If you're looking for the funniest stuff, I suggest starting with the Steve, Don't Eat It Homage and then the travel category. You're on your own with the older posts that have yet to be categorized.

Wednesday, December 13, 2017

AI has already taken over

There has been much made over the past year or so about the danger of a "rogue" AI destroying humanity or some such. The reality is this may have already occurred. Let me explain.
Not long ago, I wrote an AI to play the game Can't Stop. This is a simple dice rolling/press-your-luck game. The AI (it's actually a neural network but that is not important) looks at all the current information and decides what combination of dice to choose. Then it decides whether to roll again or stop (and end its turn). To create this AI, I start with a network, copy it and make some random changes to the copy. I then let these two networks play against each other many times. I then look at the results and if the original was clearly better than the copy, I keep it. Otherwise I replace the original with the copy. I make a copy of the winner, add some random changes and repeat. Eventually, the network learns the best way to play. Google is doing something similar with their Go playing AI.
And so is Facebook and Twitter and Amazon and Scott Adams. It most of these cases, the changes aren't random but the effect is the same. Eventually, to the extent that it is possible, they will consume all of your time and/or money. CNN, Fox News and MSNBC operate on much the same principle (as news has for a very long time). If you find something that doesn't operate like this, it is likely a failure (or on its way) or something that can ignore popularity/money/power (like a personal blog).