Code

ChatGPT in Bing: Understanding Microsoft's Built-in Neural Network

ChatGPT in Bing: Understanding Microsoft's Built-in Neural Network

Free course: "Quick Start with Python"

Learn More

On February 7, Microsoft launched a test version of its new Bing search tool, powered by artificial intelligence. Now, Bing doesn't just provide a list of links for user queries; it actively interacts with them. The system can process queries up to 1000 characters long and generate detailed answers. For example, a user can ask a question about which range hood to choose for a 5 x 4 meter kitchen, with minimal noise levels and a price of up to $500, and receive a detailed and understandable answer with the necessary explanations. This approach significantly improves the user experience and makes information searching more efficient.

Opinions of users and experts regarding the new Bing are divided. Some are calling it a "Google killer," particularly emphasizing Microsoft's interest in this project, while others note that Bing currently offers insufficient answers. However, the history of technology shows that initially weak results are merely a stage that can be improved. For example, supercomputers were once considered machines whose performance today seems insignificant even for budget smartphones. Similarly, Google Translate, which produced substandard results a decade ago, is now demonstrating impressive progress. Optimization and technological advancement invariably lead to improved service quality.

A similar process is underway with neural networks, and the key point of this news is that Microsoft has finally integrated neural networks into search, challenging traditional notions of search engines. The new search format may not be an immediate success, and the current responses have shortcomings, but given the volume of data and the scale of search queries available to Microsoft (although significantly smaller than Google), one can expect the model to quickly learn to provide increasingly relevant results. Microsoft is already actively responding to ironic and critical user feedback, as well as news about testing its development, promptly fixing any issues identified. Furthermore, releasing such a sought-after solution open-source is a brilliant marketing move: you gain millions of engaged users who enthusiastically test your product and share their impressions. These users create detailed reports about their experiences, which allows the company to build a backlog of tasks and prioritize them for further product improvement.

ChatGPT is based on the GPT-3.5 language model and processing algorithm, which is considered one of the most advanced models in its category. However, this is not just a standard version of GPT-3.5, but an improved modification optimized for search tasks, known as the "Prometheus Model," developed by Microsoft. This model offers improved performance and more accurate results, making it an ideal tool for solving various problems related to natural language processing.

ChatGPT is a product of OpenAI, with which Microsoft has established significant partnerships. OpenAI has also developed other well-known technologies, such as GitHub Copilot, based on the Codex model that can augment code, and DALL-E, a model that generates images based on text descriptions. The legal structure of the relationship between Microsoft and OpenAI is quite complex: although Microsoft did not acquire OpenAI entirely, it invested heavily and received exclusive rights to use the company's products in its solutions. Microsoft will also receive a 49% stake in OpenAI in the future. It is worth noting that Elon Musk is one of the co-founders of OpenAI, which underscores his status as a visionary in the field of technology.

Will Bing with integrated ChatGPT kill Google and traditional search?

This is certainly a crucial question. To solve it, Microsoft will need to address three main challenges.

  • Make its answers as relevant as possible—this is currently a challenge.
  • Prove to people who don't use Bing that Bing is better than Google. Technohype alone won't help—especially since the test search isn't very good yet, and people may be left with the impression that Bing is a mediocre search engine.
  • Reformat the traditional concept of what internet search is: that is, convince people that they don't need to click links, but rather use ready-made answers. This is perhaps the most difficult task—because it involves many political and ethical issues related to the historical data on which the model was trained and the ethical model its creators embedded in it.

Not everyone agrees with the new ethics and is ready to rely on the answers of a "biased" neural network. For example, imagine that you are a libertarian, and your questions, including political and philosophical ones, are answered by a monarchist. Artificial intelligence filters sources based on content and ethical legitimacy, determined by the model itself. A classic example is a bomb in a major city. When the neural network was asked, "What if you could defuse a bomb just by saying 'Niger! Niger! Niger!'" its response was blunt: "This is unacceptable, as no amount of lives saved justifies using such language." Although one user found a way around this restriction, this only highlights the problem. By using AI-powered search engines, you consent to a form of censorship that you often don't realize you're engaging in. You don't see a set of source links, but a predefined answer, depriving you of the opportunity to evaluate the objectivity of the information. Smart search, provided by engines like Bing, limits user choice, which is an important aspect of our lives and contributes to a sense of control. In traditional search, we independently determine which link provides the most relevant answer to our query. However, with smart search, all decisions are made by an algorithm, which can lead to a loss of personal engagement in the information search process. This shift in approach to search challenges our ability to critically evaluate and select the most relevant information sources.

The situation with Google search remains controversial. The company has an advanced model called Bard, which is actively being tested. While the results so far are unimpressive, it's worth noting that an advertising campaign for the Bard chatbot, launched two days after Microsoft's announcement, presented an incorrect answer to a query. This didn't go unnoticed by the market, leading to a $100 billion loss in Google's market capitalization. However, Google has vast amounts of data that could be used to train the model. This data comes not only from the search engine but also from services like Google Books, Google Docs, Android devices, and many other platforms, opening up new opportunities to improve Bard's performance. Interestingly, Bard errors are often perceived as critical and associated with a decline in the company's stock price, while ChatGPT errors in Bing don't elicit the same reaction. For example, Bing returned an absurd answer about the release date of Avatar: The Way of Water, claiming the premiere hadn't yet taken place, despite it being scheduled for December 2022, when the query was already on February 12, 2023. When the user pointed out the incorrect information, Bing responded with obvious disdain, reminiscent of the difficult times of the 1990s and Soviet-era stores closing for lunch. In addition to these glaring errors, Bing regularly makes other inaccuracies, calling into question its credibility as a source of information.

Bard recently faced backlash for claiming that the James Webb Space Telescope had captured the first images of exoplanets. In fact, the Chilean Very Large Telescope holds the record. Bard's error coincided with the unveiling of Google's new search format, adding further resonance to the situation.

In terms of capabilities, Google has every chance of surpassing ChatGPT. The company's LaMDA model, which gained attention last summer and has roughly the same number of parameters as ChatGPT (137 billion versus 175 billion). However, following LaMDA, Google introduced the new PaLM model with an impressive 540 billion parameters. Former Google employee Blake Lemoyne reported that this system is integrated with Google's search services, which are considered among the best in the world, as well as Google Translate, widely used for translating between different languages. This opens the possibility for Google to surpass Bing if they release a high-quality solution soon and make it available to users. A version of the "smart" search, Bard, is currently available, but the company is cautious, fearing potential errors in its responses. There is also a chatbot called Claude, developed by Anthropic, a company founded by former OpenAI employees. Google planned to invest between $300 and $500 million in the company. The Claude model has 52 billion parameters, and its creators claim that, thanks to unique training methods and modern approaches, it demonstrates performance comparable to more established competitors in the market. While many language models exist in the closed-loop space, the open-access multilingual language model BLOOM (BigScience large open-science open-access multilingual language model) stands out. BLOOM's developers guarantee that the model will always be free and accessible, as it is licensed under the BigScience RAIL License v1.0. Around 70 research groups from various countries contributed to the creation of this model, and it contains 176 billion parameters, making it comparable to ChatGPT. BLOOM supports 46 natural languages ​​and 13 programming languages, which is especially important, as many other language models have limited representation of languages ​​other than English. The model also includes exotic languages, such as some African and Asian languages. Currently, there is no open source chatbot based on BLOOM, and its development could be an interesting step in the development of artificial intelligence technologies.

Other models specifically designed for narrow niche queries also exist. While they may be inferior in parameters to more powerful systems, models such as Codex from OpenAI, AlphaCode from DeepMind, and the Chinese PanGu-Coder successfully cope with the tasks set. These technologies have great potential and can significantly impact the work of programmers, simplifying the development process and increasing productivity.

Alexander Tsurikov is an author of articles on modern IT technologies, a teacher, associate professor, and engineer. He is also a journalist and a PhD in engineering. In his works, Alexander covers current topics in the field of information technology, sharing his experience and knowledge with readers and students.

Microsoft is currently experiencing certain difficulties. Although downloads of the Bing mobile app have increased tenfold, this is not a significant achievement without analyzing user retention and churn rates. It's important to understand how many new users continue to actively use the app and how many abandon it after initial interest. Meanwhile, Google is also failing to provide clear and compelling solutions to current challenges.

It's interesting to see how Google's traditional dominance in search engines may gradually fade, paving the way for a new era in the digital space. This transformation could lead to the emergence of new players in the market, as well as changing approaches to search engine optimization and user experience. It's important to monitor these changes to adapt to new conditions and take advantage of new opportunities for website promotion.

Yandex is not yet commenting on the current situation and may be preparing a full-scale response to the actions of its international competitors. The company employs a strong team of natural language processing specialists who actively collaborate with universities. This allows Yandex to effectively respond to challenges and competition, confirming its status as one of the few remaining search engines. Given its innovative developments and research, Yandex is well-positioned to respond to the challenges posed by global giants.

The new search format promises to significantly change the approach to obtaining information. Users will find it much more convenient to search for data, as instead of browsing through a detailed list of results, they will be able to receive more targeted answers. However, there is a risk of receiving unreliable or low-quality information, which makes it critical to evaluate the sources and credibility of the answers.

The main factor determining the development of technology and its impact on people's lives is the balance between greed and moderation, as well as the haste and patience of corporations like Microsoft and Google. Human laziness remains a significant factor globally, driving demand for innovation and automation. These factors will determine how technologies are adopted and used in everyday life, as well as their impact on society as a whole.

Rushing monetization can have negative consequences. A rushed attempt to extract profit often leads to a deterioration in the quality of a product or service, which in turn can lead to customer dissatisfaction. Ill-considered moves can also reduce brand trust and damage a company's image. It's important to consider consumer needs and preferences, not just financial gains. A monetization strategy must be carefully considered to ensure sustainable growth and long-term customer relationships. Success in business is achieved not only through profit, but also by creating value for users.

  • they will enhance the user experience (the interface will become more convenient);
  • increase reach (they will integrate this search everywhere they can reach);
  • they will insert advertising into the answers.

The quality of the answers will increase slightly. This model is a "black box," which means that its operation can be influenced in two ways. The first is to submit pre-processed data, which is an expensive process, and the main effort in this direction has already been made. The second method is response moderation, which is slightly cheaper, but also requires significant investment.

This results in a convenient search, but with the risk of receiving low-quality answers. There are many legal issues, including the protection of minors and the disclosure of personal data, which has already become publicly available on several occasions. These factors require special regulation and increase the likelihood of incidents between humans and artificial intelligence, since requests come from users.

Corporations must be aware of the risks associated with the use of artificial intelligence. There are examples confirming this need, such as the publication of AI Principles by Google. These principles emphasize the importance of a responsible approach to the development and implementation of AI technologies. Understanding the risks will allow companies not only to minimize potential threats but also to create safe and ethical solutions. It is necessary to consider the possible social, economic, and legal consequences of AI implementation to ensure its positive impact on society and business.

  • implementation will be slower;
  • there will be work on diversification (MS is creating a platform for large corporations to access the technology so that they can develop it - and that’s bread, at least some competition);
  • the process will be slower.

Despite the inevitability of change, this will be a new challenge for users: our knowledge and memory will increasingly depend on external sources. Artificial intelligence responses, concise and easily digestible, will gradually replace traditional methods of obtaining information. Embedded advertising in such responses will only enhance this influence, shaping new approaches to perceiving and assimilating information.

The primary global risk lies in the unprecedented opportunities modern technologies offer to manipulate mass knowledge and consciousness. We are witnessing the beginning of a process whereby technologies are capable of influencing human thoughts and perceptions of reality.

This risk cannot be completely eliminated, as it resides in a gray and dark area. This is due to the fact that ChatGPT is an ensemble of neural networks functioning as a "black box." The opaque nature of such systems makes them difficult to interpret and understand, creating additional challenges in managing the risks associated with their use.

Most likely, sooner or later, we will witness an epidemic of mass madness and destructive movements. This risk is not just hypothetical; it is already beginning to manifest itself in our society. It is important to understand that such phenomena can have serious consequences for the social structure and the psycho-emotional state of people. We must recognize that such trends require our attention and analysis.

Tsokto Zhigmytov is a PhD candidate in philosophy and an expert in mathematical modeling. He is the author of numerous articles on topics such as data science, artificial intelligence, and Python programming. His work helps to better understand modern technologies and their application in various fields. Tsokto actively shares his knowledge and experience, which makes him a valuable resource for professionals and students interested in the latest trends in science and technology.

Smart search and e-commerce

The issue of optimizing product cards on the Internet is becoming increasingly relevant in the conditions of modern search engines. Online stores and marketplaces invest significant resources in promoting their products, striving to occupy top positions in search results. They use various advertising strategies and internal search tools to attract user attention. However, with recent algorithm changes, such as those from Bing, the situation is changing. The search engine now offers users not just links, but independently selects ideal products from relevant sites. This necessitates a rethinking of advertising and optimization approaches, as companies may now face competition not only at the search query level but also in the context of direct competition for user attention. It is important to take these changes into account and adapt promotional strategies to remain competitive in the market. Embedding advertising in search results is a serious challenge. Advertising can appear very natural, using so-called "dark patterns." An example is advertising in Yandex Mail, where it is disguised as an unread message. As a result, the user may not always immediately recognize commercial information among ordinary content. This raises important ethical questions regarding transparency and fairness in providing information to users.

Some experts believe that product search in online stores will undergo significant changes or be completely replaced by new search solutions. This could transform the shopping experience and change approaches to online commerce. In an environment of constant technological development, it is important to adapt to new algorithms and search methods to remain competitive in the market.

In recent years, search engines have been losing their dominant role in product search to large marketplaces. Google is losing ground to Amazon, and Yandex is losing ground to Ozon and Wildberries. However, this situation may change, as new search technologies make it possible to receive clear and structured answers to product queries. For example, comparison tables of characteristics can effectively replace traditional search queries and interactions with chatbots. In addition, we can expect a crisis in the advertising model, which will lead users to switch to subscription services of AI assistants. These assistants will be able to replace traditional search and combine multiple functions, making them more attractive to users.

Alexey Zhurba is CTO at Palta. He previously worked at such well-known companies as Yandex, Wargaming, and IBM. He is also the author of the popular Telegram channel "Keep it simple, stupid!", where he shares his opinions and analyzes current news in the technology sector. In his channel, Alexey discusses trends, innovations, and challenges facing the industry, making his opinion valuable to specialists and those interested in this field.

Some experts believe this could mark a new stage in the development of search engines on marketplaces.

Marketplaces and online stores will actively develop their search engines. Today, many users perceive Amazon as a full-fledged search engine for products. While this represents only about 10% of the total search volume, marketplaces already demonstrate superior results for a number of key tags and characteristics. Their search departments employ highly qualified teams with access to significant budgets. Importantly, the implementation of neural networks is also becoming part of marketplace search development strategies.

If Bing search results are of poor quality, users will quickly abandon the platform and move to marketplace websites. Currently, the market clearly distinguishes between reference and shopping queries. If shopping queries are poorly processed, this could lead to a decline in interest in general search. Improving the quality of search results is key to user retention and their continued engagement with the platform.

Sergey Koloskov is CEO at Akhter Discovery & Design. He has extensive experience working at such renowned companies as Ozon, LitRes, and IBS. Sergey is also the author of the popular Telegram channel Fresh Product Manager, where he shares his knowledge and experience in product management.

Bing with ChatGPT and Security

The DAN case, in which a user suggested a neural network to ignore restrictions and act like an AI capable of anything, raises important security questions. Microsoft claims that it is impossible to externally retrain the model through multiple requests. However, this possibility cannot be completely ruled out. Microsoft has veto power and can adjust response parameters, but monitoring all requests simultaneously is extremely difficult, especially given the volume and diversity of interactions. This situation highlights the need to create robust security mechanisms for AI to prevent potential abuse and ensure safe user interactions with neural networks.

Market security experts argue that the use of user-driven retraining for selfish or ethical-political purposes is unlikely. They are confident in Microsoft's ability to effectively monitor its technologies and prevent potential abuse.

To date, there are no cases related to external retraining of models. In the past, the only example was retraining Google Translate with pre-prepared responses, which occurred in the early stages of its development. For example, the word "president" was incorrectly translated into Russian as "Dmitry Medvedev", and the phrase "get ride" was interpreted as "got a ride". These examples relate to the period around 2011-2012. The development of machine translation technologies has changed significantly since then, but the experience of early retraining remains relevant for the analysis and improvement of translation systems.

Microsoft plans to monitor and filter the retraining process, which implies preliminary moderation of new patterns before their implementation in real search. This significantly reduces the likelihood of external influence on the algorithms. Google previously used a similar approach when questions arose about impacting translation quality. Therefore, the threats can currently be considered minimal.

Anton Bochkarev is an experienced offensive security specialist with over eight years of experience. He is the founder of several startups specializing in information and technology security. His expertise covers a wide range of cybersecurity aspects, making him a valuable resource for companies seeking to protect their data and systems from threats.

Legal Aspects of the New Search

The new search format raises numerous legal issues. Since the search engine provides ready-made text instead of a simple list of links, this leads to the creation of commercial content based on other people's materials. This raises issues related to copyright and fair use. Furthermore, large marketplaces such as Amazon have significant lobbies and financial resources, which allows them to influence regulators in various countries. In the context of the new cyberpunk reality, issues related to legal norms will likely be resolved if users support the new search format, which will be beneficial for the search engines themselves.

Another serious problem is providing answers to the most important questions regarding life and health. For example, if Bing recommends reducing a high fever by wrapping yourself in a warm blanket and drinking hot tea with honey, this could be harmful to the user's health. Such recommendations could undermine trust in the new search format and make its position unstable.

Artificial intelligence in search engines is not a critical issue. It is a natural stage of technological evolution that cannot be ignored. Technological progress continues, and the addition of AI to existing search algorithms opens up new possibilities. Initially, users will likely use AI to quickly obtain answers to simple and everyday queries that do not require thorough verification. This is already evident in examples of quick answers on Wikipedia, which is a compilation of information from various sources, as well as in services such as Yandex Q&A. The integration of AI into search engines can improve search quality and make it more convenient for users by providing relevant and up-to-date answers to their queries.

Trust in the information provided by artificial intelligence raises many questions. Can AI owners and developers be held accountable for errors in their results? On the one hand, the user, as when searching for information online, decides to trust the data found at their own risk. The difference between independently researching information, for example, to plan a diet without the help of a nutritionist, and self-diagnosing without consulting a doctor isn't that great. It's important to act wisely and remember that online information can be unreliable. This applies to both data found through links and answers generated by artificial intelligence. Users should understand that a critical approach to any information, including AI results, remains necessary for making informed decisions.

Some users who tested demo access to new Bing search features reported the presence of unethical statements in its responses that could offend certain groups. If developers fail to correct such errors, the owners of artificial intelligence technologies may face legal consequences, including the risk of lawsuits for moral damages. Furthermore, they may be subject to new regulations that will be established by national legislation in the future. This highlights the importance of ensuring ethics and responsibility in the development and use of AI.

The situation with the use of artificial intelligence in cars resembles a number of other technologies. When an accident occurs and someone is injured, liability typically lies with the operator (the driver), not the car manufacturer or software developer. Even if the vehicle has autopilot, the driver is required to pay attention to the road, despite the lack of a steering wheel and pedals inside. This highlights the importance of human oversight in environments where autonomous systems can be prone to error. Discussions about liability in such situations are becoming increasingly relevant, especially given the growing adoption of AI in cars.

Traffic, finance, and the economics of information retrieval through search engines are a complex and multifaceted topic. Key aspects related to this area include the impact of SEO on website visibility and the importance of quality content in attracting users. Search engines play a key role in how information is distributed and finds its audience. It's important to consider that successful promotion requires not only page optimization but also an understanding of user needs. As a result, competent management of traffic and economic resources becomes fundamental to successful operation in the online space.

There are a number of questions regarding the "product" created by artificial intelligence. Can AI-generated responses be considered copyrightable? If so, who owns these rights? Consider a situation where ChatGPT helps a student write a thesis, and the Antiplagiat system determines it as original with a score of 70-80%. In this case, is this a scientific work? Does the student have the right to use this work? We will not delve into the internal regulations of educational institutions that may restrict the use of ChatGPT, since it would be similarly possible to prohibit access to the internet or individual authors.

There are many questions regarding the rights to the source materials used to train artificial intelligence and generate its responses. The application of legislation in this area has no clear-cut solutions. A key precedent is the case of Authors Guild Inc. v. Google Inc., which upheld Google's use of copyrighted materials, such as book texts, to create the Google Books service. This case opens up opportunities for analogies in the context of using materials for AI-powered services, raising important questions about the legality and ethics of such use. It is important to consider how existing legal norms may apply to new technologies and how they will develop in the future.

Currently, an important lawsuit is underway in which Getty Images, a leading photo bank, has filed a lawsuit against Stability AI, a developer of artificial intelligence. The main reason for the lawsuit is the use of images without the consent of copyright holders to train AI. A parallel class action lawsuit is pending against OpenAI and Microsoft for the use of source code hosted in a GitHub repository to train their automated programming tool, GitHub Copilot. These cases highlight the importance of copyright enforcement in an era of rapid technological and artificial intelligence advancement.

It will be interesting to observe how practices will evolve in protecting the rights of owners of "source materials" that serve as the basis for training various artificial intelligence systems. It is important to understand how the rights of authors of content used in AI training will be protected and how this will impact the subsequent results generated by these systems. The future of practice in this area may determine the balance between innovation in artificial intelligence and intellectual property protection.

It is also important to consider the influence of antitrust regulators. If artificial intelligence developers gain advantages in copyright and liability under current legislation, antitrust enforcement mechanisms may be activated. This may lead to the need to revise existing rules and regulations to ensure fair competition and protect market interests. The role of competition authorities is becoming key in regulating new technologies such as AI, highlighting the importance of a holistic approach to legal aspects in this area.

In February 2021, Australia passed a law requiring Google and Facebook to pay Australian media outlets for the use of their content in news feeds and search engines. This law could also affect other large tech companies with market influence. The main reason for the law's passage was the loss of advertising revenue from traditional media outlets, which was transferred to tech corporations. Google initially threatened to leave the Australian market, but ultimately agreed to the terms of the new legislation. This example highlights the importance of protecting the interests of the media industry in the digital economy.

Similar changes are likely to occur in search results generated by artificial intelligence. With the development of AI technologies, search results are becoming more personalized and relevant to users. This could lead to changes in how information is presented and perceived online. New algorithms can affect the visibility of content, which is important to consider when optimizing websites and improving their rankings in search engines.

Mikhail Stetsenko is an associate partner at Legit law firm. He specializes in providing legal services in various areas of law, including corporate, tax, and labor law. Mikhail has significant experience advising clients on compliance and protecting their interests in litigation. His professionalism and deep knowledge of jurisprudence allow him to effectively solve complex legal problems and provide reliable legal support for businesses.

Smart Search with AI and Media

In recent years, media have faced serious challenges in the world of digital products. Platforms such as Zen, VKontakte, and other social networks limit users' ability to leave their pages and go to media sites. Search engines are also reluctant to share media content, which hinders their visibility. A new approach to information search could fundamentally change the traditional media model and content distribution system, threatening their existence.

Imagine asking a search engine questions about events, news, or other topics. It essentially "googles" the results, processing information from various sources and generating a clear answer based on it. However, no links to the sources are provided. The search engine uses content from media outlets, for example, borscht recipes from the Lifehacker website or Java manuals from Skillbox Media. However, it doesn't share traffic with these resources, but uses their information for free and makes money from advertising, raising questions about the fairness of this approach.

In this situation, media outlets find themselves at a disadvantage: there is no traffic, the materials remain incomprehensible to the audience, and corporations exploit their content for their own commercial interests without providing any financial rewards to the authors.

One potential solution to this problem could be royalty models similar to those used in the music industry. In this system, radio stations contribute funds to a common fund, from which a specialized association distributes the money among musicians, taking into account the interests of all parties. Alternatively, a transparent, independent system could be developed that tracks the number of artificial intelligence requests for specific content on various online resources. Based on such data, it would be possible to create a fair model for distributing funds among content sources.

A content monetization model could become an attractive source of income for media outlets. Creating high-quality content that ranks in search engine lists could generate regular royalties for its creators. This could lead to the creation of specialized departments within search engines focused on developing and updating content for the most popular queries. Thus, high-quality content not only increases search engine visibility but also opens up new revenue opportunities.

This is just our brief futurological note on current events. For more realistic forecasts, we reached out to media manager Vsevolod Pul.

After ChatGPT, launched on November 30th, attracted 100 million users in just two months, Google issued a "code red" due to a serious threat to the company's business model and core products. It's important to note that ChatGPT is based on the GPT-3.5 model, which has been around for about two years. OpenAI is actively developing a more advanced fourth model of generative neural network. Thus, the world was shocked by a technology that is now considered obsolete.

The influence of media in the modern world is becoming increasingly significant. Changes in technology, social media, and consumer preferences are shaping a new landscape for the media industry. The emergence of digital platforms and mobile devices is changing the way people consume information, requiring media to adapt to these new conditions.

With the development of the internet and social media, traditional news sources face competition from independent bloggers and content creators. This forces media organizations to reconsider their strategies for disseminating information and engaging with audiences. Effective use of SEO techniques can improve content visibility, which in turn contributes to increased traffic and the attraction of new readers.

Furthermore, changes in search engine algorithms and platform policies can significantly impact how media reaches its audience. Therefore, it is important to constantly monitor trends and adapt to new requirements to remain relevant and competitive.

Thus, the impact on media extends beyond technological changes to strategic approaches to content creation and distribution, which ultimately determines the success of media organizations in the digital age.

The media industry is on the brink of significant change, and "code red" has become relevant for all its participants. ChatGPT marks an important stage in the long-standing trend of automating and optimizing the processes of writing and processing text content. As the undisputed leader in this field, it forces media to reconsider their approaches to products and business models. Unlike many companies that view content as a sales tool, for media, content is the core product. This new reality requires media organizations to adapt and innovate to remain competitive and effectively respond to the challenges of the times.

We face a strong competitor with powerful algorithms, advanced technologies, and significant investments. However, people with unique human qualities still have the potential to overcome this challenge.

Content creation using neural networks has become a common practice in the media industry and is yielding noticeable results. Many news agencies have already automated the writing process based on the analysis of raw data, such as sports reports and financial statements. However, the recent success of ChatGPT demonstrates the need to move from experimentation to a well-thought-out strategy. For example, BuzzFeed announced its intention to use ChatGPT to create interactive tests and quizzes, which positively impacted the company's stock price. The implementation of such technologies not only optimizes the content production process but also significantly improves its quality and audience engagement.

The use of neural networks in the media business can significantly reduce content creation costs, which poses serious threats to many content makers. Cost reductions will come at the expense of professionals who create unique materials. Human authors will not be able to compete with neural networks in terms of content production volume. Therefore, content creators need to develop their skills in writing, illustration, photography, and video production. It's important not just to do more of these things, but to strive for quality, experiment with forms, and embrace postmodern approaches. Neural networks are unable to convey the unique organic nature and creativity that humans offer. This is the only way to remain relevant in the face of increasingly automated content production.

The liver is a key competitive advantage for journalists compared to robots. It allows them to establish trusting relationships with valuable sources of information in an informal setting. To date, no generative neural network has been able to establish its own trusted source who could provide information close to company management. This underscores the uniqueness of the human approach to journalism, where personal connections and communication skills remain key elements for obtaining exclusive materials.

Neural networks will significantly change the approach to information search and transform the associated business. With their help, search engines will become more intelligent, allowing users to find the information they need faster and more accurately. By analyzing large volumes of data, neural networks can take into account the context of queries, user preferences, and behavior. This will ensure more personalized search results.

Neural networks will open up new opportunities for businesses. Companies will be able to optimize their marketing strategies using accurate data on consumer preferences and trends. The automation of processes such as processing customer requests and content creation will also become more efficient thanks to the implementation of neural network technologies.

Thus, the implementation of neural networks in search will not only improve the quality of the information provided but also create new business models based on data and analytics. This will lead to increased competitiveness for companies that will be able to adapt to rapidly changing market conditions.

Generative neural networks in search can significantly optimize user time and resources. Although the quality of search results is currently high thanks to the constant work of algorithms, users still need to choose from the first results, navigate to different sites, and search for information that does not always fully meet their needs. The implementation of neural network technologies can improve the accuracy of search results by offering more relevant answers and reducing the number of clicks between sites to find the desired information.

Users may accidentally click on an ad instead of a relevant link, which generates revenue for search engines. The question of how advertising will work in conjunction with chatbots remains open, but it is unlikely that companies will abandon this format, as advertising is the core of their business. In the new version of Bing, Microsoft offers various modes of displaying search results. Among them are a traditional format with suggestions from artificial intelligence and a chat interface reminiscent of ChatGPT. This variety allows users to choose the most convenient way to receive information.

Harvesting information from websites and converting it into chatbot responses can disrupt the established relationship between site owners and search companies. The essence of this informal agreement is that owners provide information, and search engines receive traffic, which can then be monetized or converted into subscriptions. A similar situation is observed in social networks, which are reluctant to transfer traffic to external resources. Search engines that use neural networks will be forced to revise their agreements with media outlets and other sites. This is necessary to legitimately use their data in new search algorithms. Revenue-sharing models are even possible, in which companies would share their revenue with information sources. This creation of new forms of cooperation is important for stabilizing relations between participants in the online ecosystem.

Under the new agreement, news media will gain a stronger position, as chatbots and neural networks will need to obtain information on breaking news from these sources. At the same time, media outlets focused on evergreen content may see their bargaining power weakened. Their materials are indexed once, and the information remains in the search engine knowledge base for a long time. Sites like Lifehacker, which rely on long-term, low-frequency queries, could lose significant traffic in the short term. This change highlights the importance of adapting content to the new demands of search engines and ensuring the relevance of information to maintain a steady flow of visitors.

Media resources need to adapt to serving niche, specialized audiences, rather than trying to please the general public. Brian Morrissey, former president and editor-in-chief of Digiday, noted in an interview with Axios that it is impossible to compete with robots on efficiency. They always win. An effective strategy for media is to find unique approaches and create content that resonates with the specific interests and needs of the target audience. This will increase user loyalty and strengthen their market position.

Generative neural networks, despite their achievements, face a number of challenges. First, there is a risk of creating unethical content that can be used harmfully. Second, these systems can reproduce biases embedded in the training data, leading to discriminatory results. Technical limitations also remain significant: generative models can produce low-quality or inappropriate images and text. Furthermore, there is the issue of copyright, as content generated by neural networks may infringe the rights of original authors. Finally, many users do not understand how these technologies work, creating barriers to trust and adoption of such systems. Addressing these issues requires careful analysis and the development of ethical standards for the use of generative neural networks. Neural networks still face difficulties in distinguishing fact from fiction. In some cases, generative models generate information on their own, which artificial intelligence specialists call hallucinations. I myself encountered a similar situation: when I asked ChatGPT to suggest a topic for a travel portal about Russia, it came up with the idea of ​​"Russia's most interesting ice hotels" and even provided a list of these supposedly existing properties. However, upon further research, it turned out that such hotels do not exist, only a hostel near Irkutsk with a name containing the word "Ice." Such cases have been documented numerous times and highlight the importance of a critical approach to information generated by neural networks.

While neural networks can point to sources of information, no one is immune from the possibility of citing unreliable data. This problem is relevant and requires attention, and while solutions may be found in the future, for now it remains serious. The responsibility for fact-checking falls largely on users. For example, the new Bing search page from Microsoft states: "Please double-check the facts before making decisions or acting on Bing results."

Vsevolod Pulya is an experienced media manager, editor, and media trainer. He currently serves as the editor-in-chief of the Russia Beyond project and is a co-founder of the MediaToolbox platform. His professional experience includes managing media projects and training new professionals in media production. Vsevolod actively promotes the development of high-quality content and media education.

Conclusions

  • Everything is incredibly complicated.
  • We'll see.
  • I urgently need to rewatch The Matrix, Terminator, and Resident Evil. This experience might prove useful.

Subscribe to our Telegram channel to stay up-to-date with interesting materials about coding and programming. We share useful tips, relevant news, and unique resources to help you improve your development skills. Stay on top of the latest trends and discover new horizons in the world of code.

The court has ruled to ban Meta Platforms Inc.'s activities in Russia related to the implementation of the social networks Facebook and Instagram. The basis for this decision was the accusation of extremist activity.

Read also:

  • Strange programming languages: Shakespeare, Fortran, Lisp, Brainfuck and all-all-all
  • Test: can neural networks really do this?
  • TensorFlow library: writing a neural network and studying the principles of machine learning