The opening post is short, but I agree with it pretty much completely, and there's a great discussion that follows that draws out many of the arguments and implications. If it's all new to you, skip down to comment 19, which draws the distinction between classical AI and machine learning AI. In a nutshell, Daniel Lemire is arguing that the new 'Linked Data' approach, which is an heir to the Semantic Web, is an heir to the now discredited 'classical AI' approach to machine intelligence. In the classical approach, you collect all the sentences that describe the world, organize them into subjects and (especially) predicates, and link them together. "Collecting, curating and interpreting billions of predicates is a fundamentally intractable problem. So our AI researchers failed to solve real problems, time and time again."
Today: 5 Total: 112 [Direct link] [Share]
