AlphaGo and the A.I. Apocalypse

terminator

AlphaGo, circa 2050

“After humanity spent thousands of years improving our tactics, computers tell us that humans are completely wrong, (and) I would go as far as to say that not a single human has touched the edge of truth of Go.” – Ke Jei

The above quote is from a man many consider the world’s best Go player. It was stated shortly after AlphaGo, an artificial intelligence (AI), thrashed him as well as many more of the world’s best players to win 60 straight matches, an unprecedented feat. Some marveled at the accomplishment, the creation of an artificial intelligence capable of defeating the best human players in the world at a game that not only requires skill but also tremendous creativity to master. It was painted not so much as a victory for machines but for Humanity, who created a synthetic limited cognizance that showed the first signs of being able to supersede its creators not only in technique but in “imagination” as well.

While taking a walk, something profound occurred to me about this development. There are some who think that we should not be particularly concerned with the prospect of AI achieving intellectual and creative superiority over Humanity; but there was something unsettling about the situation regarding AlphaGo as I pondered it in more detail. Consider that a computer has been created that has shown clear superiority over humans in a game that many thought was beyond a machine’s inherent capability to master. To me, the fact that a machine could establish mastery over humans in this fashion makes it clear that machines likely can gain mastery over humans in every fashion, at least when it comes to cognition.

While it remains to be seen if that’s the case, what really struck me was how AlphaGo was able to do it. This isn’t so much about the “machine learning” used by it but something more basic:

AlphaGo has perfect recall.

Yes, it learns but it also remembers more or less perfectly. In other words, AlphaGo is able to access its accumulated experiences, it’s history, in a perfect way. By being able to draw from all of its experiences in totality, it was able to formulate near perfect strategy to throttle the world’s best at a notoriously difficult game.

In my post “Efficient Markets” Eliminate Risks, I wrote about prescience and how perfect knowledge of the future would inform subsequent actions to either reap gains or avoid calamity. To put it another way, complete knowledge of the future would allow one to formulate a perfect strategy to deal with it.

Consider the prospect that not only can complete knowledge of the future eliminate risk and create certainty of action, but complete knowledge of the past likely can as well. If humans were able to successfully map every action of everyone everywhere in history with a machine, would that machine be able to successfully “predict” what we would do next?

AlphaGo was not just reacting to its opponents, it was using its history to shape its future. It was using perfect knowledge of its past to inform its actions to create a future outcome. Considering the fallibility of the human mind in accurately retaining past accumulated information, is it any wonder that we are prone to repeating the same errors of judgment? Maybe all human error is simply the result of our general inability to perfectly and efficiently access our complete past experiences.

Another thing to consider is that Go is a game with specific rules in which a specific outcome is sought. What are those rules and what is the outcome we seek as a species? Even if we perfectly mapped the past, using it to inform current or future actions would be meaningless if we didn’t know what outcome we were seeking. So is it possible that it is not only important, but essential that Humanity agrees on an overall purpose?

What I think I’ve come to understand is that, in order for Humanity to survive, it should strive for two things:

  1. As complete and unbiased an accounting of our history as possible and;
  2. A clear and continuous linear set of objectives, with each subsequent objective determined only when a previous objective has been met.

In theory, if Humanity were able to record all of its history perfectly, every event, in an AI, that AI likely would be able make very high-value recommendations regarding what our future actions should be.

What would a machine with perfect knowledge of our past be able to tell us about our future? Most likely that “not a single human has touched the edge of [T]ruth.”

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s