30:28Adam Coates explains how Baidu's AI Lab connects basic research to products, with a mission to make AI technologies useful to at least 100 million people. He describes Deep Speech, a deep learning speech recognition system trained with large amounts of audio, and how the team adapted research models to meet production needs such as low latency. He also discusses Deep Voice and the challenge of building speech systems that need less labeled data and work across accents, noisy settings, and long conversations. The talk closes with the need for adaptable, self-directed researchers who can carry ideas from research through hardware and product development.