Turing Award winner Yann LeCun is proposing a new way of making AI which might revolutionize what it can do—if we can figure out how to make it. LeCun is trying to find a way to give AI "common sense", a tough task, given that it can't even get an intuitive grasp of the force of gravity. Simply crunching numbers won't work anymore; we need a radical new idea to make any meaningful progress. LeCun Proposes a network which focuses only on relevant information, paired with a configurator which determines which information is actually relevant. LeCun believes this will be the next step in AI, not large models or reinforcement learning. That said, even he doesn't know how to implement the model, or how to deal with the many questions it raises, only that it's a new and viable path to AI having general knowledge of the world.
Comments
Post a Comment