You could combine Cyc with an LLM using embeddings. First you select from the Cyc database which entries are relevant to your prompt. This may use low-level ai

You could combine Cyc with an LLM using embeddings. First you select from the Cyc database which entries are relevant to your prompt. This may use low-level ai techniques, e.g., do a word embedding and look for the database entry close to the prompt. Then add the relevant data entries data you found as context to the prompt. The LLM will then have the additional information. This is already used to generate answers to support questions. You look for relevant FAQs in a knowledge base then have the LLM write a specific answer with the additional knowledge base information. I doubt Cyc would add much though. As I recall it was filled with trivial world information. Like you typically shave in the morning. People need to eat. Things that ChatGPT has absolutely no problem with.