The main articles are:

• Artificial intelligence
• Artificial Intelligence (Russian market)
• Artificial Intelligence (Global Market)

Generative models

The first highly developed generative AI models began to appear in 2017: ProGAN, CycleGAN, RealNVP, VQ-VAE, Glow, WaveGAN and WaveNet, StyleGAN and StyleGAN2 and BERT appeared a year later. GPT integration began in 2019, and it was she who achieved success.

The most popular types of generative models for 2024 are:

• Generative Adversarial Networks (GANs),
• variational auto-encoders (VAEs) and
• Transformer architecture used in GPT and BERT.

Tokens - Basic Units of Information and Memory Constraints

A token is a basic unit of information used by the model to process and generate text. A token can be a word, part of a word, a symbol, or even a group of words, depending on how the model was trained to separate and interpret the text.

Without going into tokenization algorithms, on average, the English text contains 4-5 characters in one token, and the Russian text 2.2-2.4 characters. For example, an article in Russian in 6000 characters, taking into account spaces, will contain about 2600 tokens. It is cheaper to generate texts in English, Spydell Finance wrote.

The most significant problem of generative AI is the length of the context and the built-in algorithms of functioning.

The first editions of GPT 3.5 had 4096 tokens, the latest version has 16 thousand tokens, GPT 4 had 32 thousand until November 2023, and after November GPT 4 Turbo has already 128 thousand.

Accordingly, the maximum context length for GPT 3.5 Turbo with 16 thousand tokens is about 37 thousand characters of text in Russian, and for GPT 4 Turbo with 128 thousand tokens - almost 300 thousand characters.

What does that all mean? Within a single session, generative artificial intelligence (GII) forgets what was discussed at the beginning of the discussion, which makes it impossible to accumulate experience and detailed discussions. It is as if a person forgot every time that it was two days ago, cutting off all the past accumulated life experience.

In short, the principle of operation of the GII is to formalize the context vector (compressed representation of input data) to generate a continuation of the dialogue (output information), i.e., content compression (input information unit) occurs. What can this be compared to?

Imagine if a high-quality PNG image of 3840 * 2160 pixels were compressed to a picture of 384 * 216 JPG pixels, i.e. 100 times over pixels with an aggressive lossy compression mechanism. It seems that you can correctly interpret the information in a compressed picture (understand what is depicted), but there is an irreversible loss of details.

So with the GII. All information coming to the input of the GII in one way or another with various algorithms and compression mechanisms undergoes compression.

For example, the task is to analyze 10 million characters of legal information, which is about 4 million tokens. How will the context window of 128 thousand tokens process information that is multiple of the limit length of the context?

As of early 2024, there are two of the most commonly used admissions.

• Separation of text into blocks, paragraphs close in meaning and forced compression by X value.
• The sliding context window, when the first block is sequentially processed by 128 thousand tokens, is compressed 20 times, then the second block, and so on. As a result, the output is an extract 20 times with the inevitable loss of parts and content. Is it possible to apply this technique in legislative documents? Not yet.

Thus, GII is very useful for compiling compressed summaries of text, video and audio information. To analyze the information representing the value - not yet.

This means that if you write large-scale works of art or research projects, GII will not allow you to effectively inherit the characteristics and connections of objects in earlier iterations.

Is it possible to solve this problem in the future? The context length must grow by several parameters. Not 128k, as at the beginning of 2024, but thousands of times more!

GII problems

At the beginning of 2024, Spydell Finance highlighted the following limitations of generative AI:

• There is no self-study.

• Not suitable for complex integrated and multidimensional projects, where you need to take into account the holistic picture and dynamic connections

• It is necessary to very rigidly formalize the task to obtain the desired effect and often the time for industrial engineering exceeds the benefit of using GII (it is easier to do everything with your hands, as before). It is necessary to divide the task into many subtasks until the maximum simplification.

• Limited context length (see tokens above).

• A lot of mistakes, no fact checking. The GII masterfully imitates the truth and generates output content very convincingly, but if you dig deeper, it turns out that a significant part of the information is fake and hallucinations.

As of May 2024, there is no positive experience in integrating generative AI into research projects in the field of economics and finance. No model is functional, nothing works.

There are critical and so far insoluble problems at the architectural level of the GII models themselves.

Poor output quality: Time and resources to validate GII performance surpass any potential benefit

One of the problems is that there is no embedded control over the verification of the output data and the correctness of interpretation. In other words, the GII is not able to assess the correctness and adequacy of the generated content, there is no built-in truth criterion.

Modern GII models do not have critical thinking and verification of results, which in the context of the work of large language models (BYM) means: identifying logical connections and contradictions, evaluating arguments and evidence, analyzing data and sources, adapting the output into the context of conditions.

Available for June 2024 BYs:

• Do not check the validity of information sources or distinguish between reliable and unreliable data.
• They are not able to independently identify logical errors or contradictions in their answers.
• Cannot critically evaluate the arguments and evidence presented.
• Cannot adequately adapt their responses to specific conditions or task context.

BYUs are trained on extra large data arrays, the original reliability of which is in doubt, and in this set of information garbage data compression and weight determination take place.

Those data on which large language models were trained can initially contain errors, bias and inaccurate information, and therefore training is often based on false information.

In some ways, weights in large language models (BYMs) define the hierarchy of information interpretation, allowing the model to recognize hierarchical and contextual dependencies in the data. In other words, weights determine a measure of the connectivity of information blocks as one piece of information affects the other piece of information.

What does this mean in practice? BYMs are extremely ineffective in developing innovative semantic constructs and interpreting initially contradictory information, producing complex multi-level estimates of factors, circumstances and dependencies.

Generative artificial intelligence can be effective in interpreting universally recognized most popular facts of a regular nature, but not ways to build a hierarchy of priorities and a multilevel composition of risk factors in an ambiguous and unstructured dataset whose distribution vector is not predictive.

Therefore, complex analysis of processes and events is not subject to GII, so there is no intelligence in the GII in the broad sense. This is a highly erudite system that is quite stupid in understanding the connections and dependencies of complex systems, and sociology, psychology, political science, economics are precisely those areas where there is no rigid structuring of data and there is no unambiguity in interpretation.

It is possible to formalize mathematics or physics (here GII in the future 2027-2030 can achieve success), but it is impossible to formalize the motives and actions of society, therefore GII cannot manage business processes, cannot predict and evaluate all those areas where a person is involved (finance, economics, sociology, politics, etc.).

What does this lead to? GIIs generate a huge amount of content that is almost impossible to apply to applications due to lack of reliability.

Ideally, the system should work like a low-level program in a processor, where repeating experiments always gives the same result - there is unambiguity and predictability. The GII has too wide a range of tolerances.

As a result, the time and resources to check the results of the GII exceed any potential benefit.

The low reliability of the output content is sewn at the level of the BYU architecture, so the problem is incorrigible neither now nor in the near future, Spydell Finance wrote.

Key use cases

Initially, it was assumed that AI would replace a monotonous/repetitive, which means a formalized type of labor and a low-skilled labor force, and only at the very end, at the top of its evolution, it would hit creative professions, but it turned out differently, Spydell Finance wrote.

Generative AI threatens the creative segment of professions:

• Text creation - meaningful texts, practically indistinguishable from the human style of writing, with the ability to create jokes, poems, scripts for films, stories.

• Writing program code and formulas - if you break the program into mini modules, there will be a certain benefit, although there will be many errors initially. However, the main benefit is rather in the tool of dynamic prompts and the search for errors in compilation - suitable for experienced programmers.

• Making music. The generation of musical compositions that imitate certain styles based on the analysis of tens of thousands of works or the creation of completely new musical works (experimental neuro music).

• Create realistic images. Again, based on the analysis of patterns over a large array of real photos, the characteristics, structure and features of objects are ordered and synthetic images are generated through GAN algorithms.

• Making a video, deepfake. AI analyzes video, studying the characteristics of movement, transitions between personnel, visual and audio patterns and further using a similar method, as with images - video AI is generated. It was expected that the GII could be used for animation, creating special effects. But in 2024, this does not work for a banal reason - the GII does not take into account the laws of physics and is not able to adequately assess the mechanisms of interaction between objects and light. Problems with the integration of objects into scenes, incorrect proportions and scales, distortions and artifacts, lighting errors, unrealistic movements, inconsistency and incorrect interaction, not to mention the physics and properties of objects.

• Generation of speech, preserving timbre, intonations, rhythm of human speech. It can be used in AI consultants, simultaneous translators, navigation systems, voice assistants, in voicing books, online videos and films.

• Simulating and modeling complex systems. All that can be formalized is an ideal environment for AI. Scientific calculations are included in this segment. It can be used in expanding predictive/predictive models in finance, economics and business, in virtual prototyping (for engineers, architects, designers).

The main and most basic purpose of the GII is to compress and decompress information, and from this various derived directions are already "split."

Information compression - conspecting, summarizing, generalizing, "summarization" and so on huge arrays of text, audio or video information according to special algorithms. For example, a short retelling of a video, book, instructions for use, or any events.

Decompression of information - from previously compressed information based on logical chains and generate audio, video or text content using given scenario vectors. For example, based on a brief review of a book generated earlier by AI, write similar reviews/reviews or give recommendations on similar literary works.

The main scenarios for using GII:

• Generalization and interpretation of content - the main purpose of GII is useful for filtering and grouping garbage content by directions and topics, but it is not suitable for compression of important documents, because it almost always misses important nuances.

• News digest - in practice, it works poorly, because all leading news agencies have banned the indexation of news content by the GII.

• Translator.

• Expert system/answers to questions (what it means, how to do it, how it works, how to fix it, etc.).

• Analysis, data analytics (for 2024, the weakest link so far and the worst developed).

• Rewrite ready-made texts according to specified directions, styles and keys.

• Writing resumes, reviews, essays, simple articles on given topics.

• More efficient recognition of digital content (OCR documents, video, audio). For example, an automatic transcript of a presentation with audio into text, automatic timecodes in video, recognition and structuring of documents.

• Smarter semantic search. For example, searching for certain objects, plots, and so on in a photo or video. All this is not working yet, but Google promises integration by the end of 2024.

• Create a photo and video with the specified conditions. You can modify old, destroyed black and white photos in high quality in color.

In the future of the year, the GII model allows you to make a smart organizer by structuring documents, letters, photos and videos with given markers. For example, 30 thousand photos in the library should be divided into types and plots (people, nature, cities, cultural objects, etc.). Similar to documents, i.e. smart grouping and search by criteria. Structuring and systematizing content is what the GII is capable of doing.

Potentially useful, as a personal tutor, creator of a guide, guide in various unknown issues and directions, including as a tourist guide, is not a bad space for the role of a consultant on many issues. Some improved Google + Wiki/online library combination.

GII can be used to create a range of ideas for content, used to automate the creation of reports and template projects, and primary data analytics.

At the everyday level, the GII is very good - you can ask to compile and analyze a list of films, disassemble famous literary works, make some recipes or ask how to fix something, and many other variations.

At the level of business decisions and in science, public versions of GII are still very weak.

In practice, the seemingly limitless space of possibilities for 2024 rests only on a translator, a text proofreader, a help desk and an assistant in writing code.

Does it drag on the revolution? Hardly. GII is still some smarter combination of the search engine + Wiki with the understanding that the number of errors may be unacceptably high.

Can public versions accelerate the creation of technologies and improve the quality of existing technologies? Excluded.

However, the GII technology itself is able to influence technological progress, provided access to the LLM architecture and scales, allowing deep modernization of models for specific scientific tasks. This privilege is available only to select companies that develop GIIs.

Chronicle

2024

MTS began to use generative AI to develop sites and mobile applications

On September 4, 2024, MTS announced the introduction of the large language model (LLM) MTS AI in Front Platform - the company's internal platform, which will now allow developing sites and mobile applications using generative artificial intelligence. This will significantly speed up the creation of new products, since the transition from traditional phased development to configuring the parameters of ready-made templates will reduce the time to launch new solutions. Read more here

The capitalization of GII developers in the USA and the EU is more than $10 trillion with revenue of $15 billion and losses - $10 billion. Lists of the largest projects

The GII segment is estimated at $15 billion in revenue in 2024, excluding equipment for all IT solutions in the United States and Europe, excluding China. At the same time, the net losses of the industry amount to about $10 billion - ultra-expensive programmers, infrastructure and marketing. At the same time, the capitalization of companies exceeds $10 trillion.

Among the videos, the largest projects in Western countries for July 2024:

• Luma Dream Machine
• Runway Gen-2
• Genmo
• InVideo
• Pika Labs
• Imagen Video Google
• Noisee AI
• Pictory AI
• Haiper
• Clipwise
• Synthesia

Among the generation of images:

• Midjourney
• DALL-E 3
• Imagen Google
• Stable Diffusion
• Leonardo AI
• Playground AI
• Recraft AI
• RunwayML
• Artbreeder
• Picsart AI

Named 9 main risks of generative AI

(GIi) Generative artificial intelligence marks a significant leap in the ability to neuronets understand, interact with, and create new content from complex data structures. This technology opens up a wide range of opportunities in a wide variety of industries. At the same time, new risks are being created, as stated in the materials IDC published on July 10, 2024. More. here

Russia entered the top favorable countries for the development of generative artificial intelligence

On June 25, 2024, it became known that Russia became one of the countries with the most favorable conditions for the development of generative artificial intelligence (GII), taking a position on a par with such technologically developed states as, Israel, and Singapore. USA UAE

According to Izvestia, this conclusion was made on the basis of a joint study conducted by Yakov & Partners and the Alliance in the field of artificial intelligence. According to the publication, experts analyzed the regulatory framework and the degree of freedom for development companies in various countries of the world.

Russia became one of the countries with the most favorable conditions for the development of GII

In the course of the study, it was revealed that in Russia, as well as in other leading countries, the most favorable conditions for the development of AI and attracting investments in this area have developed. According to analysts, this contributes to the active growth and adoption of GII technologies.

The researchers also identified five main risks associated with the development of GII. Among them are an increase in the volume of low-quality content, potential harm from false responses of AI systems, a negative impact on the labor market, an increase in digital fraud and violation of ethical standards.

As the publication notes, most of these risks can be minimized by marking generative content, creating services to check generated elements and responsible user attitude to working with the technology.

In 2024, Alliance members in the field of artificial intelligence supplemented the AI Code of Ethics with a declaration on the responsible development and use of GII-based services. According to Izvestia, analysts also emphasize the need to improve the digital and legal literacy of the population to minimize the risks associated with the use of GII.^[1]

Why Russian companies are slowly introducing generative artificial intelligence

Generative artificial intelligence is rapidly developing due to the fact that it allows you to solve many professional problems. In Russia, so far it is hardly possible to talk about the massive spread of such technologies in business, but there are more and more projects. To understand why companies in the Russian Federation are in no hurry to introduce generative AI, TAdviser interviewed market participants in June 2024.

According to the forecasts of the FinTech Association, one of the trends in the next two years will be the democratization of generative AI, new business models based on it, ensuring technological sovereignty in this area and support from the state. However, by 2024, most companies in the Russian Federation are at the stage of experimenting solutions based on this technology, less than 20% use it.

One of the problems is the lack of artificial intelligence specialists. First of all, this is due to the fact that a full-fledged education in this area requires from 2 to 4 years (master's or bachelor's degree), says Eduard Zhdanov, dean of the Faculty of Artificial Intelligence at Synergy University. According to him, there are professional retraining programs, but they are quite expensive, there are many short-term courses in the educational market, but they are intended for superficial acquaintance.

Example of image generation by the Kandinsky 3.1 neural network developed in Sberbank

Another barrier that hinders the development of generative AI in the corporate sector is associated with insufficient quality results of technology. ITMO The University believes that in order to improve quality, subject adaptation of generative models for the purposes of a particular company is required.

It is necessary to further learn the AI model based on industry data (at least at the level of correct use of terminology or graphic elements), provide manufacturing (the ability to understand user questions, including professional jargon), and also "make friends" with the model with specific knowledge already available in the company: teach to "read" regulatory documents, make engineering calculations, etc., in order to replace the "hallucinations" of AI with objective data, "Alexander Bukhanovsky, director of the Megafacultet of Translational Information Technologies at ITMO University, told TAdviser.

According to him, the cost of such adaptation is within the power of only large companies that have a lot of data, and the solution can then be replicated. For small companies, "household" functions of generative AI are still available, but the development of automatic machine learning technologies for large language models can soon change the situation, Bukhanovsky added.

The development of generative AI in Russia is also constrained because the company does not fully understand these technologies and the specific possibilities for their use in business. Many have heard about generative neural networks, but only a few have tried to work seriously with them and really understand their potential, said Andrei Malov, General Director of the TTK-Cloud IT company.

He also drew attention to such a factor as the threat to information security. The corporate application of generative AI implies working with confidential documents - contracts, internal regulations that should not fall into the wrong hands. The use of public commercial neural networks may be unsafe, and the only option is the internal development of a neural network in a closed circuit. But then the entry threshold for the corporate application of generative AI increases greatly, Malov argues.

Companies in Russia prefer AI solutions to be deployed on their own servers, confirms the founder of the service, Fabula.ai Ali Ozdiev. They are not ready to send company data for processing to third-party services. Most often, they are guided by the internal charter and policy of the company, for some companies it is extremely important that all data is processed within its own infrastructure, he added.

The cost of introducing/adapting (CAPEX - capital expenditures) and using (OPEX - operating costs) artificial intelligence is really very high, agrees Maxim Milkov, head of AI at Softline Digital. On large flows, this limits the number of economically feasible projects. In addition, the use of generative AI requires significant and at the same time hard-to-reach computing power, Milkov added.

Training in artificial intelligence also takes time, especially since the initial stage of AI development gives out a lot of marriage, says Dmitry Tretyakov, general director of the Getloyalty service.

Remember, when artificial intelligence graphics programs first appeared, they generated people with six fingers and five eyes. The work of artificial intelligence needs to be carefully tested, monitored and trained, especially at an early stage, he said.

TAdviser respondents also mentioned such a problem as restricting access to foreign technologies and resources. Alexei Lebedev, head of the financial sector at RNT Group, believes that by June 2024 there are no domestic AI models in the Russian Federation that can compete with foreign solutions. Infrastructure and equipment that allow you to fully deploy AI models and ensure work with them are not supplied to the Russian Federation, Lebedev said.

In terms of the availability of equipment, we are still forced to proceed from sanctions restrictions on the supply and availability of graphics accelerators. This issue will have to be resolved at a high level, - said Ilya Petukhov, head of projects for the development of AI products at Directum.

Nevertheless, organizations in Russia are actively introducing partnerships with technology companies and startups in order to gain access to new developments and solutions in the field of generative AI, says Dmitry Vladimirov, director of the product office of BIA Technologies. In his opinion, an important step is also the exchange of experience and knowledge between companies and the stimulation of innovation through government support and investment in research and development.

Reinforcement training allowed generative streaming neural networks to work better

Scientists at the faculty's Center AI and Institute artificial intelligence for Digital Sciences computer sciences HSE have applied classical algorithms reinforcement training to tune generative streaming (networks GFlowNets). This made it possible to improve the work of GFlowNets, which have been used for three years to solve the most difficult scientific problems at the stages of modeling, generating hypotheses and experimental design. The HSE announced this on June 13, 2024. More. here

Generative AI launched on State Services

Generative artificial intelligence was launched at the Public services. This was reported to the Ministry of Digital Development of the Russian Federation on May 22, 2024. Read more here.

How generative artificial intelligence is transforming the beauty industry

In early April 2024, the French manufacturer of perfumes and cosmetics L'Oreal spoke about the advantages of generative artificial intelligence (GENI) technology can bring to the beauty industry. The company believes that such tools can revolutionize the industry. Read more here.

How generative AI will help Russian oil and gas companies earn an additional 343 billion rubles

The total effect of the use of generative artificial intelligence (Genii) for Russian oil and gas companies can be up to 343 billion rubles a year. This can be achieved by improving labor productivity and production efficiency, as stated in the study "Vybon Consulting," the results of which were published on April 3, 2024.

It is noted that oil and gas companies are actively using deep learning AI to work with large structured numerical data. At the same time, as of the beginning of April 2024, there are no industrial solutions based on Genii in the domestic oil and gas sector. Meanwhile, the introduction of such technologies allows you to cover a larger amount of information, reduce the time for its analysis, and increase the speed and quality of decision-making. GeneII, in particular, will be able to increase the automation of functional processes related to engineering and scientific and technical expertise. The report says that up to 59% of applied expertise can be automated. In general, the potential for automation of the work of professional groups related to engineering, when using neural networks, is estimated at 57%.

The total effect of the use of generative artificial intelligence (Genii) for Russian oil and gas companies can be up to 343 billion rubles per year

Analysts believe that out of 343 billion rubles of the total effect of the introduction of Genii, the largest share (about 69%) will fall on such areas as geological exploration, drilling, development and capital construction, monitoring and production management. At the first stage, Genia will be able to solve relatively simple problems: this is the recognition of text documents, search in the knowledge base, as well as the generalization of documents and the construction of key conclusions.

It is also said that the introduction of appropriate technologies in Russia is hindered by sanctions restrictions and the high cost of systems. Thus, the total investment in Genii in Russia is enough to create only one megamodel of the 2023-2024 level, while in the future they will need to increase, since the costs of creation and development can reach 100 billion rubles.^[2]

Possibilities of using generative AI to ensure the technological sovereignty of the Russian fuel and energy complex

How generative AI develops in Russia

On March 1, 2024, the FinTech Association (AFL) presented the results of a study on the development of generative artificial intelligence (GENI) in Russia. The published report is designed to simplify the interaction between developers of products and solutions based on AI and representatives of the financial market, as well as help create innovative solutions in the relevant area.

The authors of the report describe the structure of the Russian market, highlighting four basic layers. These are Genia-based products and services, tools for Genia, models, and infrastructure. The first of these categories includes horizontal and vertical application solutions available locally and in the clouds. Horizontal applications include applications in industries such as sales, productivity, finance, customer service, information technology, and knowledge management. Vertical and industry applications cover areas such as fintech, law, science, retail, health, transportation, PR/marketing and robotics.

GeniAI tools include tools for efficient and safe industrial use of basic models. In turn, the models themselves are divided into GPT solutions, domain and diffusion systems. In addition, the development of marketplaces of generative models is underway. AI infrastructure combines computing for training, programmable/specialized integrated circuits, communication networks, interconnect, basic security tools, private computing, cloud services, etc.

By the end of 2023, it is estimated that 20% of large Russian companies are using Genia solutions. For more than 60% of employees, the development of Genia will become a plus, since it will free up their time by automating routine tasks such as data collection, entry and initial processing, standard document flow, primary communications with clients, accounting, etc.^[3]

Sergey Brin called Google's new image generation service a failure

In early March 2024, Google co-founder Sergei Brin announced that the company "definitely messed up" with image generation in the large Gemini language model. Users complain about dubious answers and historical inaccuracies that the AI system allows in the process of work. Read more here.

Why banks are spending billions of dollars on generative AI

Research by Juniper Research, whose results were published on January 23, 2024, suggests that banks are rapidly increasing spending on generative artificial intelligence (GenAI). The implementation of such tools improves the quality of customer service and improves the level of security. Read more here.

How generative AI saves hundreds of billions of dollars to hospitals and clinics

By 2025, generative artificial intelligence (GenAI) will free up up to 10% of doctors' time, as well as help healthcare facilities save hundreds of billions of dollars. This is stated in the IDC study, the results of which are presented on January 23, 2024. Read more here.

2023

How the cost of generative AI is growing in the world

In 2023, companies' spending on generative artificial intelligence (GENI) averaged 4.5% of the total IT budgets. In the future, as expected, the indicator will grow and by 2027 will reach about 7.6%. This indicates the increasing importance of AI technologies in the corporate sector. Market trends are addressed in a Boston Consulting Group (BCG) survey published July 16, 2024.

Genia tools are rapidly gaining popularity. According to BCG estimates, in 2024, total investments in this area will rise by 30% compared to the previous year. Moreover, the ROI of enterprises that invest heavily in Genia projects will be about three times higher by 2027 than that of companies that prefer not to spend significant amounts on such developments. The main obstacle to the introduction of Genia is the immaturity of the technology. In addition, serious concerns among organizations raise risks related to data security and possible legal problems. The study said the overall IT budgets of businesses are increasing, yet costs remain uneven. The focus is on security and digital transformation.

Although companies are focusing on cloud and security, GeniI technologies are coming to the fore as businesses look to improve productivity, the BCG review notes.

As part of cost control efforts, IT managers plan to consolidate suppliers of almost all products, especially those related to infrastructure. In particular, 33% of respondents indicated consolidation of purchases in the field of storage, and 36% - in the field of servers. At the same time, a different picture is observed in the field of Genia and machine learning: 42% of respondents plan to expand the number of suppliers, and only 13% expect consolidation. Given that Genia technologies are relatively young, but at the same time are rapidly developing, companies need to interact with various developers and providers to find the tools that suit them in the best way. This, in turn, forces Genia suppliers to differentiate in order to attract and retain customers.

The authors of the study talk about a stable increase in interest in Genia tools in the corporate segment. As of the beginning of 2024, only about 20% of companies use Genia insignificantly or do not apply such solutions at all. For comparison, in the third quarter of 2023, this figure was 24%. Although the share of enterprises with a high level of GI implementation during the period under review remained the same (approximately 12%), the number of companies with an average level of implementation rose from 18% to 27%. In the technology sector, approximately 62% of companies show a high and average level of maturity in terms of the use of Genia. This is followed by banking, retail, the industrial sector and healthcare, where from 32% to 39% of organizations have a significant level of introduction of generative artificial intelligence technologies.

From a geographical point of view in North America and Europe, the process of introducing Genia shows in many respects similar trends: about 40% of companies reported medium and high maturity. In Asia, the results are slightly better: 45% of organizations indicated a significant level of introduction of Genia. In addition, in Asia, only 16% of respondents said that they use Genia insignificantly or do not use it at all. For comparison, in North America and Europe, this value is 18% and 23%, respectively.^[4]

China is 6 times ahead of the United States in the number of patents for generative AI

In the period from 2014 to 2023, about 54 thousand patent applications for technologies related to generative artificial intelligence (GENI) were filed on a global scale. Moreover, China leads in the number of patents in the field under consideration, six times ahead of the United States. Such data are provided in the report of the World Intellectual Property Organization (WIPO), published on July 3, 2024.

It is noted that GENII tools are used in such industries as biomedical sciences, industrial production transport, security and telecommunications. During 2014-2023, more than 38 thousand inventions in the field of Genii were registered in the PRC. In the United States, which is in second place in the number of patents, approximately 6,300 inventions were registered during this period. Further in the ranking are (South Korea 4155 inventions) and (Japan 3409). In (India 1350 inventions), which is in fifth place, there is the highest average annual growth rate among the leaders - 56%.

In the period from 2014 to 2023, about 54 thousand patent applications for Genii technologies were filed on a global scale.

The top ten companies and organizations applying for patents in the field of Genia include Tencent (2074 inventions), Ping An Insurance (1564), Baidu (1234), Chinese Academy of Sciences (607), IBM (601), Alibaba Group (571), Samsung Electronics (468), Alphabet (443), ByteDance (418) and Microsoft (377).

The total mass of GeniI patents is dominated by solutions related to image and video processing (17,996 inventions). This is followed by text (13,494 inventions) and speech/music (13,480 inventions). The number of patents for technologies using data on molecules, genes and proteins is rapidly increasing: the average annual growth rate is recorded at around 78%. In general, from 2017 to 2023, the number of patents in the field of Genia increased eight times.^[5]

Named 10 main trends in the field of generative AI for business

Generative artificial intelligence (GenAI) technologies will allow companies and organizations from different sectors to optimize processes, increase staff productivity, save time on routine tasks and reduce costs. On November 29, 2023, IDC analysts named 10 key trends in GenAI for business.

1. GenAI to help drive innovation

Generative AI will be used to jointly develop digital products and services by identifying market prospects and allocating the company's resources. Enterprises will be able to more effectively implement new projects, which will help increase income.

2. Digital investment growth to continue

In 2024, IDC believes, companies' spending on digital solutions will grow seven times faster than the economy as a whole. The transformation of the market stimulates enterprises to develop new business models and strengthen digital opportunities.

3. AI at the senior management level

According to an IDC survey, more than half of Chief information officers say their organization has or is planning a head of AI. Moreover, in many cases we are talking about the inclusion of such specialists in the top management (at the level of the executive team).

4. Generative AI in digital enterprises

Digital transformation companies will continue to implement GenAI tools to improve their business models and create additional competitive advantage. Digital enterprises, according to analysts, will actively invest in GenAI applications in 2024.

5. Digital Business Platforms

Such solutions optimize the analysis of the company's activities, allowing you to better assess the effect of investments. As enterprises digitalize, business platforms of a new generation can contribute to the growth of financial indicators.

6. AI stimulates the emergence of new digital business models

IDC believes that the combination of predictive AI, machine vision and GenAI technology, as well as the provision of on-demand services through digital ecosystems, will open up new business opportunities. Companies will be able to create products and services taking into account the interests of certain groups of customers.

7. Implementation of new performance indicators

Analysts believe that against the background of the ongoing digital transformation, companies will begin to apply new criteria for assessing the effectiveness of operations. This will help in making strategically important decisions.

8. Digital initiatives come to the fore

IDC believes IT executives will focus on improving business results, increasing operational flexibility and creating new revenue streams through the adoption of digital services and services.

9. AI will have an impact on workflows

The massive introduction of AI will create certain difficulties for employees, affecting established operational processes and training. Therefore, employees of companies will have to undergo retraining and learn skills to interact with GenAI platforms.

10. Digital technologies will help achieve sustainable development goals

Companies, according to analysts, will have to carefully plan investments to achieve business goals while ensuring sustainable development. This will make it possible to effectively use the available resources with an eye to long-term successful expansion of activities.

In general, IDC notes, the industry is entering the era of digital business: companies are looking for new sources of revenue, while digitizing operations to reduce costs and increase return on investment.^[6]

How generative AI will develop in 2024. 6 Top Trends

The introduction of generative artificial intelligence tools creates qualitatively new opportunities for both organizations and consumers. Such tools help improve the efficiency of business operations, improve the quality of customer service, and reduce the time and financial costs of performing routine tasks. In 2024, the generative AI market will continue to actively develop, and up to 60% of skeptics negatively related to technology, in one form or another, will begin to use or evaluate it. This is stated in a report by Forrester Research, published on October 24, 2023. Analysts name six key trends in the global generative AI market.

1. Improve customer experience

Generative AI tools, according to Forrester, in 2024 for the first time in many years will improve the level of customer service. Support services will be able to answer questions faster and better and solve many user problems at the first time. At the same time, the waiting time for calls to call centers will be significantly reduced, and the load on support services specialists will decrease. It will be possible to automate the summarization of cases and provide instant accurate answers using chatbots with information extraction capabilities.

2. Personalized AI Solutions for Enterprise Customers

In 2024, the 10 largest advertising agencies will spend a total of approximately $50 million on the creation of individual AI systems that will allow their customers to scale marketing campaigns, increase operational efficiency and attract new customers.

3. Cloud Tip Technique

In 2024, according to analysts, many hyperscalers and cloud platforms will introduce prompt engineering in one form or another. It is an AI concept aimed at natural language processing in particular. In the tooltip technique, the task description is embedded in the input rather than being implicitly specified. However, Forrester believes, despite the introduction of prompt engineering by cloud platforms, 80% of enterprises will involve internal specialists to solve the relevant problems.

4. Regulation in the field of generative AI

AI services create new opportunities, but at the same time create additional problems, in particular, in terms of processing personal information. Therefore, Forrester reports, companies should identify applications that can provoke incorrect use of personal data, as well as implement additional risk management tools.

5. Changes in the B2B sphere

Despite the advent of new AI-based recommendation systems, many customers in the B2B segment prefer personal communication with product specialists, considering such interaction more valuable and effective.

6. Reduce investment in improving employee experience

As generative AI platforms take on monotonous functions, companies will begin to reduce investments in improving employee experience, that is, in improving the quality of employee interaction with their organization. In general, Forrester emphasizes, generative AI, despite certain risks, will serve as the support that enterprises rely on to expand capabilities, improve business processes, and attract employees and customers. In 2024, analysts say, those companies that will introduce generative AI and continue to experiment with this technology will achieve the greatest success.^[7]

How much companies in the world spend on generative AI

In 2023, companies around the world will invest about $16 billion in generative artificial intelligence (AI), analysts at IDC predict ( the study was published on October 16, 2023). In their calculations, they included the costs of equipment, software, as well as IT and business services related to generative AI.

IDC Vice President Ritu Jyoti says generative artificial intelligence is not some kind of fleeting trend or hype. This is revolutionary technology that can have a long-term impact on business.

Thanks to ethical and responsible implementation, generative AI can change markets, change how we work, play and interact with the world, the expert added.

Such AI systems require infrastructure, including hardware and services such as IaaS (Infrastructure as a Service) and SIS (System Infrastructure Software). It is the infrastructure that occupies most of the costs of companies when investing in generative AI, and this situation will remain in the future, analysts at IDC, which released a market study in October 2023, say. At the same time, the highest growth rates in the market are predicted in the software segments (+ 76.8% per year in the period from 2023 to 2027), AI platforms and models (+ 96.4% per year) and application development (+ 82.7% per year).

IDC experts expect that business spending on generative artificial intelligence in the horizon 2023-2027. will grow at an average annual rate of 73.3% - more than twice as fast as spending on AI as a whole, and almost 13 times faster than spending on IT spheres. In 2027, the cost of generative AI will amount to 28.1% of all investments in the field of AI as a whole, compared to 9% in 2023, the study says.

According to IDC analyst Rick Villars, the growth rate of spending on generative intelligence will be restrained until 2025 due to instability in changing work processes and resource allocation related to chips, network communications, AI skills, etc. In addition, high prices will have a negative impact on the market, concerns of companies about data security and privacy. Investments in the market will develop naturally, following how organizations are moving from early experiments to active use and widespread adoption of technology, according to IDC.

Revenues of the generative artificial intelligence market may increase to $1.3 trillion by 2032, according to a report by analysts at Bloomberg Intelligence. This is 32 times more than this market brought in 2022, when profit amounted to $40 billion.

Analysts believe that the generative artificial intelligence sector will experience explosive growth within 10 years, which could fundamentally change the way the technology sector operates. According to Bloomberg calculations, this sector can grow at a CAGR of 42% over 10 years, which explains the demand for infrastructure for training neural networks, as well as devices with artificial intelligence models, advertising and other services. The infrastructure market for AI training could reach approximately $247 billion by 2032, Bloomberg predicts. Income from Digital Signage can reach $192 billion by this time, and income from artificial intelligence servers - $134 billion, the report says. The main beneficiaries may be Amazon's cloud division, Alphabet (Google's parent company), microchip maker Nvidia and Microsoft, the researchers said.^[8]

How companies are implementing generative AI in the world - Gartner study

In 2023, less than 5% of companies and organizations globally adopt software interfaces (APIs) or generative artificial intelligence ( GenAI) models and/or implement appropriate applications in their operating environments. However, by 2026, this figure could exceed 80%, as stated in a study by Gartner, the results of which were published on October 11, 2023.

Generator AI has become a top priority for corporate top executives and has led to new tools that go beyond basic models. Demand for GenAI applications and services is growing in many industries, such as healthcare, life sciences, legal and financial services, the public sector, "said Arun Chandrasekaran, vice president of Gartner.

Analysts identify three key areas that will have a significant impact on organizations using or implementing AI tools. One of them is GenAI-enabled applications. These tools help you solve everyday tasks and speed up the achievement of your goals by automating routine processes. As of October 2023, the main application area of ​ ​ GenAI-based applications is the analysis of large amounts of text information using natural language requests. Generative AI helps to extract the right data, create new content and ideas, including conversations, stories, images, videos and music. In addition, GenAI tools are used to improve the quality of digital images, edit materials, quickly create prototypes for production, etc.

The second direction is the base models. These are scale AI models pre-trained on huge amounts of data. These, in particular, include large language models (LLMs), specifically focused on performing tasks such as generalization, text generation, classification, answering questions, etc. Gartner estimates that by 2027, base models will underlie 60% of natural language processing platforms versus less than 5% in 2021. Basic models will find application in a variety of areas. For example, in the financial sector, institutions will be able to use AI-based chatbots to improve customer service, generating product recommendations and responses to user requests. Credit institutions will be able to accelerate the issuance of loans using basic models in underserved markets, especially in developing regions. In addition, basic models will help accelerate innovations in healthcare, automotive, power, telecommunications, etc. Entertainment companies can use generative AI to develop next-generation games and empower users to create deeply personalized avatars.

Gartner experts call trust, risk and security management (AI TRiSM) the third key area. This is an important basis for creating responsible AI. It's about managing AI tools, ensuring reliability, fairness, efficiency and data protection. In addition, ethical standards and user confidentiality must be observed. Analysts believe that organizations that do not pay due attention to AI TRiSM funds will face a variety of problems, including financial and reputational losses, project failures and the inability to achieve their business goals, social harm and possible litigation.^[9]

How generative AI is used in the fashion industry

Generative artificial intelligence technologies could have a major impact on the fashion industry. Such systems will help to bring clothes and accessories to market faster, sell them more efficiently and improve the quality of customer service. This is stated in a study by McKinsey, the results of which were released in early March 2023.

Analysts note that the fashion industry is experimenting with basic AI and other advanced technologies: these are, in particular, metaverse, non-interchangeable tokens (NFT), digital identifiers and augmented or virtual reality. At the same time, as of the beginning of 2023, generative AI tools are practically not used in this area, despite their huge potential. According to McKinsey estimates, by 2026-2028, generative AI could add $150 billion to $275 billion to operating profit in the clothing, fashion and luxury sectors.

In our opinion, generative AI is not just automation of processes. This is also expansion and acceleration. Fashion professionals and creative people get technological tools that allow them to perform certain tasks much faster. Thus, additional time can be released for those issues that can only be resolved by people. It also means creating new services for better customer service, McKinsey said in a report.

Creating products

2022

==

==

==

==

==