I agree.

The doubt appears in my head though when I think about the extension of human knowledge.

Take any field. We know a certain amount. PhD's extrapolate on the knowledge and search for new knowledge. They test, they experiment, they theorise.

But in some areas their search is for the patterns that connects everything.

Like Alpha Fold.

We knew that proteins folded but the exact way they folded was tricky to figure out. Humes had a go and did ok, and then Deep Mind had a go and eventually taught a machine to be better at humes at folding proteins.

Then a bit past that was to use diffusion models to generate new proteins.

Busted the field wide open.

So I ponder what other area of knowledge are similar. Physics. Chemistry. Maths. How many of them have patterns hidden in the data that we can't see but that an agentic/LLM/MoE etc machine might find.

PhD's won't disappear, but they are at threat because in the end someone has to pay them and if the people with the money think the machines can do the task, then the PhD's will be out of a job too.