Claim in Headline numbers in landmark machine learning papers

The 2022 Chinchilla paper reports that a 70 billion parameter model trained on more data outperformed the 280 billion parameter Gopher.

Verdict: no verdict yet