Google search is really just a ranked sorting of mass data into related data, returning a pointer to the raw data.

ChatGPT is a more advanced way to compress related data into ML models, where context is largely conserved, while losing raw data.

It’s at least a couple orders of magnitude better in terms of efficiency (likely 1000x+) - yet does come with some trade-offs. Instead of access to highest ranked raw references, you now get access to an interpreted derivative (model) - which is a lossy abstraction that’s dependant on the ML models efficiency in that context.

In it’s simplistic form, it’s actually just a new efficiency in data compression, and retrieval.