Rise Of The Machines
This week I read about graduates spending thousands of pounds on training so that they can beat emotion scanning robots used by City firms to filter candidates. Twenty minute interviews are conducted via a webcam. Candidates are given a range of questions, including brain teasers, and are then scanned for eye movements, breathing patterns and nervous ticks. Every time I get in the car in the morning my phone tells me about how long it will take me to get work. I have not asked it to do this. Whoever programmed my phone decided that I needed to know. Artificial Intelligence (AI) is becoming increasingly ubiquitous in our society. It is in internet search engines, online adverts, games, sensors, and self-driving cars. And it is helping scientists to improve their understanding of how the brain works, which is in turn feeding back into the development of even better AI.
The rate of development in this field is rapid. Chess playing computer programmes have been around for many years. In February 1996, IBM’s Deep Blue computer beat Garry Kasparov at chess. This was the first time a machine had won a game against the reigning world chess champion under tournament rules. Its winning strategy was to perform a massive number of simultaneous calculations. This does not imply intelligence of course. Deep Blue’s success was built on brute computing power. But in December 2017 a programme called AlphaZero, developed by a company owned by Google’s parent, Alphabet, taught itself to play chess in four hours. You might not think that that is particularly remarkable. But it was only given the rules of the game; it wasn’t provided with any sample games or human chess playing strategies. It taught itself by playing against itself. After learning how to play chess it then went on to beat the current champion chess playing computer programme, Stockfish. In a timed tournament AlphaZero won 28 games and there were 72 draws. Stockfish didn’t win a single game.
Based on the idea that as AI develops, it can then help to build even better AI, it is predicted that eventually we will get to the point where AI no longer needs human assistance. This point is known as the Technological Singularity. The current scientific consensus is that the Technological Singularity will happen somewhere between 2040 and 2050. What happens after that no one can predict.
NLP grow up in the early 1970s in California, when the computer industry was beginning to find its feet. NLP started by seeking out and then modelling excellent behaviour in humans. It maps the mind rather than the neurological structures that under pin it. On the other hand neuro-scientists are interested in mapping and understanding brain structures. So far AI has done very well performing massive numbers of complex processes, and will continue to do so. But will it ever have a mind to accompany its processing ability? Will it ever be conscious? And if it ever is, how will we know? We naturally impute consciousness onto other people and animals. We do this with people because they seem to be like us, and after all we are conscious so everyone else must be. We impute consciousness on animals too, because they seem to do things that imply thinking and awareness. We can also put them through cognitive tests. But how will we know that AI is self-aware, and if it ever is what will the world be like then?
Ever since we have been dabbling in mechanical and computer based ‘intelligence’ we have been worrying about a time when humans are no longer the smartest things on planet Earth. Through our books and films we have been playing this fear out and exploring it. But the time is fast approaching when we will need to start dealing with this issue in our reality, instead of in our imagination. It is high time for the humans on this planet to start using our own intelligence, because it wont be too long before we have some really smart company…
“Come with me if you want to live…” Kyle Reese, in Terminator.