Okay, so we're going to talk today about what AI is - Artificial Intelligence, Artificial Intelligence. And what is ML - Machine Learning? Very often, we see people use these technologies and their names interchangeably, without necessarily fully understanding what's really behind them and how these technologies really differ from each other. Artificial intelligence. Let's start with it. What is it really? What is its task? Artificial intelligence is, of course, an algorithm. It is nothing more than an algorithm that is designed to mimic human intelligent behavior to the best of its ability or is designed to perform tasks as humans would perform them. In general, artificial intelligence is about just having the best possible way, simulating maybe, I don't know if that's the best word, but it's a bit like that. It is supposed to simulate how humans behave. Of course, there are already conducted, there have been experiments more than once, which showed that “Hey, look the artificial intelligence is so cool that you can't distinguish it from a human.
Of course, there are Turing tests that continue to investigate whether we are actually dealing with artificial intelligence, etc.
Still, this artificial intelligence of course is not perfect, but we are getting closer and closer to this artificial intelligence actually replacing us in many aspects. And so, as you can see from the recent news that has brought a great deal of hype to the market, and because of this we are also recording this episode today. Because we are getting a great many questions, a great many companies suddenly want to implement this solution. There is ChatGPT. As you know, it is probably an artificial intelligence algorithm that is used to generate texts. It is a so-called generative AI, which, based on the information supplied to it, queries, texts, and so on. It can either generate these texts, paraphrase them, or, for example, generate a piece of application code. And this is great. Very many people want to use this now because they simply see that this artificial intelligence, mimics just how humans behave, that is, just generating some text, with very often high probability. Here spoiler. Note also sometimes this artificial intelligence is wrong, so we should still keep an eye on it and check what it has de facto generated and whether what it has generated is actually true, makes sense.
You have to remember, for example, that ChatGPT has knowledge up to a maximum 2021, so everything that happened in 2022 or 2023 years is unknown to it. So if new facts have appeared in a certain area, it will not take them into account. On the other hand, you can see that these Artificial Intelligence algorithms are behaving better and better, they are imitating us better and better. At the moment when, for example, it just comes to generating text, we have, for example, the task of generating emails, blog posts, etc. This is for the most part able to simply crank us out or simply help us, inspire us to do things. It simply imitates a human being in this aspect. On the other hand, what is Machine Learning? It is machine learning, which is very often confused with artificial intelligence. Machine learning is also an algorithm, only operating on a slightly different principle. Machine learning is a self-learning algorithm, on the basis of data supplied to it, images, text, or anything else or combined in different sources.
Its task is to generate predictions for us. What does this actually mean? Google, for example, has a system that is able to recognize what is in the images. It can also recognize our speech. If you see just transcription algorithms somewhere, very frequently these are the ML algorithms. Machine Learning algorithms, which are just supposed to us, based on the data collected, learn, are able to give more and more accurate predictions and spit out to us the probability of this information. How likely is this information that we're asking about likely to be in this dataset? How does it work? You can even test for yourself this Google algorithm for recognizing what is in an image. You're also surely familiar with that CAPTCHA system that asks us what we see in a given image, or to mark all the motorcycles in the images. Well, that's how we also provide data to it. If we show such an algorithm a picture with, say, a teddy bear or a dog in it, and ask it what is in the picture, this artificial apple tree has again incorrectly used these terms interchangeably.
It is this machine learning algorithm that is supposed to tell us what it sees in this picture based on what it has seen in other pictures, which users have also pointed out, which it has been taught.
He will tell us: "Hey, this percent certainty, I have the conviction that in this place in the picture, there is a dog. In this place, there is a cat, and in this place, there is a tree." And you'll have this. There will be these squares shown that will show you just the probability of occurrence that this is what we're asking, and so better and better predictions will be provided to us based on this algorithm. So moving a little bit smoothly to the next topic we were going to touch on today, what can we de facto use artificial intelligence for and what can we use ML for? Artificial intelligence is used very often. Where we use intelligent advisors everywhere. I'm sure you're familiar with Cortana, Alexa, Siri, the Google Assistant. These are all artificial intelligence. We ask an open-ended question, a question of some kind, and Artificial Intelligence analyzes the text. What's in it, what we said, and based on that it spits out some answers for us.
If we ask Siri about a joke, for example, Siri will tell us a joke that someone knows or understands. If we ask the Google Assistant to recommend to us where we should go in Warsaw for, for example, "I don't know where I should go.
some Thai food, she will, based on the dataset and what other users of what she knows have said, suggest to us where we can go. Also it will be able to suggest some other things to us. Artificial intelligence is also being used wherever we need to mimic in some way just this human behavior, that is, for example, just generating content, generating images based on the information provided. That's everywhere we use artificial intelligence. Of course, artificial intelligence has a very much broader application, and I think there is no point for me to talk about it in detail here today, because, first of all, I am not an artificial intelligence specialist, and secondly, on the Internet you will find a lot of nicely described examples of how we can use artificial intelligence, whether to analyze data sets, or to provide various information, or to suggest behaviors, suggest various things to us, and so on and so forth. What, on the other hand, can we use machine learning, or this machine learning, for?
Of course, we can use it wherever we need to have some kind of production system, based on a production system. A good example is, for example, counteracting SPAM in our inboxes.
Surely more than once if you use Gmail or Outlook or other email systems, If you receive emails, you very often see there such labels at the top: "Hey, according to me according to this inbox this is some kind of spam," some kind of spam, and how does this inbox know this? Well, that's how it knows it, because it analyzes every email that comes to us, it compares with its knowledge, with the whole set of data, with what users have reported in these inboxes as spam, and based on the just provided data it has learned. And this email that came to us in a pattern reminds her of other emails that were marked as spam and she has a pretty high probability that, this too is spam. Of course, she doesn't say that it definitely is, because she can never be 100 percent sure that it is definitely spam. He can never tell us with 100 percent certainty either, but he does suggest to us, "Hey, pretty high probability that this is spam.
" As I said before, it also allows us to analyze what's in the images, for example.
What is in the text can provide us with just a prediction of what is in the text. It can also recommend products to us, for example. Surely you know these systems from various online stores, where you buy some products and at the very end you get the information Hey, users who bought this product also bought such a product. Users similar to you bought such products, maybe you will be interested in any of them. This is also what the machine learning algorithm has learned from the data it collected, that users similar to me. This is also what it has learned, by the way, that there are users similar to me and they buy similar things. As a result, the store is able to increase its sales, because it can get me interested in something that exists, for example, that I didn't know about, and that de facto could actually be useful to me. So, as you can see, artificial intelligence or machine learning algorithms have very broad applications. And here again I refer you to Google, which also relies precisely on these algorithms.
All of them, where you will find very interesting examples of how we can de facto use these algorithms.
Right. And based on what's been going on lately, there's been a lot of buzz about ChatGPT, or other systems for generating images. As beautiful images this artificial intelligence can generate, can anyone use AI and ML? In general, as much as possible, yes. Anyone can use it. Not everyone can, of course. It is also becoming increasingly easy for us, as ChatGPT, among others, has just proved, that we can very easily use from a very powerful, very giving artificial intelligence. On the other hand, we should not be guided only by the fact when using artificial intelligence or ML that there is a hype for it lately, that some high-profile solution has been created, and immediately want to integrate it in our solutions or something else. This should always be followed by, of course, some arguments. Why do we want to use something, what real value will it give us. Because you must always think about the fact that if something is fashionable, of course it doesn't mean that we should use it.
I very much encourage it, of course, because these solutions carry a lot of value, but used as usual in an appropriate, thoughtful way.
Then we will be sure that what we do, what we realize, carries some value. Can anyone benefit? Here I would still like to emphasize one very important thing. Of course, artificial intelligence algorithms are divided, if we would like to use them, they are divided into two areas. Of course, there are general-purpose or machine learning artificial intelligence systems that we can integrate into our solutions, like ChatGPT, Google Vision and so on. Which we can integrate, which expose their APIs and expose their, let's say, such connectors, so that we can connect to that solution and use it, in our solution. To make it easier for us to work or to develop, or, To maybe not develop. Cutting. That it expands the functionality of our application. These are the kind of freely available artificial intelligence systems. On the other hand, of course, there is also the second approach, which is an approach in which we can simply create such artificial intelligence or Machine Learning algorithms ourselves. Of course, to create such an algorithm and, to use it realistically, to reap real benefits.
In order for it to work, well, first of all, we need to have huge data sets with which we will be able to feed these algorithms, teach them something about how they should function, and we should have a corresponding large team of data scientist, data engineering, etc.
Itd. Which will help us build it but this is a topic of course already more directed to companies that are involved in building such solutions. What I wanted to talk a little bit about today, to talk about that, what I wanted to tell you a little bit about today, is how de facto we can use this AI and ML in no-code and whether we can at all. The first answer I want to give will be on whether we can use AI and ML in no-code or low-code. The answer is short and simple. Of course we can. We can, of course, use as in any other traditional software. We can integrate with solutions that already exist, that is, the aforementioned Google Vision, ChatGPT and all those artificial intelligence or machine learning solutions that put out endpoints, that are public, maybe not public, but, because that's a bad term, but that make their services available for integration in other solutions. This is where we will be able to integrate any such solution into our no-code or low-code application in a very simple way.
All we need to do is simply connect via API, provide it with the right information and the right information, we will simply collect from that application.
Is such a process complicated? Of course, I'll answer like a typical consultant. It depends. It already depends on the very solution we want to use, because some solutions are very easy to integrate. We can integrate them in a pretty quick, friendly way, others just require a little more work. The question is also, of course, what are we going to do later with these results that this artificial intelligence or machine learning will provide us with, What are we going to de facto want to do with them, How will it affect other functionalities in our application? Is it going to be something or is it just going to help make a decision for users that the user is doing something in our application, they don't know what to do. This is where the artificial intelligence comes in and simply displays the text "Hey! Based on the data provided, we suggest you do this and this," or will it be something more complicated? If we would, for example, want to implement integrations based just on machine learning into the store, that is potentially something more complicated, because here we have to implement a system of user research, information about our goods, say which are similar, which are not similar, why they are similar, and so on.
Itd. This is already a much more complicated process, and of course, as in the no-code itself, we can integrate these solutions quite easily. This is one side of the coin. The other side of the coin is, of course, what de facto solutions we will use. As I said, will we also have to configure these solutions, train them in our own fashion, or will it be something like Google Vision? That is, this algorithm is already taught to recognize images, and we simply, for example, want to use it to study whether users upload some inappropriate material to our site. Suppose we are creating social media, we want to build some kind of social media platform, and we would like to protect ourselves in any way against users uploading pornographic material there, for example. We can build it in such a way that the moment a user uploads a photo to the platform, it flies right to this Google Vision. The photo is examined by this Google Vision to see if there is pornographic content there.
If, for example, Google is more than 50% sure that there is pornographic content on it or 20%.
Here, of course, this pain point will depend on how we set it, then such a photo is automatically deleted and blocked from uploading to the platform. So here as much as possible we can use such solutions, and we can integrate them. What can we actually do in the no-code approach? Integration with these platforms is unlimited as long as we just want to use them, as long as it is based on, for example, an API or some piece of custom code. Just as I just gave you an example of using the ML algorithm for image recognition, we can implement any other solution where we can plug in. Whether it will be image recognition, whether it will be recommendations, whether it will be just generating text, whether it will be generating images, whether it will be some assistants. This is where our possibilities are unfettered, just like the traditional approach to application development. To summarize, what we know from this episode is: What is artificial intelligence? What is machine learning? What are the differences between these solutions?
We talked about what these solutions are most commonly used for, and what we can use them for, and learned whether we can implement them in our no-code / low-code solution. As much as possible, it is possible. It is doable. Of course, it always depends on what we want to build and whether it will be of real value to our users. I hope you found today's episode interesting. I managed to tell you a little bit about what intelligence is, how it works, what we can use it for, and to hear you in the next episode of the Just No Code podcast. Bye!
.svg)
.svg){kind=small}
