AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is There a Dance Between a Bigger and a Smaller Model?
There's this dance between compute getting cheaper, but also, at the same time, these algrams getting more efficient. mgogle just came out with their language modeon 25 x more efficient on a compute basis than previous models. And it's interesting to think of when the limit actually is hit, now that everyone is trying to just go as big as possible. As things get ynear, the rubber starts to meet the road more and more on applications, and it takes less n less expertise to leverage these things too.