See If You Can Outsmart A ‘Mind-Reading Machine’ From 1953 (We Failed)

Forget being smarter than a fifth-grader. These days, the real flex is if you’re up to par with artificial intelligence.

To start things off on a very arguably easy note, here’s an interesting test that pits you against a 70-year-old “mind-reaching machine.” Can you trick an algorithm so it can’t guess your next move?

The simulator is based on a system outlined in a 1953 memo by American mathematician Claude Shannon (via Boing Boing). In most experiments, it involved a human seated at a table with a lever, which they had to push left or right. While the participant decided, an accompanying machine would secretly predict the human’s move. 

Each time the algorithm made an incorrect projection, the human participant scored a point. If it guessed correctly, it earned a point. 

The concept has been turned into an online game called The Mind Reader Game by MIT researchers Guy Satat and Ramesh Raskar.

this was very stressful for some reasonhttps://t.co/XCdDn4OCiK pic.twitter.com/0qfMo1aXBA

— Maxim Leyzerovich (@round) January 14, 2021

The test is pretty hard to beat. Shannon, often regarded as the “father of information theory,” explained in his paper that this is because the machine looks out for patterns in the human player’s behavior and remembers them to jump a few steps ahead of its opponent.

The “mind-reading” algorithm makes a randomized guess at the start of the round. Once it has a grasp of your behavior, it begins to strategize, which makes the game harder as it progresses. It’s a primitive version of machine learning, if you will.

During our try, we scored a close 22 points against the machine’s 25 points. Give the algorithm a go here.

Then, send this link to a psychic to see if they can pre-predict the robot’s moves.

[via Boing Boing and MIT, images via various sources]