What is DeepMind's GPT-3 Rival Chinchilla?

It significantly outperforms Gopher (280B) and GPT-3 (175B) on a large range of downstream evaluation tasks. What!? How is this even possible?

Maybe it’s time to use an App instead of Email.

First of all in this case just go directly to the paper.

https://arxiv.org/pdf/2203.15556.pdf

Things are moving really fast in Language Models.

Check out site for comparing MMLU Benchmark numbers https://paperswithcode.com/sota/multi-task-language-understanding-on-mmlu

Note that Deepmind is also looking for a…