ChatGPT

The main articles are:

• Chatbots (Chat-bot) Virtual interlocutors
• Speech Recognition (Technology, Market)
• Speech technology: On the path from recognition to understanding
• Neural networks (neural networks)

ChatGPT is an artificial intelligence chatbot that can work online.

ChatGPT can support natural language dialogue, write poetry, generate texts and program code, translate, answer all kinds of questions as a person, etc.

2024

Attackers use ChatGPT to create malware

OpenAI reported the suppression of 20 malicious operations in which the company's generative artificial intelligence (AI) was used for illegal actions, according to a report published by OpenAI. This was announced on October 15, 2024 by the press service of Anton Nemkin, a member of the State Duma Committee on Information Policy, Information Technology and Communications.

OpenAI clarified that in order to determine the key areas of use chat boat for illegal purposes, OpenAI analysts analyzed the key cases of 2024. So, often the chatbot was used to create, harmful software generate fake content phishing and send out. For example, cybercriminals they used - AI technology to write scripts used as part of a multi-stage chain of device infection.

The mission of OpenAI is to ensure that artificial intelligence benefits all of humanity. We are committed to identifying, preventing and suppressing attempts to abuse our models for harmful purposes. We know that in this global election year, it is especially important to build robust, tiered defenses against state-linked cyber attackers and covert influence operations that may try to use our models to conduct misleading campaigns on social media and other internet platforms, the company said in a statement.

Artificial intelligence copes well with the tasks set, said Anton Nemkin, a member of the State Duma Committee on Information Policy, Information Technology and Communications. "The key problem that the global IT community is facing today is that AI is doing an excellent job of implementing illegal actions. The number of IT crimes committed with the help of neural networks is steadily growing, "the deputy said.

Even an inexperienced attacker can create a phishing newsletter. According to Kaspersky Lab analysts, the number of shadow Telegram channels has recently grown, it is possible to find ready-made scripts for creating fake mailings. In addition, the number of personalized phishing emails is increasing. So, AI, based on open data, analyzes the interests and consumer habits of the user, on the basis of which it subsequently forms a phishing mailing with a user proposal, the deputy explained.

This is not to say that the vast majority of attackers use AI to implement attacks, but the popularity of the tool is likely to only grow. Existing models of generative artificial intelligence contain protective mechanisms: for example, if the code contains a malicious architecture, then the chatbot will refuse to execute the assigned command. Of course, this complicates the life of intruders, but definitely does not deprive them of the opportunity to use the tool, the deputy emphasized.

It is necessary to develop systems of trusted artificial intelligence, Nemkin is sure. Trusted artificial intelligence, in addition to uninterrupted operation and technical stability, is based on the norms of morality and ethics shared in society, and also complies with the provisions of legislative acts.

At least 12 research centers are developing trusted AI in Russia, the deputy emphasized.

A chapel has started opening at any time with a smartphone

At the end of September 2024, it became known that a unique chapel was opened in Poland, which can be accessed at any time of the day using a smartphone application. Another feature of the construction is the use of artificial intelligence tools. This is said to be an innovative way to combine spirituality and advanced technology. Read more here

The number of users of corporate versions of ChatGPT exceeded 1 million

The number of paid users of corporate versions of ChatGPT has exceeded 1 million people. The company OpenAI told about this in early September 2024. They also explained that the named figure includes the total number of people who registered to use its ChatGPT Team and Enterprise services, as well as people at universities using its ChatGPT Edu product.

The company introduced the ChatGPT Team subscription for small teams in January 2024, and launched the Enterprise version in August 2023.

Audience growth up to 200 million users per week

In August 2024, the number of ChatGPT users exceeded 200 million people. That is how many people, according to OpenAI, use the service at least once a week. Over the year, the size of the audience of the neural network has doubled.

Now people are using our tools as part of their daily lives, making real changes in areas like health and education, "said the CEO of OpenAI. Sam Altman

It is noted that by August 2024, 92% of Fortune 500 companies, a list of the 500 largest enterprises, USA use OpenAI products.

Top 10 ChatGPT analogues available to users in Russia

Generative artificial intelligence (GENI), capable of creating new content and ideas, including conversations, stories, images, videos and music, penetrates many areas of activity and business processes. One of the most popular chat bots based on Genia is ChatGPT, developed by OpenAI. However, this service is not available to Russian users. TAdviser has prepared a list of ten AI systems that can become an alternative to ChatGPT when performing certain tasks. You can use these chat bots, including from the Russian Federation. Read more here.

Brazilian courts start using ChatGPT to save money

In mid-June 2024, the Brazilian Government announced that the judiciary would use OpenAI services, including ChatGPT, to speed up the review and analysis of thousands of lawsuits in order to avoid costly legal costs that place a heavy burden on the federal budget. The AI system will identify lawsuits that require immediate intervention, as well as analyze trends and potential impact zones, sending the results to the attorney general. More detailed here.

52% of ChatGPT answers to programming questions are incorrect

In May 2024, researchers at Purdue University presented the results of a study that found ChatGPT answered more than half of the programming-related questions incorrectly. At the same time, the chatbot responded so confidently that it convinced more than a third of the study participants.

According to The Register, the team of scientists analyzed ChatGPT answers to 517 questions asked on the StackOverflow service and compared them with real answers from programmers. The answers were checked for correctness, consistency, clarity and conciseness.

ChatGPT incorrectly answers more than half of programming-related questions

Our analysis shows that 52 percent of ChatGPT responses are incorrect and 77 percent are too detailed, the team concluded. However, ChatGPT responses are still preferred in 39.34% of cases due to their completeness and well-formulated language style.

Scientists noted that when an error in the chatbot's response is obvious, users identify the error, but when the ChatGPT response is difficult to verify or additional software or documents are required to verify, users often cannot determine whether the response is correct.

Moreover, even when the error is obvious, two out of 12 users still chose the chatbot's response. One of the main reasons was how detailed ChatGPT's responses were. In many cases, participants were not confused by the volume if they received useful information from long and detailed responses. In addition, two other reasons were positive sentiments and politeness of responses.

The authors note that ChatGPT responses contain more 'motivational attributes' - formulations that suggest achievement - but do not describe risks as often as stack overflow messages. So, ChatGPT often uses phrases like 'Of course I'll help you' or 'This will definitely help'.

A linguistic analysis of responses from ChatGPT and StackOverflow found that bot responses are more formal, analytic, show more effort to achieve goals, and show less negative emotions. Their sentiment analysis found that ChatGPT's responses are more positive than StackOverflow's.^[1]

A computer mouse with ChatGPT inside entered the market

On April 17, 2024, the company Logitech announced the Signature AI Edition M750 computer mouse. Its feature is a special button for quick access to the Logi AI Prompt Builder tool, which, among other things, allows you to interact chat boat ChatGPT with the based. artificial intelligence More. here

2023

Russian Defense Ministry demanded to ban ChatGPT

At the end of December 2023, the Ministry of Defense of the Russian Federation announced the need to prohibit the use of generative artificial intelligence in critical areas. According to the department, such tools can be used to collect certain information for intelligence purposes, to organize cyber attacks based on advanced social engineering methods, as well as for "cognitive impact on the population."

We are talking about services like ChatGPT, Midjourney, Dall-e 2, StableDiffusion, etc. They can generate different types of content - audio, video, text and images. According to Nikolai Lishin, Deputy Director of the Department of Information Systems of the Ministry of Defense, such algorithms are not capable of producing objective results. Generative AI causes a number of ethical problems, especially in terms of the authenticity and integrity of the content being created. Deepfakes, for example, can be used to spread misinformation.

Generative AI models have "intrinsic content from the programmers" who create them, Lishin said. As a result, the "worldview" of such algorithms coincides with the worldview of westerners aged 25 to 35, and their users find themselves in a "cognitive trap."

The Ministry of Defense, according to CNews, calls the systems of generative AI dual-use technologies, comparing them with nuclear weapons. It is thanks to the presence of nuclear weapons that Russia retains the status of a superpower, and therefore, the department believes, it is "extremely important for the Russian Federation to develop its own technologies." In general, the Ministry of Defense believes, in Russia it is necessary to exclude the introduction and use of foreign large language models in "critical" areas, such as decision support platforms, medicine, military affairs, CII facilities, etc.^[2]

Implementation in the application "Majordom"

The IT company "Philosopher," which creates digital products for real estate, is introducing ChatGPT into the mobile application for smart homes "Majordom." In the first iteration of the implementation, a function will be available for the management company. This was announced by "Philosopher" on December 21, 2023. Read more here.

The world's first computer mouse with ChatGPT support has been released

In mid-November 2023, a smart wireless manipulator called Cheerdots2 debuted. The developers call the device the world's first computer mouse with support for a chatbot ChatGPT based on. artificial intelligence More. here

ChatGPT Integration into RocketData Platform for Comprehensive Reputation and Online Brand Presence Management

On September 7, 2023, RocketData announced that business work with content in geoservices and social networks will now become easier thanks to the integration of ChatGPT into the RocketData platform for comprehensive reputation management and online brand presence. The neural network will generate simple posts and is designed to free up time for company employees to perform global tasks. Read more here.

Implementation in Trivio Work Travel Management Service

Trivio released an update to its online travel service on June 21, 2023, introducing a virtual assistant based on ChatGPT technologies. The added functionality will allow customers to quickly receive collections of hotels, information about cities where to eat, where to go and even life hacks - how to prepare for a business meeting or write a speech. Read more here.

ChatGPT accounts began to enjoy increased interest among hackers

Group-IB (Russia, Singapore) on June 20, 2023 informed TAdviser that in recent months cyber crime has shown increased interest in user accounts in the chat bot system based on OpenAI ChatGPT artificial intelligence. The details of access to them over the past year have begun to be actively sold on markets on the darknet.

In total, Group-IB experts identified 101 134 devices with logins saved on them passwords and to ChatGPT and infected Trojans with -infostilers (that malware steal data access details).

As of May 2023, Group-IB experts estimated that 26,802 stolen data arrays containing, among other things, information about ChatGPT accounts are in a one-time turnover on criminal markets.

40.5% of this data was taken away from users from the Asia-Tacean region. In general, this is where the largest number of accounts compromised by information dealers falls - 40,999, where they hold unconditional leadership. India There countries CIS are only 754 such incidents.

Information dealers collect a variety of data: information about, credit cards details of access to crypto wallets, files cookies, history browsers and other information that can be stored in browsers on compromised systems, including details of access to ChatGPT.

These programs are not selective: their task is to collect as much of any information of value to cyber crime as possible and send them to a control server, where they will be sorted in subsequent stages. Sometimes the data collected in this way is sold on the darknet directly unfiltered, only the domains found in the logs and information about the IP address of the compromised host are highlighted.

It was these data arrays that were studied by Group-IB experts, who discovered an abundance of logins and passwords for ChatGPT accounts in the logs of information dealers. The publication of Group-IB does not directly say that it is the data from the negotiations between users and the chatbot that is put up for sale, but when their accounts are hacked, attackers, as a rule, gain access to everything at once.

Many organizations integrate ChatGPT into their daily operational processes. Workers sometimes share secret correspondence with an AI bot or use it to optimize proprietary code. Considering that in its standard configuration ChatGPT stores the contents of all dialogs, for attackers this can mean a storehouse of critical information, to gain access to which it is enough to seize the details of ChatGPT. We at Group-IB constantly monitor underground communities to identify such accounts in a timely manner, said Dmitry Shestakov, head of cyber threats intelligence at Group-IB.

Group-IB recommends regularly changing passwords for ChatGPT and implementing two-factor authorization.

ChatGPT also allows you to disable data saving or manually delete call logs with it. Accordingly, it will be best practice to immediately delete logs where some important technical or business information may be present that should not fall into the wrong hands. As you can see, the decisions of some of the world's largest companies to ban the use of ChatGPT for security reasons turned out to be a sound decision. noted Anastasia Melnikova, Director of Information Security at SEQ.

The expert added that the observed increase in interest in chat bots will inevitably be accompanied by an increase in interest in cyber crime, and this will affect not only ChatGPT.

Fraudsters in Russia began to use ChatGPT in phishing attacks

Fraudsters in Russia began to use ChatGPT in phishing attacks. This was announced in mid-June 2023 by the managing director of Kaspersky Lab in the Russian Federation and the countries of SNGANNA Kulashova.

Artificial intelligence is a tool, and as any tool it carries both capabilities and threats. And, of course, it will be used from all sides, including we have already seen attacks using ChatGPT, "she said in a conversation with RIA Novosti.

According to Kulashova, the neural network is used for targeted phishing when attacking an organization. Such actions can be resisted, in addition, the use of such technologies remains expensive by mid-June 2023.

Previously, it was possible to distinguish phishing email by its usual speech literacy: as soon as you understand that the letter is strangely composed, there are errors, and so on, it means phishing. With the use of ChatGPT, errors in these letters have become much less, - said the managing director of Kaspersky Lab in the Russian Federation and the CIS countries.

T.Hunter confirms that scammers have begun using ChatGPT for phishing. This allows you to automate the mailing process, increase the reliability of texts and, accordingly, the likelihood of deceiving users, experts say.

Earlier in 2023, the head of OpenAI, which developed ChatGPT chatbot Sam Altman, said that the scope of AI needs to be regulated, since if something "goes wrong with the technology, everything can go completely wrong." He noted that, according to the developers, the advantages that AI gives exceed the risks, but ensuring security is important for the activities of developers.

Altman proposed creating an agency that would license the development of large AI models, develop rules and testing for them.^[3]

ChatGPT is increasingly leaking important data

By May 2023, several cases of data leaks due to the use of the ChatGPT neural network are already known. The company itself, OpenAI developing the project, confirmed the existence of the problem.

In March 2023, ChatGPT users began reporting en masse that their accounts displayed other people's chat stories. Moreover, according to PC Magazine, some ChatGPT Plus subscribers also saw other people's email addresses on their subscription pages.

As a result, the OpenAI developers were forced to temporarily disable this functionality in order to fix the error. At first, the company reported that the cause of the failure was related to unnamed open source software used by OpenAI, and stressed that due to the bug, people saw only the headlines of other people's conversations, but not their contents.

Subsequently, OpenAI said that the cause of the ChatGPT data failure and leak was an open source library error Redis. The leak affected data from 1.2% of customers. The company called the figure "extremely low."

In April 2023, it became known that the corporate use of ChatGPT by Samsung employees to solve official tasks turned into a major data breach. The South Korean electronics manufacturer has deployed artificial intelligence to look for problems in the source code of its products. Thus, confidential data got into the public domain, becoming part of ChatGPT's own database.

At the same time, according to some experts, cyber incidents with ChatGPT and other chat bots are mistakenly called data leaks. The fact is that in the usual terminology, leaks come from those systems in which the data is protected. But if a user transfers confidential information for processing to ChatGPT, this can be compared to uploading the same data to a public table from Google.^[4]

China first arrested for fake news written by ChatGPT

On May 9, 2023, it became known that the first arrest of an Internet user who used the ChatGPT chat bot based on artificial intelligence to write fake news was made in China.

A man named Hong is suspected of violating the law, which, in particular, prohibits the illegal use of AI services and neural networks to disseminate false information. According to law enforcement, Hong, through the ChatGPT bot, generated articles describing the railway disaster. These materials said that passengers were killed in the train crash.

In total, about 20 slightly different fake notes were found indicating different locations of the incident: by changing the presentation, Hong bypassed the means of detecting recurring content. All the notes were made public on the Baijiahao blog platform, which is owned by Baidu. In total, these fake news received more than 15 thousand views. In such a fraudulent way, the suspect made money from traffic.

It is emphasized that ChatGPT is blocked in the PRC, but you can access the chatbot through a virtual private network (VPN). Hong was arrested by police in north-central China's Gansu province . The man has already confessed to illegal use of ChatGPT. He faces charges of provoking conflicts. In addition, a man can be accused of violating public order and (or) creating riots in public places. The wording of the relevant article is rather vague: it has been repeatedly criticized for the fact that it can strangle freedom of speech on the Internet and be used to arrest activists who criticize the Chinese government. Defendants could face five to ten years in prison.^[5]

Creating a Voice Communication Service

On April 19, 2023, the 3iTech Company (DSS Lab LLC) announced the creation of a service that allows you to communicate orally with ChatGPT in Russian. Integration with ChatGPT is made in the form of a telegram chat bot 3i Olivia. Read more here.

Implementation in the Internet bank "Points"

Bank Point On April 19, 2023, he announced that he had implemented GPT in the search string Internetbank-. The service will help the entrepreneur get answers to the most common and important business questions. here More.

Using in the Sunlight app

On April 7, 2023, SUNLIGHT announced that it had begun generating release notes for its mobile application using the ChatGPT neural network. Read more here.

Implementation in the Point service

On April 3, 2023, the Pointer service announced that it had built a chat bot with artificial intelligence ChatGPT from OpenAI into its feedback functionality. Feedback processing will now take less time and users will get personalized responses faster. Read more here.

Fraudsters in Russia began to use ChatGPT for phishing

Fraudsters in Russia began to use ChatGPT for phishing. T.Hunter information security specialists told about this at the end of March 2023. According to experts, cybercriminals are actively using AI capabilities in order to increase the accuracy of texts, automate the process and increase the likelihood of deceiving users. Read more here.

British newspaper used ChatGPT to write news and became frustrated with results

ChatGPT chat bot based on artificial intelligence, developed by OpenAI, is not yet suitable for the role of a reporter. This, as it became known in early March 2023, is evidenced by the results of an experiment conducted by a British newspaper.

The publication, which requested anonymity, gave the ChatGPT bot the task of preparing material about the brutal robbery of an elderly woman in Buenos Aires. At the same time, using the provided source text, it was necessary to add names, specify a specific scene of the incident and insert other details.

As noted, the news generated by the neural network at first glance looked rather tolerable - especially given the fact that artificial intelligence independently figured out how to structure a report with a sharp introduction, short sentences and a consistent presentation approximately corresponding to the inverted pyramid.

But the subsequent analysis of the text, performed by the editor, revealed many inaccuracies and dubious moments. It turned out that ChatGPT incorrectly indicated the name and age of the victim, as well as the crime scene. In the generated news, it was indicated that the offender is unknown and remains at large, but in fact he was identified and arrested. Probably, quotes from the mayor of Buenos Aires (when specifying the name of which an inaccuracy was also made) and the children of the victim were fabricated. In addition, the description of the existing video material did not correspond to the personnel.

As a result, the newspaper concluded that ChatGPT in its current form is not suitable for writing full-fledged news. On the other hand, many publishers are considering using this chatbot to generate certain content using strictly controlled source material. This could be, for example, the processing of any statistical data.^[6]

The US Air Force told how ChatGPT will help the Pentagon

At the end of February 2023, US Air Force Chief information officer Lauren Barrett Knausenberger announced that the ChatGPT chatbot could improve the work of the US Department of Defense. In particular, according to her, artificial intelligence is able to simplify document management and synchronization of tasks between all military units.

Knausenberger believes that AI can help quickly piece together information, such as finding out who is responsible for organizing in a single operation. ChatGPT will be able to perform the duties of three officers who process and analyze information at once. US officers can be entrusted with another front of work, which will save a significant amount of time and effort for the army, she explained.

By early March 2023, Knausenberger is about to leave the post of Chief information officer of the US Air Force. But before leaving, she promises to complete a project in 2023 to simplify, automate and comprehensively synchronize the devices used by the military in her department. Including this problem will be solved thanks to AI, as well as the integration of the ChatGPT chatbot functions. The chief information officer reports that the wider the computer infrastructure, the more difficult it is to manage and synchronize between devices.

According to Defenseone, the Pentagon plans to introduce AI into projects such as financial and audit accounting by 2025. The U.S. Air Force aims to bring its various services together in a single tool stack. And another of the urgent tasks that they will try to solve at the expense of AI is to optimize incoming and outgoing traffic between all departments at the US Army military bases scattered around the world.^[7]

UAE introduces ChatGPT-based online tutors in schools

In early March 2023, UAE Education Minister Ahmed Belhul Al Falasi announced the launch of AI-based online tutors across the country in collaboration with global tech giants. Read more here.

Growth in demand for ChatGPT specialists in Russia by 13 times

In Russia, the number of vacancies on HeadHunter, where skills with ChatGPT are required, in February 2023 amounted to 38, which is 13 times more than a month earlier. The recruiting service announced this in March 2023.

As Vedomosti writes with reference to a representative of HeadHunter, according to ChatGPT, the main hiring is from IT companies and companies from the financial sector, and programmers, PR specialists, Internet marketers and analysts are indicated as specialties.

The developers themselves rarely need the skill to work with ChatGPT and similar sites, this is a skill primarily for "content writers, editors," a recruiting expert and founder of Facancy Alyona Vladimirskaya told the publication. With the help of neural networks, they are trying to make bots, texts, SEO analytics, she added.

The skill of owning the same ChatGPT will most likely not be affected by the salary level, the newspaper notes. For example, the average salary of a copywriter both with and without the mention of a chatbot, hh.ru is at the level of 30,000 rubles. The developer of the Center for Artificial Intelligence "SKB Kontur" Dmitry Ivankov noted that it is necessary to distinguish between employees who know how to work with neural networks as a developer (create, train and use neural networks when creating various services) and as a user. The demand for the former, according to him, is consistently high, for the second category of workers, it began to grow in 2022.

According to HeadHunter, the number of vacancies where the skill of working with neural networks is needed increased by 62% in six months - from 312 ads in August 2022 to 506 in February 2023. Also among the services based on neural networks that applicants should be able to work with, most often the Colorize application is indicated, which colors black and white photos, as well as DALL-E 2, Midjourney and Stable Diffusion, which generate images based on text descriptions.^[8]

The emergence of paid APIs 

On March 2, 2023, OpenAI allowed developers to integrate ChatGPT into third-party applications and services through the API. The first third-party customers to take advantage of the released APIs were Snap, Quizlet, Instacart and Shopify.

The cost of using the ChatGPT API for integration into neural network models of developers is $0.002 per 1,000 tokens - about 700 words, according to the OpenAI blog.

We listened carefully to the feedback from our developers and improved the API service terms to better meet their needs! We believe that AI can provide incredible opportunities for everyone, and the best way to achieve this is to allow everyone to create something with it, the developers from OpenAI report.

According to OpenAI President Greg Brockman, the API is based on the same AI model as the "gpt-3.5-turbo" version of ChatGPT. GPT-3.5 is the most powerful text generation model OpenAI offers for March 2023; the turbo version is more optimized and responsive.

According to OpenAI, in order to avoid incorrect and false answers, the company has implemented the ChatML markup language. It passes the ChatGPT API text as a sequence of messages along with metadata. In addition, developers will begin updating ChatGPT more often with the release of "gpt-3.5-turbo." This will give users more control over the load of allocated capacity. Customers will be able to use "gpt-3.5-turbo" with a contextual window weighing up to 16 KB, which will process four times more tokens than the standard ChatGPT model. This will allow you to insert entire pages of text in the question and receive reasonable answers from the model.^[9]

ChatGPT uses 10,000 GPUs

The ChatGPT AI system has been trained on more than 10,000 GPUs. This was announced at the end of February 2023 by Jensen Huang, CEO of Nvidia, whose technologies are used in the popular chatbot.

As part of a seminar at the Walter Haas School of Business University of California, Berkeley , Jensen Huang shared his thoughts on this development. Similar ChatGPT systems, he said, require a lot of computing power. AI for 2023 already provides business opportunities for GPU chip manufacturers, mainly Nvidia, but AMD Intel Google also produces hardware to ON support deep learning workloads. Such systems require a lot of computing power.

According to Huang's statements, as of March 2023, AI mainly uses processor the Nvidia A100 graphics and DGX A100 platform, which can provide performance at 5 petaflops. So unless AMD or another company has a strong competitor for Nvidia, the green team will remain the biggest player in the AI industry.

According to Huang, the advantages of this ChatGPT technology are that those areas that previously could not use AI will now be able. Drawing a further analogy, Huang said that thanks to the development of ChatGPT, now everyone can actually become a programmer, regardless of their profession.

According to TrendForce, the number of parameters used in training ChatGPT models in 2018 was 120 million, and the number reached 180 billion in 2020. In a published report, the number of GPUs required by the ChatGPT model to process training data reached 10,000. In the future, the number of GPUs required to commercialize the ChatGPT model is estimated to exceed 30,000. However, the exact number of GPUs used for training will depend on the specific implementation and resources available.^[10]

Large corporations have started using ChatGPT for marketing

At the end of February 2023, information appeared about Coca-Cola's decision to use AI tools and chat bots for marketing and communication tasks. In particular, we are talking about ChatGPT, DALL-E and Codex solutions. Read more here.

Start of sales of books written by the ChatGPT neural network

In mid-February 2023, it became known that books written by ChatGPT based on artificial intelligence were available for ordering in the Amazon online store.

According to Reuters, the author of one of these projects is Brett Schickler. According to him, the idea to generate a book appeared after the launch of the ChatGPT neural network, which is able to form texts in a natural language. Schickler, as noted, created a 30-page illustrated children's book called The Wise Little Squirrel: A Tale of Saving and Investing in a matter of hours. The images for the product are also generated by AI.

The book tells the story of the squirrel Sammy, who finds a gold coin and hides it in a piggy bank from an acorn. After that, Sammy opens an acorn trading business and becomes the richest squirrel in the forest. But this causes envy among her friends. The book is offered for $2.99 in electronic version or $9.99 in printed version. The work brought Schickler less than $100, but this turned out to be enough to inspire him to create other books using artificial intelligence.

As of mid-February 2023, the Amazon Kindle store had more than 200 e-books listing ChatGPT as an author or co-author. But due to the features of ChatGPT and the fact that many authors do not disclose information about the use of the bot, it is almost impossible to estimate the real number of works written using this neural network.

ChatGPT seems poised to upend the established book industry as aspiring writers looking to make a quick buck turn to this literature-making software. Hundreds of guides have appeared on YouTube, TikTok and Reddit showing how to make a book in just a few hours, Reuters notes.[11]

Ban on use for Chinese IT giants

On February 22, 2023, it became known that regulators China banned large technology companies from offering services based on the ChatGPT AI chatbot to the general public.

According to Nikkei Asia, Tencent Holdings and Ant Group, as well as the fintech branch of Alibaba Group Holding, were instructed not to provide access to ChatGPT services on their platforms either directly or through third parties. IT companies will also have to report to regulators on plans to launch their own services similar to ChatGPT.

The ChatGPT AI bot, developed by startup OpenAI, is officially unavailable in China, but some Internet users bypass restrictions via VPN. In addition, dozens of mini-programs have been released by third-party developers to use ChatGPT on Tencent's WeChat. Under pressure from regulators, Tencent suspended several such third-party services whether they were connected to ChatGPT or simply emulated that AI bot.

The introduction of new rules banning IT companies from offering services using ChatGPT is dictated by an official backlash regarding the chatbot. In Beijing, anxiety is growing about ChatGPT's uncensored responses to user requests. According to regulators, this service "can lend a hand to the US government in spreading disinformation and manipulating global narratives in its own geopolitical interests."

From the very beginning, we understood that ChatGPT would never be able to penetrate China due to censorship problems, and the PRC would need its own versions of the bot, "said one of the leaders of the leading technology company.

After the emergence of ChatGPT, major Chinese corporations, including Tencent, Alibaba and Baidu, rushed to unveil their own plans to develop similar services. However, these companies were careful to formulate their intentions, noting that their services are similar to ChatGPT, but do not integrate ChatGPT itself.^[12]

Google unveils chatbot competitor

On February 6, 2023, Google unveiled a chatbot called Bard based on artificial intelligence. The system is expected to compete with the ChatGPT service launched by OpenAI. Read more here.

Based on a Telegram bot for creating product descriptions

On February 6, 2023, the company inSales announced that it neuronets ChatGPT had created a database Telegram boat that, based Internet on a simple request, can generate texts for cards of goods sold in stores and marketplaces. More. here

Application audience - 100 million users

In January 2023, ChatGPT reached 100 million active users. It took approximately 2 months^[13].

Hackers began to use neural network to create viruses

On January 6, 2023, Check Point Research released the results of a study that suggests that attackers, including novice hackers and people without any programming experience, are using the advanced ChatGPT neural network to create malicious code.

Cybersecurity experts give several examples of using the ChatGPT chat bot based on artificial intelligence to generate malware. In one such case, the neural network wrote Python code with a number of cryptographic functions: this, in particular, encryption of system files using the Blowfish and Twofish algorithms. Theoretically, such code could form the basis of a ransomware program. In another case, the neural network formed a Python script to steal information: such code is able to search for certain types of files, for example, PDF, copy them to a temporary directory, compress and send them to a server controlled by cybercriminals.

ChatGPT can also reportedly generate malicious code in other languages, such as Java. Attackers are testing the ability of a neural network to create fraudulent chat bots designed to deceive Internet users. Other scenarios for malicious use of ChatGPT can be the formation of phishing sites and deeply personalized web pages, writing emails of certain content, etc.

Cybercriminals even demonstrated a code fragment generated using a third-party API that allows you to receive current prices for Monero, Bitcoin and Ethereum cryptocurrencies as part of a payment system for the darknet market. In addition, the neural network was used to create a marketplace scenario designed to buy and sell stolen accounts, credit card information, malicious code, and more.^[14]

New York City school ban

On January 6, 2023, it became known that the New York Department of Education banned the use of the ChatGPT AI chatbot developed by OpenAI in schools.

In general, the neural network received positive reviews, but some experts doubted the actual accuracy of the answers issued, and also began to talk about the problems that the chatbot can create in some areas, in particular, in education. Educators are worried that students will start using this AI en masse to do homework, solve math problems and write essays.

Although the tool can provide quick and simple answers to questions, it does not develop critical thinking and problem solving skills that are necessary for academic success and lifelong success, said Jenna Lyle, a spokeswoman for the New York City Department of Education.

In response to the ban, an OpenAI spokesman said the company "does not want ChatGPT to be used for misleading purposes in schools or elsewhere." On the other hand, a number of education workers say institutions will need to find ways to integrate neural networks such as ChatGPT into their curricula rather than banning them entirely. Even if ChatGPT is banned, students will be able to use the chatbot on their personal computers. In addition, as noted, even in schools, ChatGPT can be applied when it is provided for by the educational program - for example, when studying modern AI technologies.^[15]

2022: ChatGPT launch

ChatGPT is an advanced neural network with wide capabilities. It was launched on November 30, 2022 and immediately attracted public attention.

Fundamental advantages:

• Attention mechanism for weighing relevant segments in unstructured arrays;

• Parallel processing - simultaneous processing of all tokens (the basic basic unit of input information);

• Lack of recurrence, which avoids problems associated with long-term dependencies in the data, and therefore eliminates the accumulation of errors in the system;

• Scalability-Quickly and efficiently learn from a large data array.

ChatGPT is based on the use of transformers in NLP, Big Data processing methods, deep learning with the active use of reverse error propagation and GAN methods, and most importantly, contextual understanding, i.e. the ability to take into account the wide context of input data based on the history of requests.

Thus, the success of ChatGPT is the combination of 8 AI features and 8 AI directions at the same time:

• Generative AI
• NLP AI
• Cognitive AI
• Machine learning
• Deep learning
• Predictive AI
• Decision AI
• Robotics AI (if necessary).

2017: Google leads AI development

Google was the leader in AI development and in 2017 introduced the then revolutionary BERT, but all laurels were later taken over by OpenAI, which managed to bring Google's technology solutions to mind by introducing an affordable and extremely effective model.

Notes