back
1 comments
Wow.

The progress of DLNs amazes me. The recent Ng et al. Zero-shot learning through cross-modal transfer[1] paper showed how a DLN could be taught to recognise classes of objects in images (cats vs trucks vs mice etc), then taught from non-image data about other classes (ie, "reading text") and then could distinguish between cats and dogs.

Now this shows a DLN can learn to program.

They say that technology means long-term change is usually great than expected while short term change seems slower. Right now the short term change seems damn fast to me.

[1] http://scholar.google.com.au/scholar?cluster=143306265984748...