How does AI work in practice?

Although it’s not a new idea (see A Brief History of AI, https://t.co/A7c53luxwL), it IS a different form of programming. To see how it contrasts with traditional programming we should perhaps start by asking ‘what is intelligence’? According to one dictionary, it is “the ability to acquire and apply knowledge and skills”. Humans are already wired to do this, but machines have to be shown the way.

This means that instead of telling systems exactly what to do, we programme them to learn from the data we give them and let them figure out how to do it. The arrival of big data has also given us enough data points to create effective AI models. Here are a few ways we’re seeing AI work.

For example; in online games, it’s all about managing large permutations. Chess was an early example of technology beating the human brain. Deep Blue became the first computer to defeat a reigning world champion (Gary Kasparov) in a match during 1996. AI is needed for managing the move permutations in chess, which start going up exponentially very quickly.

But of course, the Chinese territorial game GO was the game-changer for AI. The options are equivalent to the theoretical number of particles in the universe. So in order to beat humans at GO, the system has to know how to learn and strategise. The AI then invented a brand new move, dubbed ‘move 37’. This strategy had never been seen before in a game over 2,500 years old. It is often used as an example of a potential sign of creativity or original thinking by AI.

In image processing we’re seeing some brilliant cases of AI’s ability to spot patterns. Trading systems and economic forecasting aside, AI is being used as a second opinion in cancer detection. The AI system is fed the same scanner images that human Dr's use to diagnose cancer however it can occasionally detect instances that the human clinician missed.

Traffic management is another good example of where AI is useful in filtering mass amounts of data to allow humans to make strategic decisions and not be distracted by 'noise'. If we end up in a world where the sky is full of autonomous drones AI can be used to filter out the ones that are 'behaving well' and highlight rogue ones. This would allow a human operator to then intervene and take immediate action on the drones that need attention.

Lastly, AI isn't just Machine Learning. Genetic Algorithms can optimise the design of electricity-generating windmills. If we can say Machine Learning is inspired by how a human brain works, Genetic Algorithms are the artificial form of Darwin's Natural Selection. In the case of wind turbines, data about height, sail length etc can be given attributes and values to generate random examples. An algorithm or simulation can then score hundreds of random samples. The attributes of the best scoring ones can then be mixed to create a second generation of the same size. This new generation can be scored in the same way, and so on perhaps for hundreds of generations until we get to a suitable solution.

AI is a new approach to programming with practical benefits. By combining it with large data sets and human ingenuity, we can turn AI into a serious force for good. AI is not here to replace us but to work with us and the best systems always have a human in the loop.

If you enjoyed this article please subscribe to CKX as we will be releasing similar content aimed at demystifying AI.