Artificial intelligence and the future
ARTIFICIAL intelligence (AI) simply means intelligent behaviour exhibited by machines. Examples range from a thermostat deciding to turn off the air conditioning because the room is now cool enough, to a complex set of algorithms that can autonomously drive a car.
The idea of building intelligent machines has been around for millennia. One of the big breakthroughs in the 20th century was Alan Turing’s mathematical formulation of the notion of computation, and the subsequent development of digital computers. However, some of the earliest work in AI was inspired by neuroscience, and how biological brains compute. That inspiration continues to this day. Interest in AI has exploded in the last few years due to a combination of algorithms originally inspired by neuroscience, new developments in computer hardware which allow those algorithms to run very quickly, and the vast amounts of data now available from the Internet. The power of these algorithms comes from learning. From around the 1950s to the 1980s scientists thought they could create AI by writing down formal rules for behaviour, such as what combination of features you expect to see in a photograph to recognise it as a dog instead of a cat. They didn’t get very far. Instead, AI in the 21st century relies on learning statistical rules from seeing lots of examples: for instance photos of dogs and cats, or internet search histories. One of the most important algorithms is called “deep learning”, which is based on artificial neural networks. Another is “reinforcement learning”, which grew partly out of studying how animals learn to make good decisions, for instance which way to turn in a maze to get the cheese.
Together these algorithms have produced astonishingly good performance on problems which people used to think only human intelligence could crack. For instance, Google DeepMind’s program AlphaGo used a combination of these algorithms to beat the world champion at Go, a strategy board game. There are still many areas where AI lags far behind human abilities - for instance the ability to understand language (as anyone who has used Google Translate will know). However it is improving rapidly. Experts don’t agree on whether AI represents a threat or an opportunity for the future of mankind. But either way, it will certainly have a major impact on our lives over the next few years.