‘The Antikythera Mechanism, as it became known, was discovered in the wreckage in 1902 but it’s only in the last 10 years or so that scientists have begun to understand what it was for and just how clever it was.’
Around 100 years before the birth of Christ, somewhere in the Mediterranean – let’s call it Rhodes – lived a man called Hipparchus. He was a mathematician and astronomer who liked a challenge, and one particular challenge was this: if you are going to train for the Olympics then it would be helpful to know when the next one is but how, in ancient Greece, could you figure that out precisely?
The Hipster made a little machine with lots of cogs and gears that would show not only when the next opening ceremony would be but also things like eclipses, the Moon’s orbit and so on. It must have taken him ages, what with only having some bronze and a pencil (and maybe not even a pencil).
When it was ready he popped it in the post. It got as far as the Island of Antikythera, where the boat carrying it sank. The device wasn’t seen again for 2,000 years and nobody would make anything even close to that complexity until the fourteenth century.
The Antikythera Mechanism, as it became known, was discovered in the wreckage in 1902 but it’s only in the last 10 years or so that scientists have begun to understand what it was for and just how clever it was. It’s likely that it was developed from earlier devices and nobody knows for sure who made it, though Hipparchus is a strong suspect. It’s considered the first analogue computer and I would say that it marks the birth of AI.
You might say: “How is that AI? It’s just a machine that does some calculations!” That’s true but that’s what artificial intelligence is. The “simulation of intelligent behaviour in computers,” as one definition has it, is usually just a machine carrying out calculations. It’s a digital computer, rather than an analogue one, and we call the steps for completing the calculations ‘algorithms’ but the principle is the same.
“But,” you might say, with a persistent but hypothetical objection that helps me make my point, “the machine isn’t thinking!” And you would be right because machines cannot think. They do not think. They probably never will.
We don’t even understand human thought and how it works or what it is. How on Earth would we be able to build a machine capable of doing it?
At ContactEngine we have been feeding data into one of the industry's best Language Understanding Services – a machine-learning service for understanding natural language. When people ask to reschedule appointments, there are dozens of different responses that they might send. A human will understand nearly all of them and reschedule appropriately but getting a machine to do it is difficult.
The Language Understanding Service gets it right about 25 per cent of the time. Why? Because it’s not intelligent. It’s not thinking. Making it better means ‘training’ it, which usually means a human classifying each possible instance until the service has more to go on. That’s human intelligence being automated, not computer intelligence being created.
The problem is the language. We are doing amazing things with computers today. They are capable of making calculations and processing data in ways that couldn’t be dreamt of a generation ago. Describing what they do as ‘artificial intelligence’ makes people assume they are doing something they are not. For a lot of people, the phrase ‘artificial intelligence’ means magic.
The Antikythera Mechanism probably looked like magic too. It wasn’t, though. It was science, maths, some hard work and a lot of human observations. All etched into the cogs of a wonderful machine.