Out of idle curiosity yesterday I spent a bit of time asking the internet the question “What do I need to know to start exploring the world of Artificial Intelligence?” The answers that I received from those nice Mr Brin & Page’s marvellous machine were illuminating.
There was a lot about mathematics. Statistics in particular. There was also quite a bit about programming. And a great deal of things about the management and manipulation of data.
But there were some great, gaping holes. Nothing about psychology. Nothing about sociology. Nothing about ethics. Nothing about aesthetics. Nothing about linguistics. The view of “intelligence” that is being created in the field of contemporary AI is very, very narrow. It’s intelligence viewed through a STEM filter. And that’s an issue.
It’s an issue because intelligence is so much more than logical reasoning and maths. And if we allow AI to be dominated by logical reasoning and maths, then logical reasoning and maths will be used to describe things that aren’t logical or reasoning. Things like art, beauty, emotions, music…
And whilst logic and reasoning can be used to create art, that’s only a very limited palate.
Up until the rise of data-driven synthetic intelligence in the past decade, AI focused much more on the idea of creating cognition in machines. Today big data techniques based on correlation over understanding causality and meaning have short-cutted to working solutions in fields like machine translation or voice recognition that previously seemed nearly impossible to solve.
These hacks, though, can be remarkably fragile. Whether the failure of Google Flu Trends, or strange patterns that can totally bamboozle, there’s no real intelligence. And until we start to again broaden out the discussion about what construes intelligence I fear the algorithms won’t become any more robust.