OpenAI development - ChatGPT - gained 100 million users in two months. Ukrainian business has also become an active user of the platform. However, this artificial intelligence technology requires detailed study.
ChatGPT is a search engine capable of responding to a user's query with an open answer, rather than a simple list of sites. The chatbot works on the language model of artificial intelligence GPT3.5. This algorithm, the developers assure, can perform any tasks. The success of ChatGPT is explained by the ability to respond to human language as closely as possible and, no less important, to understand the interlocutor contextually. The chatbot got such capabilities from the language model (LLM), which was trained with the help of huge amounts of data.
"Historical moment - we decided to hire ChatGPT instead of an SMM manager at 3DLOOK," the founder and CEO of the startup Vadym Rogovsky wrote on Facebook.
The functions of the chatbot from the OpenAI company are the generation of articles, letters, computer code, etc. In two months, 100 million people used it, and the daily number of unique visitors was 13 million. Strategic investor Microsoft presented the integration of the "new generation" of ChatGPT in its Bing search engine and Edge browser. So far, use is limited to a dozen prepared questions.
Artificial intelligence may become the hottest topic of 2023, just like the creation of the PC, the graphical interface, and the Internet. "I think artificial intelligence can join the fourth of these three most important discoveries in digital technology," says Microsoft co-founder Bill Gates. Is Ukrainian business ready to master breakthrough technology during the war?
Cooperation of artificial and human intelligence
It will not be possible to completely replace the SMM manager with artificial intelligence. ChatGPT writes posts adapted to social networks, but the selection of photos and technical tasks are on the shoulders of a digital marketer.
With artificial intelligence, business communications will become more personalized and human, predicts Alisa Kyrychok, the founder of the communication agency EmailConsulting.top. "At the same time, the project will be worked on by two or three highly qualified "bot drivers" who know how to set tasks correctly," she says.
But before making artificial intelligence a part of business processes, companies need time for testing in different departments. Chat capabilities can save time for marketers, analysts, PR people, writers, SMM-niks, recruiters, believes COO Genesis Artem Kopanev. But ChatGPT's responsibilities are still small - maintaining the Twitter page of the Genesis corporate blog.
"According to some data, ChatGPT can optimize up to 70-80% of routine tasks," says Andriy Lazorenko, co-founder and CEO of IdeaSoft. From 2021, the Kharkiv outsourcer is part of the Ukrainian-Swedish IT company Sigma Software Group with more than 2000 specialists.
The parent company globally is in no hurry to delegate work to artificial intelligence, but IdeaSoft from Sigma Software is testing ChatGPT as a tool for writing small pieces of code or finding bugs.
The basic skills of a chatbot allow people without knowledge of programming languages to "enter IT". The director of the Kyiv School of Economics, Tymofiy Mylovanov, wrote a Python service for himself, which selects 10-30 of the most important articles per day and makes extracts from each. "ChatGPT explained to me how to install it and work with the underlying codes," he wrote on Facebook.
But this is not enough for professional developers. "It is not enough for complex programming languages, such as Rust or building correct and complex architecture, business programs or game development," says Lazorenko.
Do not scare the people of Haiti with automatic code - since 2021, they have had another product from OpenAI at their disposal - Copilot. With terabytes of source code on GitHub, he learned how to automatically complete the code. The fate of the tool is complicated - a court case was initiated against its creators for violating the copyright of developers who published their open source code on GitHub.
Such tools help to automate the routine of developers and free their hands for creativity. "And further, it will depend on each specialist individually whether he can use the freed time for skills in creating strategies, creative and critical thinking, which are not subject to AI," says Lazorenko.
Unreliable assistant
The praise of ChatGPT goes hand in hand with sanctions against it - it has already been banned from schools in New York state. In practice, cybercriminals have also successfully mastered the tool.
A chatbot can be a nuisance to a business. For more than 10 years, Ukrainian Grammarly has been developing software for checking literacy. More than 30 million people use the product every day, the company reports. But will there be demand if the text is written by AI? "Improving generative AI technologies opens the way to support other stages of communication, such as creating a concept of a message or interpreting it," answers head of product Rahul Roy-Chowdhury.
But for now, business has time - ChatGPT is still wrong. Even the Bing search engine, which will work on an improved version of GPT3, has not passed fact-checking.
"ChatGPT can provide any answer, even when there is not enough data for this," says Hleb Dobzhanskyi, VP of Engineering at IT company Master of Code Global. – To do this, it is necessary to additionally train the model for specific and updated information, and to control and filter the AI responses.”
Globally, this threatens the emergence of deepfakes and the spread of false news. Relying on it for important business decisions is unreliable.
Another nuance is data security. Is the information a user provides to ChatGPT secure from hackers? "With the help of similar services, it is better not to work on sensitive financial data of the company or compose letters to clients with project details protected by NDA," says Roman Mykhailyshyn, Global Head of Intelligent Automation CoE at Ciklum.
Threats of ChatGPT
Daron Acemoglu, a professor of economics at the Massachusetts Institute of Technology, believes that a bot with artificial intelligence threatens disaster not only for employees, but also for consumers and even investors.
The problem for workers is clear: the number of jobs that require strong communication skills will decrease, and therefore the number of positions that pay well. Cleaners, drivers and some other manual workers will keep their jobs, but the rest should beware.
“Think about customer service. Instead of hiring humans to interact with customers, companies will increasingly rely on generative AI like ChatGPT to soothe angry callers with intelligent and comforting words. Fewer entry-level jobs will mean fewer opportunities to start a career—a continuation of a trend set in motion by earlier digital technologies.
Consumers will also suffer. Chatbots are good for solving completely routine questions, but the problems that drive people to call customer support are not routine. When a real-world situation arises, such as an airline flight cancellation or a pipe burst in your basement, you want to speak with a highly skilled, responsive professional who can allocate resources and generate timely solutions. Naturally, you don't want to spend eight hours waiting for an answer, but you also don't want to waste precious time on a verbose, but ultimately useless chatbot, either," says Acemoglu.
