Jun 29, 2017 11:51 AM EDT
The use of artificial intelligence is becoming more prevalent today from home systems to scientific research. At present, artificial intelligence learns through supervised training but the next generation AI can develop its own "human-like" intuition.
The new wave of artificial intelligence will be able to understand and learn the world around them like a baby. Then as it learns more about the world around it, so does its understanding of it. This is what Yann LeCun, director of AI research at Facebook, said.
He explained that at present, there are two types of machine learning - supervised learning and reinforcement learning. The first one is the most dominant method while the second one exist mostly in games.
However, there is a third type of machine learning that is emerging - predictive learning. This type of learning is unsupervised, not needing to be trained with millions of "human-labeled" examples.
He used the baby metaphor to explain this. He said that when a baby is shown an image and another image moves in front of it, the other one is still there. But if an object is not supported, it falls. He said that no one teaches these things to the baby but rather he learns by observation.
Predictive learning is the next goal for artificial intelligence. Scientists don't know how to create these types of machines yet but research has been made how to go that next level.
When predictive learning becomes possible, it will save a lot of time and money curating huge databases of labeled data. With predictive learning, you just show the AI thousands of hours of video and it will understand how the world works and is organized.
Geoffrey Hinton, an AI pioneer and an engineering fellow at Google, said that the next generation AI will have "common sense" just like humans. One day, they will be big and sophisticated enough to hold that amount of real-world knowledge.
© 2017 University Herald, All rights reserved. Do not reproduce without permission.