IFC with GPT

We already explored ChatGPT’s ability to generate the IFC schema. We tried “zero-shot” and “in-context learning”. The results were – in my opinion – impressive. Nonetheless, there is room for improvement. In previous experiments, GPT’s job was to provide a valid IFC schema. This time, we take a different, “indirect” approach. Instead of generating an […]

OpenAI released a new flagship model, GPT 4o, also in the chat version. GPT4o is claimed to be faster and more efficient than its predecessor, GPT4. Importantly, it is available to all users, not only the premium ones. It is a good occasion to perform another IFC-generating experiment. ChatGPT4 did well in our last experiment,

ChatGPT is an AI model created to work with language. It understands and generates natural language content, such as messages, articles, or descriptions. It is pretty good also in other “languages”, such as programming code. After all, both natural languages (e.g., English, Mandarin) and programming languages (e.g., C#, Python) have their own logic, conventions, and