Voice and chatbots for business. TAdviser Overview

The development of the intellectual chatbot industry in Russia as a whole follows global trends. ChatGPT rightly enjoys market authority, the release of this product is called a key event. The emergence of LLM models influenced the development of the industry for years to come.

One of the leading may be the NLP (Natural Language Processing) trend. Another equally important trend is the development of multimodal models.

The impact of generative AI on corporate communications ecosystems will only grow towards the integration of voice and text robots with corporate CRM, BI systems and databases. The trend on information security in all areas deters companies from using foreign cloud solutions, while work on installing LLM in the customer loop is intensified.

Companies will increasingly use user data to create personalized experiences. The focus shifts to a comprehensive customer engagement strategy for the entire duration of CJM and seamless handling of client data.

The level of personalization in marketing is growing, and at the same time risks for the security of personal data are growing. The use of AI also raises questions of ethics and morality. The demand for digital doubles of media people is growing. This affects the field of copyright. In social networks, minds excite deepfakes.

Not all experts are unequivocal in their positive assessment of the future of AI due to the limitations and extensive nature of the technology development.

The future of the dialogue assistant industry is now closely intertwined with generative artificial intelligence, experts say. According to the August 2024 Gartner^[4]geneAI^[5] passed the "peak of inflated expectations," and business attention continues to shift from the excitement around base models to specific application options that increase ROI. In this regard, Maxim Ivanov the director of artificial intelligence Sberbank Business Soft noted that the "hype" trend on LLM is gradually ending, and it is replaced by the appearance of the first applied B2B and B2G cases.

The construction of new data centers has already begun, and investments in their development are colossal, he said. - We all have to learn how to use LLM as a working tool in our lives and in business with maximum benefit.

Alexander Pavlov, Managing Director of Robovoice of SL Soft, also believes that the superpopularity of AI is replaced by a stage of real implementation. New technologies are showing better results where the ceiling would seem to "reach." Despite existing limitations, such as the need for large computing resources and data volumes, the development of AI does not stop, but, on the contrary, investments in infrastructure and improvement of models are intensifying.

The observed cooling of the hype may also be due to current LLM constraints. Researchers from the Higher School of Economics checked the GPT-4 for reliability in the framework of teaching and came to the conclusion that the model could be mistaken even in the simplest tasks.^[6]

You never know where to wait for mistakes, - commented Yaroslav Kuzminov, scientific director of the Higher School of Economics.

According to the expert, at the moment GPT-4 is an assistant for whom everything needs to be rechecked, which is not particularly encouraging.

Vladislav Belyaev believes that the key natural limitation of the development of AI on the current technological track is the inaccessibility and high cost of the necessary amount of resources for its training and work. Resources should be understood as suitable video card chips, the monopoly on the production of which (and high prices) is held by Nvidia. Therefore, there are discussions about the construction of dozens of new chip factories from other manufacturers.^[7]. The amount of energy consumed by these video cards should also be taken into account. To ensure the right amount of electricity, AI leaders are also investing in promising generation technologies.^[8]

At the same time, the expert noted a certain cooling of rush interest due to the low economic efficiency of large-scale AI projects.^[9]

We think that within 2-3 years there will be an optimal balance between the continued growth of AI costs and products for specific business tasks, which we are now working on, "he said.

On the other hand, tensions are growing in various social groups due to the consequences of the introduction of AI and its unethical use, said Nikita Murenkyi, head of the Kode dialog product development team. According to him, business and states are working to determine the legal framework and ethics of the application of AI, and this work does not tolerate delays.