, the sky is blue, blah blah blah What happens here in the end You have to think about what this model was trained on It was trained on a lot of language, it was trained on dialogues It wasn't trained on mathematical models for the smartasses Like the people from Schlaumeier who think they should ask if 5 plus 5 is 10 or something It wasn't trained on numbers, it's not a mathematical model, use something else It's a large language model You can see my hand gestures, I don't have a camera Anyways, the sky is blue, but You get what I mean It won't complete what we're writing here You can imagine it like this I found this very helpful When we ask a question We can always use it as a form of an exam So ask a question Or ask a question that you could answer in an exam Whatever exam it might be But give all the necessary information So if you want to do this If you say, explain quantum physics, no, quantum computing in one simple sentence Then you get exactly this It's in a simple sentence That gives you a lot of information So the information you see here is extremely relevant You have to pay attention to how you formulate it If I change one word in a single sentence Then it can change paradigms Make complex calculations faster than with classical computers So this is something that not everyone would understand So it depends on how I ask this question And here's a first little life hack That I find very useful Explain quantum physics Task Explain quantum computing Goal group Computer scientist Because I'm a computer scientist, you can easily adjust this for yourself Complexity Simple No knowledge of physics And now we get the whole thing exactly as I would like it to be You see that it works with terms that I, of course, know as a computer scientist So in contrast to classical computers Here the large language model simply assumes that I know the things that are based on bits, i.e.
0 and 1 Quantum computers use so-called quantum bits So qubits And then I get an explanation of what these quantum bits are So I get exactly what is relevant to me So your output is only as good as what you write And I actually recommend you to try out this structure Task, goal group, double point and complexity Maybe as a first little start As I said, I'm going to go a little more into these basics, prompting things But we also want to look at a few frameworks that have been developed in the meantime Yes, there are actually frameworks like interacting with AI I know it sounds ridiculous at first But you just get better results And that's the goal of this series I just want to make it possible for you to get a little better with this You are welcome to tell me what you expect from this series And yes, we'll hear from you next time See you, bye
