4 ms·
> Reasoning and abstraction and high level AGI concepts that I don't think apply to deep learning What do you mean? Each layer is selecting features and abstra
by nojvek 9y ago
> Reasoning and abstraction and high level AGI concepts that I don't think apply to deep learning
What do you mean? Each layer is selecting features and abstracting previous layers. A cat neuron abstracts all possible pixels that form cats.
He's just saying that we need better ways of composing this `deep` abstractions.
- espadrine 9y ago> we need better ways of composing this `deep` abstractions Tensor2Tensor[0], from the Google Brain team, has some strong recent results in that direction. Related paper: "One Model To Learn Them All"[1]. [0]: https://github.com/tensorflow/tensor2tensor https://github.com/tensorflow/tensor2tensor [1]: https://arxiv.org/pdf/1706.05137.pdf https://arxiv.org/pdf/1706.05137.pdf