Another Breakthrough for AI
DeepMind's AlphaGo Reaches Milestone and Looks for New Challenges
FUTURE PROOF-BLOG BY FUTURES PLATFORM
One week ago, in May 2017, AI finally won the most complex strategy game known to humankind. After defeating the brilliant Lee Sedol with a 4-1 win last year, DeepMind’s AlphaGo, which is part of Alphabet (Google), continued to make improvements to its AI, culminating in a 3-0 win over the game's number 1 player and 19-year-old prodigy, Ke Jie. The question is: what is AlphaGo next move?
When in 1997 DeepBlue, IBM’s chess AI, beat the reigning world champion Garry Kasparov, many thought it was the dawn of a new era. And, indeed, it changed the game of chess forever. But while chess is a highly complex game, as late as last year experts believed that Go, a more intuitive game with more than 2.08*10^170 possible positions, or more possible positions than there are atoms in the observable universe, was still far from AI’s powerful reach for at least another decade.
But the “God of Go“, who learns by playing against itself, proved them wrong by beating Lee Sedol, Ke Jie, and winning 60 consecutive games online against Go masters (Ke Jie was one of them, prior to their official match). No clearer statement of supremacy could have been made.
So it is no wonder that DeepMind, the developer of AlphaGo that Google bought for around $500 million in 2014, is now looking for new challenges. It has the potential to bring new developments to a variety of industries, from healthcare to the finance world.
Already last year, Google’s Eric Schmidt said that there would probably not be a single area where Alphabet, Google’s parent, would not use DeepMind’s technology. From self-driving cars to home assistants to ads technology, DeepMind can bring profound developments in any area, sooner than anyone expects.
The company already has a partnership with UK’s National Health Service, where it can apply its algorithms to improve diagnoses or treatment plans. Although fundamentally different from winning games, making the right diagnoses or medical choices absent of mistakes is essential. It can even potentially help people live longer lives by successfully predicting a variety of health problems for a given individual and prescribing helpful courses of action.
Financial institutions can also use it to make use of all the available data to make good investment decisions. Coupled with a powerful search engine, AlphaGo could potentially use information in a way that no investment software currently does. Should DeepMind choose to make it open-source, it could level the playing field for millions of investors, changing the nature of the game.
While humans are still better at picking up patterns than AI software, many believe we are approaching a singularity. AlphaGo, Watson, and other AI programs certainly seem to be working hard for it. Just like experts believed that Go was still a decade away from being conquered by AI, many believe that our emotions and intuition, that which makes us human, are still far from being imitated by AI.
But this may not be so. The key takeaway is: the exponential growth in artificial intelligence has shown us that it may not be that long until technology behaves indistinguishably from human beings.
Back in 2014, Ray Kurzweil, famous futurist and "Googler," made a stark prediction: by 2029, we would have human-like computers. Now, only a decade away, we look at whether this prediction still makes sense.