Comment by protocolture

My takeaway:

1. This used an awful lot of compute.

2. The solution to the issues regarding whether or not OpenAI stole the result, would normally be to move to a self hosted solution, however those researchers are unlikely to be funded for 1.