Jsevillamol comments on NVIDIA and Microsoft releases 530B parameter transformer model, Megatron-Turing NLG

[Jsevillamol]

Here is your estimate in the context of other big AI models:

The top 3 training compute in the graph are Megatron-Turing, GPT-3 and AlphaZero

An interactive visualization is available here.

[Jack Stone]

Thanks—this is very interesting