My point is, what level of confidence do we have that we are not just pattern matching engines running on superior hardware? How can we be sure the difference between human intelligence and an LLM is categorical, not incremental?