By Patrick Lambe
From Kevin Kelly back in October, a nice nuanced piece on the three converging factors "that have finally unleashed AI on the world". The three factors are: (1) cheaper parallel computation; (2) big data, that gives algorithms thousands or millions of examples to see patterns in and learn from; (3) more efficient architectures for networking processors in support of "deep" (and fast) learning. It's (2) in particular where knowledge organisation methodologies contribute. Big data can only see meaningful patterns across many diverse data sources, if concepts expressed through multiple vocabularies and in diverse structures, can be resolved to the same things. Kelly also points out the dependency on very large networks of users. Artificial intelligence learns effectively only on large scale, not small scale (which is very different from how humans learn, by the way). All the more, the problem of diversity of description and structure needs to be resolved.
RSS Feed