4 ms·
Dah, that's nothing. We have models that we built specifically to determine the best way to train other models. After 2 years of tweaking these models, we learn
by hellofunk 10y ago
Dah, that's nothing. We have models that we built specifically to determine the best way to train other models. After 2 years of tweaking these models, we learned that the best way to optimize our learning was to build new models that train these models for determining how to train our other models.
- beachbum8029 10y agoSo it's models all the way down?
- randcraw 10y agoOf course. Even Plato knew that. But since most models are wrong... there's life yet in us 'bags of mostly water'.
- Capt-RogerOver 10y agoModels are not wrong. They are just models. Ie, != reality.
- notgood 10y agoUntil we realize such matrix of models analyzing each other is what we tend to call "consciousness"
- joshvm 10y agoThe obvious citation is the wonderfully titled Learning to learn by gradient descent by gradient descent https://arxiv.org/abs/1606.04474 https://arxiv.org/abs/1606.04474