Transcription
If you believe that neural networks are some kind of toy or tool that does not yet deserve your attention, then perhaps you are already losing competitively. It doesn't matter how a business or how a specialist. Hello everyone, you are on the channel "Protitization of Business" with Alexovich. My name is Andrey, and today we will talk about neural networks and how to remain a competitive specialist or an effective business in the era of digital business. Let's first understand what neural networks are. 12-15 years ago, I studied under the head of marketing at Yandex, Andrey Sebrant. Many thanks to him for this. And he explained what a neural network is. They were just appearing then. He says: "Imagine a situation, there is some kind of conveyor belt, cucumbers are moving along it, and your task is to select and put into a jar small cucumbers, which are the crunchiest, which, in fact, are our product. And there is an auntie, she takes these cucumbers from the belt, she understands that this small gherkin is for here, and this one is a little bigger, no, this is for another product. And she, based on her experience, understands, this one is worth taking, this one is not worth taking. How can this task be solved by removing this, let's say, Granny Nyurka from the conveyor belt and automating this process? Technically, we can get each cucumber from the conveyor belt and measure it somehow. We can look at the size, we can look at the weight, and so on, we can write an algorithm that, in fact, analyzes this and puts one cucumber or another into the jar. But the trick is that we need to input specific parameters into this algorithm. We need to say that a cucumber up to 3 cm is, yes, a small gherkin, it fits these conditions. And if it's large, respectively, larger than this size or weight, then, accordingly, it doesn't fit. You need to write an algorithm, and then the algorithm will do it. And there is no wow, no novelty in this.
And now imagine another situation. Imagine that we have a box that simply looks at this belt, and it can make a decision, whether to put this cucumber into the jar or not. It works as follows. First, we set up this box, a camera, and this granny spends several hours making decisions, and this box watches and understands, oh, somehow it develops neural connections inside, that this cucumber can be put into the jar, and this one cannot. After that, we remove the granny. And the miracle machine has learned by itself, and now it makes decisions by itself and puts the right cucumbers into the jar. This is exactly the story about this black box, and it can learn something by itself. And there is, in fact, a neural network. That is, you don't input an algorithm and get a strictly defined result, as the algorithm is programmed, but you teach it using examples, and this thing works effectively. But before I tell you about the use cases of neural networks, write in the comments about how you used them, for what purpose, what results you got, if you are already a user of neural networks. And the simplest use case, which you have most likely already seen, is that you can ask a neural network any question that interests you. Before, you had to go to Google, type in your query, see a certain number of pages, click on these pages, and try to understand, in fact, where the information you are looking for is, and draw some conclusion, combine this information. Perhaps gather it from several sources, to understand a certain issue. Now, you ask a question to a neural network, and the neural network gives you the answer directly on a platter. You don't need to go to any websites to search for this information at all. Yes, you can, for example, use specialized neural networks like Prexti, which will provide not just an answer, but also links to the websites from which it took this information, to read in the original source. If we take the same Google or the same Yandex, they have already integrated such mini-prompts, mini-answers from neural networks to the queries that the user types directly into the browser. I think that in the future, only these prompts will remain, because simply searching through ordinary articles will no longer be as relevant as before. This, by the way, will kill the entire business of classic search engines, because they sold clicks for queries, and now the market is transforming, and you can no longer sell these clicks, these queries. The first 3, 5, 10 positions in search engine results were advertisements. Now, why use this? Either you have already received an answer from a neural network in the search bar? Or you just opened a neural network, the same ChatGPT or DeepS, and asked it a question, and it answered you, and that's it. You don't even visit such sites at all. Why would users use the old Yandex, with its search bar and results, if there is a lot of advertising there? If you can easily get an answer to the question you are asking from a neural network. And thank God, our Yandex has a lot of other services, and wherever it has put its hand, and on what they earn, and most likely now their share of income from contextual advertising is declining.
Let me give you a personal example. Last Saturday, I was buying a watermelon. I went into the store, and there were 12 watermelons on the shelf. I thought: "Damn, what if I try to use a neural network, just ask, which one of them is ripe?" I took a photo, sent it to ChatGPT, and lo and behold, in a few seconds, ChatGPT replied: "Take the watermelon from the second row, far right," because it looks like this. These are signs that it is ripe and it will be sweet, it will be tasty. And it even gave me a mini-guide on how to choose watermelons. And it advised: "If you don't choose this one, choose this one. This is also a good sign of a good watermelon." I naturally took the watermelon that GPT recommended, and I was satisfied because it was indeed a sweet, tasty watermelon. I have never chosen watermelons in a store so quickly and easily.
So, let's move on to how to use neural networks in business, in your business activities. The most general example, of course, is that businesses can also make requests to neural networks. For example, there is some question, I don't know, about the 2-NDFL certificate. You can even ask directly, try to explain this complex question to me in the simplest language, and I don't know, as if for a child. And neural networks indeed explain the most complex questions quite simply for understanding. Neural networks are also actively used now in various video conferencing services, in Zoom, in Yandex Telemost, and so on. They are also built into the system, as they say. You talked with your client or had an internal meeting. A neural network analyzes and converts your negotiations from voice to text. And it can easily and simply analyze who said what. That is, by the tone of voice, it can understand that this is speaker one, this is speaker two, and so on, who was at the meeting. And it can transcribe this phone conversation into text, and then process the text through itself, analyze it, and immediately write conclusions about what was discussed at the meeting and what was agreed upon. That is, such a mini-summary, with specific conclusions, with specific agreements, who promised to do what, by what deadlines. The neural network provides it, and as soon as the meeting ends, you get this result. That is, you don't have to sit for another hour or an hour and a half after a two-hour meeting to write a report on the meeting. Of course, it is worth checking the results for the first time, correcting the text itself that the neural network provided, but globally, 80-90% of what it provides is already edible. You can use it. You can already take it as a basis in business. And it is very convenient when, for some agreements that you reached at a meeting, you can recall this meeting, the specific moment when it was said, and point out directly to the interlocutor, the opponent: "You promised this at that time." A person will no longer be able to deny their words.
Another application, and we are already using it in our business in our products. We are releasing our CRM system, a sales management module for 1C. By the way, if you are interested, follow the link in the description. You can see in more detail there, what functionality our CRM system has, and what it will be useful for. There is an option, accordingly, to connect an AI for transcribing phone calls. But besides transcribing phone calls, if it's a sales department, there is an opportunity to analyze the entire phone call, whether the manager established contact with the client during the conversation or not, whether there was an attempt to sell the product or not? Were there any client objections that the manager worked through or not? You can give a specific conversation, and it will analyze it according to these parameters, well, each of them, and draw a conclusion about how successful the manager was, for example, on a point-by-point scale, what can be improved for the manager. What does this give? It provides the ability to see a tabular report on all manager calls. And immediately, the Head of Sales, we built such a report into our CRM system, this is what it looks like. Immediately, the Head of Sales sees that this call is red, maximally unsuccessful, apparently. He can listen to this call, he can see by what parameter the call was evaluated, analyze this call, for example, at a general meeting of managers, or vice versa, those calls that are maximally green, where the manager was a star, consulted the client, worked through objections, and a deal happened. The same. You can immediately visually see that the call is green, and also analyze it at a meeting of managers. Thus, this is a working tool for the head of sales to improve customer interaction, and for the sales department. And you don't need to hire a separate person, some girl, who will listen to calls, who anyway cannot listen to all calls. In the best case, she listens to 10% of the sales department's calls. Here, the AI can listen to absolutely all calls and give recommendations, draw conclusions on all calls. That is, in essence, the AI replaces the quality control department, the payroll fund of at least one person who listens to calls in the quality department. And this is already 10 times more profitable than keeping a separate person.
Since we have touched upon the topic of replacing specialists, let's scare you a little. What real professions are already dying out? If someone is still working and working in the old way, these are already candidates for dismissal tomorrow. The simplest ones are the guys who do translations. The AI has learned to translate from one language to another very quickly. It translates very coolly. The quality of translations is absolutely different compared to what translation programs could do 10-20 years ago. Translators, in principle, are already practically not needed. Another profession that is under serious threat is designers. It's cool, of course, to draw one picture or one screen and coordinate it for a long time with the client and so on. And design coordination sometimes takes days or weeks. This is simply a horror. When the client himself can get the first picture, formulate what he wants to see in the neural network, and then, by correcting, by simply entering text, saying, "remove this logo, add this, and so on," get a finished product that can simply be taken and used in their activities without paying a lot of money to a designer. That is, the savings on designers are not even tenfold, they can be many times more. I know marketing companies that had departments of 100 people who were involved in content preparation. These were designers, their managers, and so on. And they simply reduced their staff tenfold, from 100 people to 10 people. The same team of 10 people produces the same volume of content with no worse quality, but in the same volumes as the team of 100 people did before. Therefore, designers, learn AI, learn how to write prompts correctly. Not so much the inscription and work with your hands will be important, but it will be important for you to understand how to compose a prompt to get a cool picture, what to tell the neural network to get the result you need. Therefore, designers are under serious threat, in my opinion, at the current moment. In fact, more and more professions will be under threat, because AI is becoming smarter and more capabilities will appear in the upcoming releases in the coming months. But it is important to understand that a person, of course, will not disappear anywhere. A person will engage in more creative things, more intellectual things. All this routine that people dislike will be done by robots, by AI. And if you want to see a video about what other professions AI will replace, then subscribe to our channel. In the very near future, we will also shoot a video about this.
And in fact, technologies where they can be applied will only increase in the very near future. The advertising that you see on the internet around you on various surfaces will become even more, more personalized. AI will select exactly the advertising that is important for you to see and watch at the moment when you are most ready to perceive this advertising and based on this, make some conclusion about purchasing a particular product or service. Therefore, I believe that thanks to neural networks, the advertising market will change significantly, and the advertising market that is specifically tailored to a specific person, to a specific user. This is advertising. Continue to follow neural networks, improve your life thanks to them. Subscribe to the channel. And with you was I, Andrey. See you in new videos. Bye.