I too doubted, from the beginning, that neural networks will be the basis of AGI. As impressive and useful as LLM's are, they are still a long, long way from AGI.