> But nobody knows how GPT works. They know how it was trained, because the training scheme was designed by humans, but the algorithm that is executed during inference was not intelligently designed but evolved, and it is implicit in the structure of the network, and interpretability has yet to mature to the point where we can draw a symbolic, abstract, human-readable program out of a sea of weights.
I object. ChatGPT executes in computer logic and is ultimately electrical signals in gates representing 1 and 0.
ChatGPT is vast and impressive, sure. Emergence[1] may get it past a Turing Test, fine. But it remains discrete logic.
In contrast, natural intelligence has not been reproduced organically, much less, fully understood. There is no repeatable experiment going from inorganic matter to self-aware, self-replicating life.
In summary, ChatGPT is impressive, but nowhere near capable of doing the impossible, e.g. predicting the weather with fidelity substantially into the future.
Nor can I bring myself to fret that Skynet is immanent.