[ramesh31]

>as with a majority of ML research

Plus "we did the same thing, but with 10x the compute resources".

But yeah.