[Talk Summary 9] The Next Frontier in AI: Unsupervised Learning
Yann LeCun, Director of AI Research at Facebook, and Silver Professor of Dara Science, Computer Science, Neural Science, and Electrical Engineering at New York University, held a talk about unsupervised learning, the next frontier in AI, on Friday, 2016/11/18 at CMU.
At the beginning of the talk, prof. LeCun introduced Neuroscience, supervised learning, deep learning, multi-layer neural nets, convolutional network architecture, very deep convNet architectures, Memory-augmented networks. He presented different kinds of application using machine learning such as image recognition and question answering.
The main part of the talk presented by prof. LeCun was about obstacles to AI. The challenge of the next several years is to let machines learn from raw, unlabeled data, such as video or text. This is known as unsupervised learning. AI systems today do not possess "common sense", which humans and animals acquire by observing the world, acting in it, and understanding the physical constraints of it.
To review approaches to unsupervised learning, prof. LeCun specified the architecture of an intelligent system. The most important of AI systems is learning predictive forward models of the world because there are a lot of uncertainty. How can we deal with uncertainty is a topic of unsupervised learning. In more details, he presented energy-based unsupervised learning and stated that the best idea ever is adversarial training. In his opinion, adversarial training is the coolest idea in machine learning in the last ten year. Finally, He showed some work about image prediction and video prediction with adversarial training.
Carnegie Mellon University