30 years ago there were amazing inference platforms like CyC [0] that were supposed to address problems similar to those that are now submitted to ChatGPT like platforms.

It could deal with special cases because it had a community-of-agents architecture -specialized reasoning modules- each with its own data structure and algorithm.

Yet their size were tiny with respect to LLMs, roughly they were 1000 smaller.

What if they were scaled to today's size? Would it still be similar in capabilities to ChatGPTs and not better?

[0] https://en.wikipedia.org/wiki/SHRDLU

[1] https://en.wikipedia.org/wiki/Cyc