AI will be writing your business strategy sooner than you think
Today marks an important milestone for AI. Remember where you were when you read this.
A computer taught itself to play chess with no prior knowledge of the game other than the rules. It then beat the current world champion computer chess program.
There is a story in the Guardian here.
I remember when a computer program first beat Gary Kasparov. Back then the program was tuned by chess and AI gurus and used brute force searches to evaluate millions of moves. Yes, it won – but the match up seemed unfair on the human.
This time the computer knew only the rules of the game. It taught itself to play by playing lots of games against different versions of itself. It uses a neural network, which is analogous to the human brain. It has lots of very simple processing elements (like neurons) connected in a complex network. Like the human brain, neural nets are good at pattern recognition. This one evaluates far fewer moves than the brute force approach because it learned to recognise the most promising possible moves and evaluate those in greater depth.
This is really impressive, because it is a general purpose AI that doesn’t need human knowledge about a problem. At the moment it is limited to very well structured problems – games. But it is an important step on they way to genuine thinking (and learning) machines.
This sort of generalised learning has enormous commercial (or social or military) potential if combined with systems dynamics and agent-based approaches (where complex real world systems, such as markets, are modelled as large numbers of autonomous interacting units representing people or companies, say). For example, a business could use this approach to develop the optimal pricing strategy, including responses to competitor actions.
I believe that, sooner than we expect, business strategy (like driving) will be considered too important and risky to be left to humans.
I am interested in talking to businesses about exploring these applications.