Sometimes I wonder if AGI/superintelligence/whatever will be like flight, which was not successful until we stopped trying to copy nature’s flapping wings and studied flight at a more fundamental level.

Yes, but one should still acknowledge the technical success _Snowbird_:

https://news.engineering.utoronto.ca/human-powered-ornithopt...

LLMs and other neural nets are already further from biological brains than planes are to birds.