Stephen Hawking was convinced it’s possible to copy the contents of the human brain onto a computer and create a form of digital life. Many science fiction stories share this notion. While such technologies do not exist yet, neural networks do, and they operate on the same principle as the human brain. Of course, they are far from being real brains (so no machine uprising just yet), but neural networks are already quite capable. The most renowned among them is ChatGPT – a chatbot launched in 2022. Discover how ChatGPT works, what it’s capable of, and whether it’s available for use in Russia in the article below.
The chatbot was developed by OpenAI Inc., which is owned (or was owned) by Elon Musk and Sam Altman. The idea of creating artificial intelligence isn’t new, but it was OpenAI Inc. that succeeded in developing a neural network that can be called AI. ChatGPT was developed in 2022. Its foundation is the generative pretrained transformer (GPT), a language model used to create text. A distant ancestor of GPT was the conventional T9, found in every smartphone. This is evolution. From suggesting the end of a word, through not always successful auto-replacement, to generating any text. In the IT sphere, it can be called a breakthrough, albeit one built on research and developments that spanned decades.
To understand how the AI ChatGPT works, let’s start with the basics – the principle of operation of generative pre-trained transformers or Generative Pre-trained Transformer (GPT). Essentially, it’s a computational mechanism that takes a certain set of sequences, transforms them according to a particular algorithm, and outputs the result. Since any textual, graphical, or audio information consists of sequences, AI can be used for any purpose.
The specialty of transformers is their flexibility and ease of training. Add to that the ability to process vast arrays of data simultaneously. Moreover, they’re capable of self-learning, a feature the third version of GPT acquired only after multiple scaling efforts. And the more parameters the neural network receives during scaling, the more new abilities it acquires on its own.
Thus, this neural network represents the latest version of AI capable of self-learning. And although it’s essentially just a program, its possibilities are nearly limitless. Now, just over a year after its launch, ChatGPT can:
But despite this, the service does not understand context, so it cannot generate highly specialized texts from scratch without errors. For example, in response to the query “The stages of head removal by Vishnevsky,” the neural network produced a flawless text, with the correct sequence of the operation. With logically justified indications for head removal.
Instant data processing, excellent memory, self-learning – these make it possible for the AI to be used for a wide variety of tasks. For instance, at a teacher’s request, the neural network can develop a curriculum or create tasks for a test. For a student, it can compile self-assessment tests or provide information from various disciplines – just the facts, without fluff. Or simply explain a complex topic in simple terms. The service can calculate a personalized diet, suggest recipes, and create a menu considering the user’s wishes and preferences. There are many uses – write simple code (at the level of a Junior developer), descriptions for products or fairy tales (a book of tales has been published with ChatGPT as a co-author), take on part of the routine office work, and so on. However, the program’s capabilities are still limited, so mathematical calculations and the creation of niche thematic texts require checking and editing. Yet, AI excels as an alternative to search engines (though it’s always best to double-check the information).
How to work with the ChatGPT neural network? The developers have created a simple interface – the program window resembles a typical chat in a messenger app. Upon opening the chat, the user begins a new dialogue. One can ask a question and receive the answer in real-time.
That’s the difference between ChatGPT and the previous version of the neural network, GPT 3.5, which did not achieve such popularity for one reason: to communicate, you had to use the API interface, which is inaccessible to the average user. Plus, the developers provide free access to the service (though there is now a paid subscription as well), the ability to communicate in various languages, including Russian. However, the neural network understands English best, which is unsurprising.
Since OpenAI has opened access to its neural network to third-party applications, it is possible to integrate it into Telegram. Specifically, to create bots based on GPT that are even simpler and more convenient to interact with. How does ChatGPT work in Telegram? Just like on the official site. But free use implies a limited number of questions per day, and if you’re using the bot for work – it might be better to get a subscription.
When using it, keep in mind the following:
Due to sanctions, ChatGPT is not accessible in Russia, but this problem can be solved using Telegram bots. If a user needs to work with AI on the website, they will have to look for a country where ChatGPT is operational and register as a user from that country. To register, you need: