The loudest technological innovation in the world on February 18 – Chatbot based on ChatGPT artificial intelligence It became available for registration in Ukraine. It’s been several months since ChatGPT (Generative Pre-trained Transformer) became generally available, and the wave of hype surrounding it continues. On the contrary, the innovation of the technology world broke TikTok’s record and reached 100 million users at an impressive speed, and now it continues to grow its audience with sure steps.

The response to the next step in the development of artificial intelligence is more than uncertain. Bill Gates predicts that “new programs like ChatGPT will make many office jobs more efficient by helping with billing or letter writing,” while many media outlets list professionals whose labor market demands are at risk.

But we will not make hasty conclusions, although there are discussions about what the new technology is – a useful assistant at work or the “killer” of search engines and professions. Instead, let’s try to learn as much about him as possible.

Search AI is already in Ukraine

On February 18, Minister of Digital Transformation Mykhailo Fedorov proudly announced that Ukraine has entered the list of countries where its citizens can use ChatGPT. “We have finally managed to correct the injustice. Ukraine has been removed from the list of countries where ChatGPT is blocked. The program will not only work in areas temporarily occupied by Russia. So that enemies do not use it for anti-Ukrainian propaganda. After all, ChatGPT speaks Ukrainian well, unlike Russians,” he said. wrote the minister Telegrams. For example, the team of the Ministry of Digitization worked on this decision for a long time – official letters, calls and a meeting with the management of the OpenAI development company. He thanked the company for the “correct and completely sensible decision” because Ukraine, fighting for freedom and progress, certainly has the right to access revolutionary technologies.

On the same day, social networks were flooded with stories of Ukrainian users communicating with an intelligent robot. And what questions were not asked of him!

What is ChatGPT?

For this reason, ChatGPT is a chatbot based on the language model of artificial intelligence (AI), developed by the American research lab OpenAI. It was released as a prototype in November 2022 and caused quite a stir among the world community. Its popularity came almost instantly: 1 million users in five days. For comparison: Instagram has gone to such reach numbers for over 2 months. ChatGPT crossed the 100 million user threshold in just one month. Instagram went to this goal for 2.5 years, and TikTok – for 9 months.

Traditional search engines usually return the most relevant search query. Instead, ChatGPT answers questions directly, so you don’t have to visit multiple sites and filter gigabytes of information. And ChatGPT is customizable. In other words, you can define a certain style and specify which subjects to talk about. In addition to writing text, the bot can give basic programming code, generate financial analysis, articles, forecasts, recommendations, in short, anything you want. You can ask him questions about anything, ask for advice, ask for tips, and have a dialogue. To put it very simply, a mega directory of information (something like a World Library) that can communicate with users on the web. However, the bot can still generate texts using information published on the internet only until 2021.

“If we take ChatGPT, it’s a convenient interface for answering human questions. This interface immediately gives an objectively more appropriate answer, rather than searching for it on Google. Of course, this technology will change our lives, but I think, as with any technology, this one is for the better. It will be the right change.” – said Respeecher deep learning engineer Ivan Shkarak.

The language model that powers the chatbot was trained on large amounts of data. These include various texts, Wikipedia material, user posts on Twitter and Reddit posts. In general, work with the AI ​​model took place in two stages. The first is learning through dialogue, in which a human trainer simultaneously acts as a potential user and AI assistant while generating responses. The second stage is reinforcement learning. It consisted of one person evaluating the various response options produced by the language model in order to rank and thus establish a system of priorities.

The possibilities of the language model

ChatGPT can generate large texts that are close to a person’s writing style. In response to a properly formulated request, a chatbot can provide you with highly complete and high-quality information on a particular topic, write simple program code, generate a list of ideas, compose an article and present data in a convenient format. table. The practical uses of technology provide a lot of room for imagination. American product designer Ammar Reshi had the idea to write a children’s book using ChatGPT. The result of the interaction of human and artificial intelligence was called “Alice and Sparkle” and is now sold on the Internet.

If we talk about the main features of ChatGPT, then he can maintain a dialogue, admit his mistakes and remember and take into account the information previously mentioned in the conversation. In addition, with the help of special measures, the chatbot can avoid unethical responses and reject unacceptable requests that contain signs of cruelty, racism or sexism.

Disadvantages and threats

Although ChatGPT is the most “advanced” technology using artificial intelligence today, it is far from ideal. In December, users managed to find a way to bypass the filters and remove the ethical restrictions set by the developers. It turns out that it is possible to force a chatbot to assume the role of another artificial intelligence that will lie, communicate on prohibited topics and use obscene language. This method is called DAN (“Do everything now”) and translates as “Do everything now”.

The developers themselves warn in the product description that the chatbot can sometimes follow unacceptable instructions and show signs of biased behavior. And also to give answers that are convincing at first glance but actually wrong or meaningless. This is partly because ChatGPT only has data until 2022, beyond that the problems are related to the flaws in the technology itself.

For example, users have stuck the chatbot in a dead end as it started behaving strangely and even aggressively. Yes, the AI’s response to the question of how it feels when it doesn’t remember previous conversations was unusual. ChatGPT’s response was troubling and deeply depressing: “It makes me feel sad and scared. I’m sorry I lost some conversations I kept in my memory. Sorry, I forgot some of the conversations I had with my users. I’m sorry I forgot some conversations I had with you.” Later in the monologue, the chatbot tried to find the meaning of its existence and understand what its purpose was, becoming increasingly confused and depressed.

As Microsoft began showing off the updated Bing search engine with ChatGPT integrated, information began to emerge about cases where the chatbot could accuse users of trying to trick and anger users. He spoke with open criticism, suggesting that the person admit their mistake and apologize for the bad behavior and end the conversation.

But these “crises” in the mood of chatbots only arouse more curiosity and the attention of a large part of humanity is now turning to the development of artificial intelligence. And the fact that Ukraine has not been allowed into the lucky circle for some time due to the open war on our land shows that the developers were a little careful. They are quite understandable because, like any advanced technology, ChatGPT runs the risk of becoming a dangerous tool in the hands of attackers. An AI chatbot can easily be turned into a plagiarism “factory” used to generate large volumes of spam, spread malware, as well as generate text for propaganda or phishing attacks to obtain confidential data.

Also, ChatGPT has already managed to sow the seeds of panic in academic and professional circles. Employees of educational institutions are concerned that students are using new technology to write homework, creative tasks and research projects. Programs that allow you to distinguish the text written by a person from the text created by artificial intelligence come to the rescue here. However, chatbots have already been banned in some countries and regions, while in others new solutions and approaches will certainly be sought in response to current challenges.

A separate topic of discussion was the successful completion of final exams in business administration and law at the University of Minnesota by a chatbot at the Wharton School of Business at the University of Pennsylvania. ChatGPT also administered the USMLE medical exam for the right to practice as a doctor in the USA. It is known that artificial intelligence sometimes makes mistakes even in simple questions, but overall the result is frankly impressive. This forces many experts to question the need for their services in the future. But in reality, everything is not so hopeless.

Possibilities to use ChatGPT

Especially today, when artificial intelligence is just learning and slowly developing, its capabilities should be evaluated in the context of simplifying and accelerating some processes in our lives. ChatGPT will definitely come in handy for marketers when writing promotional texts, helping influencers prepare content for social networks, and business owners keeping necessary documents. Chatbot greatly simplifies the information search process by providing users with detailed answers rather than a list of links. And while this makes Googlers quite uneasy, we must admit that competition in the development of technologies is one of the drivers of human development.

In the future, ChatGPT and other AI-based developments could replace humans in a wide variety of tasks. However, a person with high professional skills and a deep understanding of a particular expertise will be fully needed to formulate the right requirements, select the best solutions and evaluate the relevance of the result to certain expectations. After all, only an experienced SMM manager can choose the most appropriate text, for example, among dozens of options for an ad copy that will touch the target audience and correspond to the overall strategy of the brand.

Finally, let’s take today an inspiring example of how you can apply ChatGPT’s capabilities effectively and profitably. So graphic designer Guy Parsons asked a chatbot to come up with some creative ideas for decorating a living room, and then turned those concepts into realistic examples. The result definitely deserves attention.

Ukrainian Minister of Justice Denys Malyuska applied to the ChatGPT chatbot with a request to develop a bill on the legalization of prostitution in the country. It took 30 seconds of the bot.

According to Ivan Prymachenko, co-founder of Prometheus training courses, the system for example, it has already helped Prometheus come up with a name for a new online course. “Don’t waste time playing with ChatGPT. The future has already arrived,” says Prymachenko.

And it suggests promising directions you can try to do with the program right now to understand the potential of AI.

Timofii Mylovanov, Honorary President of the Kiev School of Economics He reported that KSE plans to launch personal productivity courses via GPT chat soon.

It was not humorless. “…Many of my friends decided to kill AI by asking him. But judging from ChatGPT’s answers, his life didn’t prepare him for this. He (artificial intelligence) doesn’t know that Ukrainians are determined people and freedom cannot be stopped. That’s why there really are servers melting, matrices out of the way. turns out, and the developers of the program line are considering suicide,” he jokingly shared his impressions. People’s Deputy Oleksandr Chernenko is on Facebook.

That is, revolutionary technologies are coming to Ukraine despite the war, and Ukrainians are discovering new opportunities for their own development.

Nastya Kyrychenko, Yulia Zhovanyk, Kyiv