An important feature of a learning machine is that its teacher will often be very largely ignorant of quite what is going on inside, although he may still be able to some extent to predict his pupil’s behavior… This is in clear contrast with normal procedure when using a machine to do computations [as] one’s object is then to have a clear mental picture of the state of the machine at each moment in the computation. This object can only be achieved with a struggle. The view that “the machine can only do what we know how to order it to do,”‘ appears strange in face of this. – Alan Turing
The above section of Alan Turing’s text “Computing Machinery and Intelligence” caught my attention in relation to the encounter back in March between Google’s AlphaGo system, and Lee Sedol (one of the top Go players in the world today). Widely regarded as the most challenging game on the planet, Go was due to it’s great complexity and astronomically vast move-tree thought un-suited for AI’s to tackle, as the usual “brute-force” (if/then) approach common to chess engines (the Deep Blue chess engine of IBM notoriously toppled world chess champion Garry Kasparov in 1998) is simply too ineffective to handle Go at the level at which top professionals play the game. However, AlphaGo is not a system based on brute-force analysis. It has achieved it’s level of play through machine learning, and deep neural networks. After an initial period of being given feedback by human supervisors as to the validity of move suggestions, AlphaGo was eventually let loose on databases containing millions of professional games—teaching itself through evaluations and comparisons the way human professionals play the game of Go. What was astounding, however (and a great surprise to a great many, yours truly included), was that not only did AlphaGo display a level of play on par with human professionals, it several times throughout the match came up with what commentators and other Go professionals alike could only describe as “creative moves”, moves that stumped not only Lee Sedol, but even AlphaGo’s own creators. Moves and sequences invented (yes, why not?) by AlphaGo during those five games have since been put to further use by, among others, Sedol himself.
For me, one of the most fascinating aspects of AI is the way in which it can force humanity to rethink what concepts such as “creativity” (or even “intelligence” for that matter) really implies. Be it in the field of Arts or abstract strategy games, further development of AI will hopefully see humanity humbled, renegotiating it’s place among other intelligences, be they electronic or organic in origin.