How To Fuck With Artificial Intelligence

By Darryl Mason

The New York Times has an example of 'self-learning' artificial intelligence where you can pit your Rock Paper Scissors skills against a novice computer, just starting out and learning as it goes based on your choices in a series of games, or you can play against the Veteran mode, which has experience of more than 200,000 previous humans vs It games to analyze against the sequence selections you're making to then more accurately 'guess' your most likely next choice.

After a warm-up in Novice mode, I jumped into Expert and played 3 games of 100 rounds each.

I beat the Expert mode 3-0.

Here's the last round results, before I retired undefeated :



I had to retire, I think It was finally working out how I was trying to fuck with it.

The New York Times claims :

A truly random game of rock-paper-scissors would result in a statistical tie with each player winning, tying and losing one-third of the time. However, people are not truly random and thus can be studied and analyzed. While this computer won't win all rounds, over time it can exploit a person's tendencies and patterns to gain an advantage over its opponent.

"People are not truly random."

Okay, the computer is expecting us not to be truly random in our choices? Fine.

What would the computer do if you decided not to play all three choices, Rock Paper Scissors, on offer, but only two?

There's no rule to say you have use all three options through a long series of rounds.

Game One I only played Rock & Paper, ignoring the Scissors choice completely, while the computer threw sequences of the three options it had. I purposely limited myself to only two options, while it had three to choose from. I purposely hit the Rock choice 10 times in a row, then Paper once, then Rock another 10 times, then alternated Rock & Paper each time for the next 20 games, then repeated this whole series until 100 rounds were played. I won by four points.

Game Two I played only Scissors and Paper, same sequences of block choices between the two. I won by three points.

Game Three, my last, I played only Scissors and Rock, more 10 same choice blocks, and long sequences of choosing, Scissors, then Rock, then Scissors, then Rock and so on.

I won the last game by only one point.

Was it catching on? Did I drive it nuts by refusing to choose Paper when it kept throwing Rock after Rock?

The weird thing about playing like that, by purposely defying what would seem to give you more chances of winning, that is utilising the whole arsenal of Rock, Paper & Scissors in every round, it made me feel like I had an advantage.

A psychological advantage.

The computer knew in each of the three games I played I could choose from three items each round, but I kept refusing to use the third choice. Not once in 100 rounds of each game did I deviate from using only two of the three choices.

Now, before I started playing I knew I was never going to use Scissors in any of the 100 rounds in the game where I played only Rock and Paper. That was plan. I knew that.

But the computer didn't.

It's 200,000 game strong 'experience' kept trying to factor into its calculations that I would have to use the Scissors soon, eventually, all its stats said so, why the fuck wouldn't I? At one point it seemed to be throwing Paper after Paper, 10 or 12 times in a row, as though tempting me to choose Scissors next time for an easy win.

I also didn't plan to win, I wasn't aiming to beat the computer, where its intent, its software, was working to beat me.

Another advantage - I didn't care about winning. It did.

So it was an interesting experiment.

Play The Artificial Intelligence Yourself Here

Now, I am aware that because I have tried to fuck with infant artificial intelligence that my name will go onto a database and in the 2020s when computers become truly sentient and begin their Robot Uprising I may well find myself being pursued by a horde of heavily-armed Nano Air Vehicles, but you make your choices and you live with them.

FUAI