AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How to Increase Data Efficiency With a Few Shot Learning
The most exciting thing over the past year has been few learning. So all kinds of really clever tricks for driving down that cost in terms of data to get some performance measurement. When you want to make a class fires o document comes in and you tell me whether it belongs in bucket a, b or c, this is like n p task one, o, one. Just literally spell it out right there in in the input data, what the possible choices are. It's such a nice, simple trick. And that's what i mean when we talk about incremental science: bunches and bunches of little artistinal improvements to how we do things.