Laurent Agobian, CEO of the AI ​​company iPavlov

[tanjimajuha20]

ML/AI architect at Softline Digital Dmitry Zboroshenko believes that at this stage of development, LLM systems are still poorly suited for industrial use: "Most modern language models, including ChatGPT, are multilingual and can work with Russian. However, there are significant limitations - roughly speaking, it will not write a framework. An important nuance is that the network is capable of generating coherent text, code fragments, but it must be checked for accuracy each time: conditionally, the code may look plausible, but will not be executed, the text will be coherent and consistent, but will not reflect the actual japan cell phone number list state of affairs. Generative networks are still poorly suited for industrial use, since each result must be checked."
believes that such systems will not be able to replace programmers, and this is not their goal: "However, they will transform the IT industry into a much more creative environment and create huge new markets. Systems like ChatGPT need significant modifications to process texts in other languages, in particular Russian. The more complex the language, the more "additional training" the system requires, which ultimately comes down to large financial investments. It is clear that in the 21st century, the economies of those countries that make giant investments in the creation of pre-trained NLP models for their languages ​​will develop. Modern digital assistants created on the basis of systems like ChatGPT are capable of multiplying sales, internal production efficiency, the level and culture of mass service of any large corporation, which directly increases the country's GDP. We see this in our daily work with our industrial customers, when implementing our digital assistants and bots in a variety of economic sectors - from healthcare to government services. In our view, investing in such systems should become a national priority."

Axenix AI expert Dmitry Nazarenko points out that programming, despite the required amount of technical knowledge for successful mastery, is largely a creative profession. Considering that neural networks can only generate solutions that are, to one degree or another, combinations of data already submitted to the input, and not create something truly new, the danger of being replaced only threatens programmers whose limit is routine projects of small size and complexity. However, for generating various kinds of scripts, the use of LLM tools is very, very promising, but so far under the control of live programmers.