Is the learning/semiotic process a 1-to-1 mapping like the biological learning/semiotic process? No, but there is analogue. One could argue that our "big data" training set is our billions of evolutionary history encoded in our biological processes (more than DNA) and our millions of years coevolving with spoken language.

What about abduction and intuition? This again is a result of our experiences, linking disparate signals together in ways that others may have not previously encountered. Two points in high dimensional space can seem to be either close or far from each other depending on the projection to a lower dimensional space.

Another aspect on the belief in how far (if even feasible) LLMs and modern AI are towards AGI probably depends on how strongly you believe Gödel's Incompleteness Theorem applies to our everyday lives. If you believe that us and everything in the natural universe is the product of some axiomatic system, then it seems that we are moving in the right direction (universal approximation on neural networks). On the other hand, if you think that there is something about our first person experience that exists outside the system that generated our physical aspects (moving towards a dualism perspective) then it is unlikely that there would be an argument to convince you that any AI system would be convincing enough to you (computers and Searle's Chinese room though experiment)