NVIDIA AI Podcast: An Argument in a Bar Led to the Generative Adversarial Networks
This is a tale that starts off like a joke: A researcher walks into a bar… and ends with a revolutionary advancement in deep learning.
Nobody’s laughing now at how Ian Goodfellow, a staff research scientist at Google, got the idea for generative adversarial networks (GANs).
In this week’s episode of the AI Podcast, Goodfellow explains that a major obstacle in deep learning is the need for “a massive amount of labeled training data,” which requires a lot of work to be done by humans.
“If you take a deep neural network and you teach it to read, to actually look at photos and recognize letters that it can see in the photo, it can do that about as well as a human being can,” says Goodfellow in conversation with AI Podcast host Michael Copeland. “But the process of learning to do it doesn’t look anything like the process that a human follows to learn to read.”
Read the entire article here, An Argument in a Bar Led to the Generative Adversarial Networks
via the fine folks at NVIDIA.