Ключевое слово: «large language models (llm)»

The article discusses two fundamental strategies for integrating the Theory of Inventive Problem Solving (TRIZ) and Artificial Intelligence (AI): “TRIZ+AI” and “AI+TRIZ”. Using the Compinno-TRIZ and AutoTRIZ software packages as examples, their capabilities and limitations are demonstrated. The authors propose a methodology for assessing the level of inventive thinking in AI based on the component model of TRIZ. It is shown that modern AI models, such as GPT, demonstrate a level of inventive thinking comparable to that of a second-year university student. The most promising areas of their application are the stages of problem analysis and refinement of solutions, whereas the generation of fundamentally new solutions remains the domain of the human expert.