The cheat code for the 1988 video game Contra (by Konami) on the Nintendo Entertainment Systems was the first video game cheat code I ever knew. There were books and magazines that would tell how to get high scores on most video games, or where the hidden coins, mushrooms, etc. were located -those maps or sequence of movements didn’t have the same satisfaction of knowing this hidden code. OK, it was terribly hidden and was known by most 10-16 year-old males at the time, but it still had the satisfaction of feeling like a secret hack.
Sitting in front of Contra and playing over and over again could get extremely boring…but become an ‘expert’ at it with a simple code and the game was like magic. Cheat codes and hidden maps; It was the steroid era for video games, and soon all games had them. Today people have realized that they can monetize these ‘cheat codes’. Want to grow your farm, city, castle faster? Pay for a pile of coins or diamonds. Want to buy those better weapons more quickly? Pay for in-game, fake money…with real money! These paid advances have taken away the novelty of perhaps knowing something other players don't, but plays on the emotion of wanting to succeed and the economics of the value of time.
Fast forward almost 30-years to 2015…Google’s DeepMind AI is able to learn how to play the Atari game Break Out. It essentially was introduced to the game without any instruction. It watched its own gameplay, came up with a ‘reward’ system so it knew when it was improving and learned to win all by itself - faster than and greater than any human competitor - all of this from beginning to end in 2-hours. Forget cheat codes. This machine didn’t know 2-hours ago it was sitting in front of a game. It learned how to play to expert levels because it never gets bored . The computer can play and learn so fast that the technique of frame-skipping was employed to allow it to play more games more quickly. This brute force attack was able to solve and master the problem of a video game in virtually no time - it is no wonder why computers are not allowed to be used in casinos.
Games can be hacked by computers to allow them to compute outcomes based on rules and probabilities of outcomes. Experts can be brought in to program situational awareness - in the case of IBM’s DeepBlue, chess masters were part of the programming team. In the case of Google’s AlphaGo, a champion was brought in to play hundreds of games against the machine to allow its programmers to study its faults. Early this year, the AlphaGo computer beat the worlds best Go player in a best of 7 game match, a feat that has been expected - but years ahead of the most optimistic opinions. The most shocking part of the match is well documented almost by as many AI researchers as Go masters. A single move, known simply as "Move 37" showed someting that researchers are still trying to explain (Google it)...but in summary - The most shocking thing about what the AlphaGo was able to accomplish in that move was not that it was made because of some huge brute force study of game play…the move appeared to be an understanding of the game in a way that humans couldn’t. There was beauty to the move that still astounds people, and debate continues that the computer understands the nuance of the game.
There are still many things that humans will have to teach machines and game play is a great way for us to do it. Far removed from cheat codes and brute force attacks, new AI is beginning to show true understanding. People will conintue to provide input into situations where brute force number crunching is not always the answer - programming intuition may be the next hurdle. There are great things that can be done once we are able to transfer our knowledge about our processes, buildings and transportation. What would happen if computer programs were allowed to use the same type of human programmed AI to make them more efficient? Maybe what could happen is the better question. Shouldn’t we expect hyper efficient factories, buildings and homes and the end of traffic and car accidents. It’s taken us 30 years to get from hidden cheat codes to an AI that can outthink human players at a board game. We should embrace the possibilities this could bring.