Semiconductors are characterized by a large number of acquired data in all phases, including product design, prototyping, and mass production.
For example, in the area related to semiconductor elements called the front-end process, the number of processes is well over 100, and there are many management items in each process, so the number of data to be handled is enormous.
GPT will undoubtedly alleviate this data collection phase and the work of graphing and creating explanatory materials. More than that, however, I would like to expect a future in which LLM will be used for detecting abnormalities in the manufacturing process.
In order to make the production line work well and reduce defective products, a huge number of parameters (variables) must be managed, and their management is beyond human knowledge.
Therefore, if AI analyzes the correlation between parameters like genome analysis (although it is debatable how to train the data of the company's own production line), we may find hints for improving the quality rate that we have not noticed before.
In other words, if "genome analysis of production lines" progresses, there is a possibility that the semiconductor manufacturing cycle itself, which still takes about four months, can be shortened. This is an evolution that greatly affects the business feasibility of the semiconductor industry, which is often plagued by business cycle problems.