Fair enough. I wouldn't go quite that far — at absolute most I would accept since the first prototype of a neural cell — but the estimates I've seen for what data/training AGI would need if it requires a simulated re-run of modern human evolution to be trained is more than current (or at least recent) AI.