Google’s DeepMind Artificial Intelligence
Greetings. The world is changing and we have a new brand of machine friends. For those of us who have been reading up on Tech news in the last few years, it is a no-news that the ace innovator, Google had acquired DeepMind Technologies Limited in 2014.
A slight brush up of the memory- DeepMind Technologies Limited is a British artificial intelligence company. Founded in 2010 by the trio Demis Hassabis, Shane Legg and Mustafa Suleyman, it left everyone gasping in disbelief as it successfully created a Neural Network and a Neural Turing Machine. They brought into creation the computer that learns and remembers like the human brain. In 2016, their program beat a human professional at the Chinese abstract strategy game, GO – a first in the World. Go is way more difficult for computers to learn as the possibilities of moves are many more than other games such as Chess. And other computers had been playing it only at amateur level and not with professionals.
And for the mystery mongers out there, DeepMind post Google’s acquisition not only has some key venture capitalists as investors but also a mysterious ethics board for AI research – embership of which till date remains unrevealed.
The goal of DeepMind Technologies is stuff that science fiction is made of. They put it simply as to “solve intelligence”. They bring together the very best in terms of technique from the two very complex streams of machine learning and systems neuroscience to put together general-purpose learning algorithms. Theirs is truly a deep mind game because they are majorly focused also on understanding the human mind.
What is phenomenally interesting is the sheer range of understanding that DeepMind has thrown up. Witness to that is Google Research which has released a paper on safety in Artificial Intelligence. Clearly, we are dealing with unfathomably exciting but a completely not envisaged future! And all of this through games. Because indeed, DeepMind centers its research on computer systems that can be taught to play games and focuses on developing it further. And according to its founder, Shane Legg, machine intelligence can achieve the levels of human intelligence.
But what makes Google’s DeepMind a class apart is that unlike other AIs, like IBM’s Deep Blue for instance, is that it is not purpose driven, pre-defined and pre-programmed. It “learns” from what it goes through – like the human experience – this is machine experience. It uses what in learning theory is referred to as deep learning and reinforcement learning. Using pixels for data, it does not restrict itself to just a game.
And, this is evident by its history of collaborations. DeepMind has been collaborating with Hospitals for the analysis of anonymised eye scans in which it would search for early signs of diseases that may lead to blindness and to develop algorithms to differentiate between healthy and cancerous tissues, automatically.
And if you thought it would still be “just a computer”, you might be proven wrong soon. What has broken the Tech news in recent times is DeepMind’s WaveNet.
This is a voice very close to being human. And a tip-off – head off to Deep Mind’s WaveNet page to hear a few of the samples. Statutory Warning: You may get really creeped out. But the possibilities of DeepMind are endless.