It’s official – the future belongs to Artificial Intelligence (AI).
That’s according to the world’s best player of probably the world’s most complex game. Ke Jie, 19, was speaking before his match against AlphaGo AI, in a three-match series of Go. The ancient (2,500+ years old) strategic board game involves two players, with the aim of encircling the other’s territory. Its complexity stems from the fact that it has more possible moves than there are atoms in the universe.
Google’s AI ended up beating the Chinese champion by just half a point (3-0 overall), the closest margin possible. At first glance this suggests a closely fought game. However, the nature of AlphaGo is that it does just enough to win, a (deep) mindset prompting Ke Jie to observe: ‘AlphaGo will always be a cold machine. Compared to human, I can’t feel its passion and love for Go. Well, its passion might only come from overheating with the CPU running too fast.’
Of course, computers beating humans is nothing new. It’s over 20 years since chess grandmaster Garry Kasparov lost a game to IBM’s Deep Blue. The chess-playing computer relied on a system of brute force – calculating every possible move before finding the correct one. In contrast, AlphaGo’s win was based on the Monte Carlo Tree Search. This is a method for making optimal decisions in artificial intelligence (AI) problems, typically move planning in combinatorial games.
AlphaGo takes things one stage further by introducing machine learning, to mimic the neural networks of the brain instead of relying on programming rules. What’s more, AlphaGo requires no knowledge about a game or situation to make reasonable decisions. Naturally, this offers exciting possibilities for solving some of humanity’s most pressing and intractable problems.
AIphaGo is the product of DeepMind, world leader in artificial intelligence research. Founded in 2010, the London-based company was acquired by Google in 2014. Its stated mission is ‘to push the boundaries of AI, developing programs that can learn to solve any complex problem without needing to be taught how’.
One example is DeepMind Health, where the technology is being applied to support medical professionals in their healthcare.
Use cases include analysing data from tests for faster diagnosing, and alerting doctors when a patient’s health deteriorates. It’s done via Streams, an app. ‘Streams is saving us a substantial amount of time every day’ explains a consultant nurse. ‘The instant alerts about some of our most vulnerable patients mean we can get the right care to the right patients much more quickly.’ Although there have been privacy concerns over how DeepMind Health stores and processes patient data. The app’s functionality has attracted the attention of the UK’s medicines and healthcare devices regulator.
Data centre savings
What impact does all this next-generation processing power have on data centres?
For Google, machine learning reduced its cooling bill by a staggering 40%. Its data centre team used historical data – such as temperatures, power, pump speeds, set points – to train a system of neural networks. The aim was to predict the data centre’s temperature and pressure by the hour, to ensure maximum Power Usage Effectiveness (PUE).
Google now plans to roll out the algorithm in other data centre environments. For example, improving power plant conversion efficiency, reducing usage of semiconductor manufacturing energy, and boosting water usage efficiency.
From strategy, to health, to energy efficiency. There are few use cases where innovations like AlphaGo can’t be applied. And as data volumes and insights grow, there are even fewer limits to its potential. ‘What’s exciting is that AlphaGo just keeps getting better,’ said Go commentator Hajin Lee. ‘It was already so good before.’