Artificial intelligence is one of the burgeoning industries that big data will provide with many helpful technological improvements. However, how much will increased computational power provide to an industry that has made very few ground breaking developments in its several decade history. We have yet to encounter a computer that can really learn, let alone has anything approaching consciousness.
Reading about the most recent developments and predictions emanating from Microsoft’s research labs in the NYT this week reminded me of an article I had read a few weeks ago warning about this exact sort of overzealous predictions. Though machine learning applications are slowly growing; there is hardly a revolution going on when the best improvement is a shift from an error rate of around 20% to around 12.5%. While this is certainly a helpful step in the right direction, it hardly seems to qualify as true artificial intelligence. The problem with most discussions of artificial intelligence seems to be the bounded nature by which the computers are able to interact with its environment. True intelligence requires a certain type of creativity and cleverness that either cannot be programmed or must be done in an incredibly abstract way. This is not a criticism of status quo programming but rather an acknowledgement of the difficulty of attempting to mimic that which we understand very little (the brain and more spefically the act of learning). Where do you think the necessary innovations will come from?