Setere Group CEO Oleg Ivchenkov expressed
Posted: Mon Jan 20, 2025 8:00 am
cording to Sergey Lunegov, Director of Products at Axiom JDK, 50% of Java developers use AI for testing. "AI copes well in situations where it is necessary to go through a billion different variants of code or library operation to detect an error or vulnerability. AI is ideal, for example, for fuzzing, one of the most effective methods of dynamic testing, which is included in the FSTEC certification requirements and is actively used by our team as part of the industrial process of developing secure software - such scenarios for using AI, according to the expert, are prevalent. - AI is very convenient and well suited for south korea whatsapp resource solving typical tasks, however, a qualified engineer is needed to form a correct request and evaluate the quality of the AI solution. Typical tasks can be: a request to create code that implements a certain algorithm, a request for a sample server configuration, database, etc. AI copes well with tasks that are often encountered in the work of many people and which are discussed a lot. AI has an infinite number of use cases. And this can significantly reduce labor costs, but AI must be implemented under vigilant supervision."
doubt about the percentage of supporters of using AI: "Our experience shows that in development and testing, the most reliable use of AI is in composing birthday greetings for employees. Otherwise, rechecking will take more time than the initial work of people on the task. We are very concerned about the reliability of products, and this has not failed us so far."
Roman Smirnov, Commercial Director of Devlonika (Softline Group), believes that the combination of "neural network + human specialist" already provides an increase in productivity, but is still far from widespread use, so developing software using only the neural network is difficult and probably even impossible: "There are inaccuracies and incorrect values that AI can produce when generating code, there are limitations associated with high security and confidentiality requirements on projects, and many other complex issues. GPT, whatever it is, will not write you a ready-made ERP system, but it will help to significantly reduce development time if the tool is used correctly by trained employees. Based not only on the data presented, but also on our research conducted in 2024, we can talk about reducing labor costs by 20-50%. The result largely depends on the direction and specifics of the work, as well as on the choice of the tool used."
Dmitry Svalov named the main
doubt about the percentage of supporters of using AI: "Our experience shows that in development and testing, the most reliable use of AI is in composing birthday greetings for employees. Otherwise, rechecking will take more time than the initial work of people on the task. We are very concerned about the reliability of products, and this has not failed us so far."
Roman Smirnov, Commercial Director of Devlonika (Softline Group), believes that the combination of "neural network + human specialist" already provides an increase in productivity, but is still far from widespread use, so developing software using only the neural network is difficult and probably even impossible: "There are inaccuracies and incorrect values that AI can produce when generating code, there are limitations associated with high security and confidentiality requirements on projects, and many other complex issues. GPT, whatever it is, will not write you a ready-made ERP system, but it will help to significantly reduce development time if the tool is used correctly by trained employees. Based not only on the data presented, but also on our research conducted in 2024, we can talk about reducing labor costs by 20-50%. The result largely depends on the direction and specifics of the work, as well as on the choice of the tool used."
Dmitry Svalov named the main