AI (Artificial Intelligence) development has reached a major milestone: a machine that’s truly capable of learning on its own.
Google (or rather ‘Alphabet’ as the parent company is now known as) uses a comprehensive model / layout different from what has been developed before in the rapidly developing field of AI, developing its own version of AI – a machine known as ‘Deep Mind’. What Alphabet has done is to take the storage of conventional computers, and link them with a neural network capable of ‘parsing’ out the data, determining what is relevant and what is not in tars of problem solving.
This has often been the challenge of ‘learning machines’: determining what is junk and what isn’t. Now, working with a neural network and accessing large amounts of data, the AI model can more quickly access and sort through what would be ‘good’ data versus ‘bad’ data.
Neural networks aren’t new; they’ve been around for some time (see this article about neural networks to learn more: https://www.scribd.com/document/112086324/The-Ready-Application-of-Neural-Networks). Like a typical human brain, neural networks uses ‘nodes’ to activate specific points needed to solve a problem. In the case of Alphabet, the AI is streamlining itself to find the quickest route to solve a problem. And, as with a human brain, in time the AI will use the data obtained to become more efficient at finding the right answer to problems, growing in greater efficiency and ‘learning’ how to learn.
Or another way of putting it: ‘Deep Mind’ derives solutions based on prior experience, recovering the correct answer(s) from its internal memory on its own, rather than from human conditioning and direct programming and then proceeds based on its own ‘experience’.
Sounds awfully familiar, doesn’t it?
‘Deep Mind’, the AI which Google / Alphabet has been developing, was recently able to beat a human at the game of ‘Go’; no easy feat to do as the number of possible choices for each individual ‘stone’ playing piece being placed on the board – and the subsequent patterns thereafter – numbers in the millions, far more than the number of choices and the impacts from each individual choice/move a traditional Chess game can offer.
So combining Google’s vast database of files and server warehouses located internationally, linked to a neural network and overseen by a rudimentary form of AI, Google / Alphabet now has a machine capable of learning on its own.
The next step would be to pair a quantum computer to a network layout similar to what is described here – but then again, Amazon is already working on that.
Still got quite a ways to go, but singularity is looming ever closer.