If you're interested in AI, this interview will not disappoint. Ian Goodfellow shares his personal story, working his way from humble grad student (building machines in his mother's garage) to staff research scientist on the Google Brain project.
Credit: Andrew Ng's "Preserve Knowledge" channel on YouTube.
Goodfellow was the first to figure out how to run ML code faster by using a computer's GPU (ordinarily reserved for computer graphics and gaming) rather than the CPU (the central processor that normally runs software applications). But his real claim to fame is inventing the generative adversarial network (GAN), a neural network architecture that enables a computer to have an imagination, of a sort.
A GAN is composed of two neural networks pitted against each other. The discriminator network trains on a set of digital images to learn what a particular type of image looks like. The generator network then starts trying to generate that same type of image, starting with a multivariate normal distribution for data. Both networks are trained with backpropagation. The goal of the generator is to get better and better at dreaming up images that look like the ones in the original training set. And the goal of the discriminator is to get better and better at being able to spot synthetic images (those created by the generator) versus real images.
The video is short (about 15 minutes) and is very approachable. You won't find any complex math formulas here, just the personal account of someone who made a big impact on AI research, including some interesting stories about how Goodfellow came up with some of his ideas.