I'm not sure where you get 'we should spend less money on ML' from the abstract....

Veedrac · on Dec 4, 2020

> We end with recommendations including weighing the environmental and financial costs first

‘Train for less time,’

> investing resources into curating and carefully documenting datasets rather than ingesting everything on the web

‘on less data,’

> carrying out pre-development exercises evaluating how the planned approach [...] supports stakeholder values

‘while spending less’

> and encouraging research directions beyond ever larger language models

‘on smaller models.’

pseudalopex · on Dec 4, 2020

Choosing worthwhile projects doesn't mean spending less.