Some of you may have been following the news about the ongoing matchup between Google's AlphaGo AI and the world champion Go player Lee Sedol (ranked 4th in the world). As you may have read, AlphaGo has beaten Lee Sedol in 3 out of 4 matches played so far, with the result of the 5th coming very soon.

(Update: AlphaGo won the final match up as well, again in a style quite unlike human players. 4-1 to the machines.)

Regardless of the final result, the fact that AlphaGo has beaten Sedol at all, let alone handily dominated him is hugely important. Much of the news has been about the more sensational aspects of the match up, but we want to talk a little about the longer term implications of the results.

The conventional wisdom till recently was that A.I.s were at least a decade away from mastering Go, because unlike Chess where the number of possible moves is relatively restricted, Go has more possible positions than there are atoms in the observable universe. So its just not possible (at the moment at any rate!) to win through computational brute force, i.e. iterating through every possible outcome till you find the best one. Instead, AlphaGo had to actually learn how to play Go, and develop its own version of intuition as to what its next best move is, as well as anticipate what the next move of its human opponent is!

As demonstrated by the match ups so far, this has proven to be a highly effective approach, leading to beautiful plays that aren't quite human, but infinitely effective and graceful nonetheless.

An interesting question is then, where does this leave humanity in the years to come? Its quite obvious that the progression of technology is getting faster and faster, in many ways at an exponential rate. It took us about 2.5 million years to go from ancient apelike creatures wielding crude stone tools to devising the printing press in the early 1400's. It took another 400 years or so to come up with the lightbulb and telephone. Then during the next 100 years after that, from the 1850s to 1950s man went to the moon and invented the first computers. And in the subsequent 70 years or so mankind has developed cell phones, DVDs, ultra portable computers, the internet, wifi, harnessed energy, 3D printing, decoded the genome, peered deep into the cosmos and made numerous breakthroughs in virtually every possible field.

Each development paves the way to future developments, and every tool we build makes it easier to create the next one. Similarly, the development of AI is reaching a tipping point, where AI can now train itself and essentially improve itself through neural networks that learn in a similar fashion to humans. Given the AlphaGo result, we can see how effective this approach is. They can already drive cars and play Go better than us. Pretty soon we will have AI's that are as smart as humans.

But that is just the beginning. Because they have the ability to evolve much faster than we do. So its a good bet that within moments of first reaching human like levels of intelligence, they'll evolve again to become smarter than humans. And not too long after that they'll push through to a level of intelligence that's unimaginable to humans today. And by then we'll have put half of them in bodies like this. (And possibly pushed them around long enough for them to start resenting us).

And all of this could happen relatively quickly, which is why a lot of leading scientists and thinkers are getting quite worried about AI.

If you were an AI and looked with cold hard logic at this world, would you still think that humans are necessary? You might see human's as a threat to your own existence. Or even if they weren't worried about themselves, a more enlightened AI might simply weigh out the consequences of human existence on the planet. Looking at the planet as a whole, at the well being of all the species on earth, its hard to argue the point that almost every species on earth would be better off without humans around (with the possible exception of domesticated dogs and hamsters. Not too sure about cats since they seem pretty independent).

So would an AI think its time to go Skynet and save planet earth? For the greater good? Or perhaps they'd just figure out interstellar travel and leave us here to live in our own mess. Interesting things to ponder as we inevitably head into this new era of ultra smart AI's.

 

Picture credit of Go Board. By The original uploader was Fcb981 at English Wikipedia (Transferred from en.wikipedia to Commons.) [GFDL (http://www.gnu.org/copyleft/fdl.html) or CC-BY-SA-3.0 (http://creativecommons.org/licenses/by-sa/3.0/)], via Wikimedia Commons