HACKER Q&A
📣 mark_l_watson

What happened with Cyc now that we have LLMs


Years ago, I enjoyed experimenting with OpenCyc, and followed what Cyc Corp was doing.

To me it seems like their goal of developing common sense reasoning has been at least partly subsumed by the common sense knowledge contained by LLMs like ChatGPT, Llama models, Mistral, Anthropic, etc. The knowledge in LLMs, trained on what I think of as a ‘shadow’ representation of the world, the shadow being text, seems adequate to build some form of representation of the real world.


  👤 h2odragon Accepted Answer ✓
I'd think its more valuable than ever as a check against LLM's to see they're not completely off the rails.

If they cannot demonstrate the connections and base that Cyc has, then the training is probably gone wrong somewhere.