Turns out using 100% of your AI brain all the time isn't most efficient way to run a model

Neural net devs are finally getting serious about efficiency Feature If you've been following AI development over the past few years, one trend has remained constant: bigger models are usually smarter, but also harder to run....