> ChatGPT has several billion conversations every day. I estimate that the model that solved Navier–Stokes was trained on data from nearly a trillion conversations.

How many of those trillion conversations were about Navier-Stokes you reckon?