I have often written and spoken about “The Singularity.” It may sound like science fiction, but it is real. [Check out Kurzweil’s book at the right for a deep dive.] All serious students of AI of whom I am aware agree that at some point, computers will have more computing power than a human brain, or even all human brains. I’m not talking about consciousness, sentience, souls, or anything metaphysical. I’m talking about sheer power to process data. Almost everyone agrees that the power of computers will accelerate at an extreme rate as they teach themselves without human assistance (control?). This has given rise to warnings by Elon Musk and Stephen Hawking, among others. Some of these are clearly “clickbait,” but among the descriptions of The Singularity are:
– “runaway technological growth, resulting in unfathomable changes to human civilization,”
– “end of the human era, as the new superintelligence would continue to upgrade itself and would advance technologically at an incomprehensible rate”
– “by 2029, computers will have human-level intelligence.” And just a few years after that singularity humans will merge with machines.
Estimates as to when The Singularity will occur are all over the place, but the substantial majority range between 15 and 40 years from now. Most that I have seen lately are around 30 years (i.e., 2047).
I write all of that to provide perspective for this: The AI world is buzzing today because Google has made a significant stride toward The Singularity. Remember the recent breakthrough when AlphaGo beat the world’s best human Go player? (Go is vastly more complicated than chess.) Now a Google AI (Deepmind) has taught itself how to play Go and has beaten AlphaGo again and again (100 to 0).
Previous versions of AlphaGo built to compete against human masters of the game required hours and hours of training on Go gameplay, but AlphaGo Zero was able to teach itself to play in a matter of days using a technique called reinforcement learning. Details about this new tech, called “AlphaGo Zero” are here and here. Sample coverage from all over the world is here and here and here and here; and the Chinese are trying to catch up.
- In this post, Ken Grady covers a couple of the major hurdles that need to be overcome for AI to really challenge human intelligence. Those are, ability to “reason” and achievement of “general intelligence.” (Sheer processing power is not a problem, and metaphysics is not really relevant.) This is related to AI someday actually practicing law.
- I have recently been impressed by Apple’s inclusion of an AI-specific processor in the upcoming iPhone X. Recent press releases suggest that all major smart phone players are doing the same. And in related news, while some say Intel has been falling behind NVIDIA’s graphics chips in the AI race, this new class of chips may help Intel narrow the gap. “Intel Nervana Chip Aims To Revolutionize Artificial Intelligence.”
- Here’s an example of DLA Piper’s AI practice’s work.
- If you’re in the mood to read more about AI generally, this article has many excellent links.