Data CentresLondon, UK, 17 Feb 2020
Take smartphones for example
A simple voice prompt of a virtual assistant will let us know what the weather will be like on a given day, and social media feeds provide news updates and advertisements that reflect personal preferences.
Nevertheless, it’s tempting to dismiss artificial intelligence as mere tech hype given the many definitions that exist of it. Doubts have also arisen about its potential given the images of sentient computers that come to mind when the words 'artificial intelligence' are uttered. Given the likes of HAL 9000 in the movie 2001: A Space Odyssey are still science fiction, this sentiment is understandable but misplaced.
The impact of artificial intelligence is sufficiently broad for infrastructure and operations professionals to take notice. A tremendous amount of compute is required to replace the human element with a machine that can exercise ‘judgement’ based on past behaviour as humans do.
In other words, artificial technologies require tremendous amounts of computing power for enterprises to realise benefits such as simplified customer interaction. Andrew Ng, chief scientist at Baidu’s Silicon Valley Lab, told Datacenter Dynamics recently that training a Baidu Chinese speech recognition model requires four terabytes of training data, in addition to 20 exaflops of compute, or 20 billion, billion math operations across the entire training cycle.
As a result, data centre design and implementation will have to accommodate artificial intelligence, however it’s defined in future, given the undeniable benefits. The cost savings and additional revenue, and more data-driven ways of working are efficiencies organisations, and customers alike will strive to achieve in the future.
For a discussion on the evolution of data centres, please contact us.