Deep Learning for Artificial Intelligence

It’s 2009, and why don’t we have artificial intelligence yet? A recent intriguing trend known as “deep learning” might be the key. In a deep architecture, the input is passed through multiple layers, for example pixels become edges become shapes become faces. This is similar to how, in the brain, an input signal is passed through many layers of neurons, from one to the next to the next and so on. Deep architectures for AI are plausible from theoretical, neurological, and intuitive perspectives. However, most current AI techniques are shallow. Why? People have been talking about deep learning for a while, but only as of two years ago do we really know how to train them. This talk will informally introduce deep learning. I will also suggest applications to information retrieval which are enabled by deep learning. Are you good at seeing how an important trend in one field can have an impact on everyday lives today? Then maybe you’ll have other ideas!

← All talks