A few weeks ago I wrote about AlphaGo’s victory over Lee Sedol in a highly-profiled Go challenge match of man vs. machine.
As you may have heard, machine won.
One of the things that struck me was not that a computer could defeat the best Go player in the world, but rather, how suddenly we learned about the technology. My brother is a serious Go player (but certainly not professional quality). Less than two years ago, he honestly estimated that he was about as good as the best Go-playing computers.
Then, out of nowhere, AlphaGo emerged and was able to defeat the very best human Go player. No one had any idea the technology or anything comparable existed. And then it did.
This technology was probably under development for a long time. But we didn’t know about it until it was ready to compete against the best. Its developers did not go public until they knew, to a high degree of probability, that the technology was ready to impress.
I think there’s a lesson to draw from this.
Developments in artificial intelligence will appear to arrive very suddenly. But this will only partially reflect reality. The technology itself will progress incrementally, but the public will only learn about it “when it’s ready.”
To use a tired metaphor, the AI we know about is only the tip of the iceberg. Floating beneath the surface, 90% of the technology under development will pass unnoticed to the public until its creators have a financial or status-related motive to reveal it.
As the sophistication of AI increases, and we progress toward strong AI, the firms creating the technology will have even stronger incentives not to reveal the details of the technology to the pubic, as governments and other firms will have a powerful interest in stealing, nationalizing, or otherwise co-opting the technology.
And so we should probably expect more “sudden” developments in artificial intelligence. But we should also be aware that floating beneath the surface, there’s a lot more happening than we realize.