Purely intellectual murder. Why is the chat bot ChatGPT called the "Google killer"?
January 08, 2023
With the advent of ChatGPT, experts predict global changes in the information search process. However, today the new technology is too expensive.

At the end of November last year, the non-profit organization OpenAI, specializing in the development of artificial intelligence technologies, presented to the general public a chatbot ChatGPT using neural network technology. The new chatbot generates meaningful responses to the proposed queries, and according to many users, the quality of responses from the neural network turned out to be higher than that of the Google search engine. The tech giant saw this as a direct threat to his business. On December 22, Google announced a "code red". The company is discussing the development of AI platforms as a priority.

Musk has inherited again
ChatGPT surprisingly easily copes with the most incredible tasks assigned to it by network users: it issues competent historical references, explains mathematical theories and even writes poems about cryptocurrency. ChatGPT attracted 1 million users in the first 5 days of its operation alone. Such a success of the project surprised even its creators, as the head of OpenAI Sam Altman admitted on Twitter.

"There are a lot more requests to ChatGPT than we expected; we are working on increasing capacity," he said.

OpenAI was founded by Sam Altman and Elon Musk seven years ago – they set a goal to develop artificial intelligence technologies that can "benefit all of humanity." In 2018, Musk withdrew from the project due to disagreements about its further development, but now he also paid tribute to the ChatGPT capabilities.

ChatGPT was created on an Azure AI supercomputer based on the GPT-3.5 language model from OpenAI. The chatbot was trained using an array of texts from the Internet and a reinforcement learning system based on human feedback Reinforcement Learning from Human Feedback. Then the neural network was retrained several times, using its own answers to make them more accurate and correct.

Red danger level
The new smart bot gives simple and understandable answers to user questions, not limited to a set of links, and the quality of answers, including complex questions, is higher in the neural network than in the Google search engine. Moreover, he is able to generate various ideas from business strategy to topics for blog posts. ChatGPT analyzes information from millions of sites and gives answers in the style of live human speech. 

The developers stated that they aimed to make AI easy to use, correct and "human". ChatGPT is still "damp" and far from perfect, which the creators themselves admit. But even in its current form, it poses a tangible threat to Google.

As The New York Times learned, the company's management was forced to declare a "code red". Google CEO Sundar Pichai took the management of the AI product development strategy into his own hands and personally engaged in the restructuring of workflows. There is an opinion that the search giant is on the threshold of what the largest players in Silicon Valley are so afraid of – global technological changes that can destroy their business. For more than 20 years, the Google search service has served as the world's largest entry point for users to the Internet – this business brings the company 80% of profits, and it does not want to lose such an asset.

He is wordy, likes to lie and is very expensive
Experts believe that ChatGPT technology will be able to push traditional search engines out of the market. However, not everything is so simple. So far, such chatbots do not allow the introduction of targeted advertising, which brings the main income to search engines. In addition, ChatGPT is an expensive technology.

By the time the neural network attracted more than 1 million users, Musk asked the head of OpenAI, Sam Altman, how much the chatbot's work costs. Altman replied: 

"Probably a few cents per correspondence; we are trying to find out more precisely, and also looking for ways to optimize." 

He also noted that sooner or later ChatGPT will have to be monetized, as the cost of computing is "impressive". Third-party estimates say that the work of the neural network requires $ 100 thousand per day or $ 3 million per month. 

And as already mentioned, the technology is still far from perfect. A chatbot can make mistakes and give out false information, that is, ChatGPT does not guarantee the accuracy of the information, although even incorrect neural network responses look "convincing". According to users, the bot often "fantasizes" and is too verbose. And yet, ChatGPT does not suffer from political correctness, but is it a disadvantage?