Short excerpt from a YouTube interview (Geoffrey Hinton) where the speakers discuss near-term excitement in deep learning: improvements in machine translation, better document understanding, and semantic/theme-based search. The speaker expects substantial progress in these areas over the next 5–10 years.
What are you excited.
About right now in terms of deep learning? I'm.
Excited about this machine translation, even though I myself haven't made any direct contribution to it. Some of my students are doing it. I think over the fairly short term, we're going to get much better translators, and I'm excited for things like understanding documents. I think there's going to.
Be huge progress there over the next five to 10 years.
Instead of giving Google some words and it'll find you all the documents have those words in you, give Google some themes and it'll find the documents that are saying that, even if they're saying in different words. And so you're going to be able to search for documents by what it is the documents are claiming, independent.
Of whatever the words are. And that would be kind of useful. Sa.
