Intelligent paper

The lesson is important because of the discussion about what artificial intelligence actually is – and whether AI is even possible.

It can also serve as an introduction to The Candy Computer, as it helps us better understand how the computer from candies learns how to play a game.

The activity was developed by Paul Curzon from Queen Mary University of London and is published in the CS Unplugged collection.

Discussion before the game

Tell the children that the piece of paper you are holding is smarter than everyone in the room. Lead a discussion towards the question of how this could be proven. Acknowledge that the paper’s intelligence is not general, but limited to never losing against a human at playing a specific game. While it cannot play chess, it is so good at tic-tac-toe that it has never been defeated.

More ideas on how to conduct the introductory discussion - which is important for understanding the message of this activity - can be found in the additional materials for teachers.

Game

1. Explain that the paper has no eyes or hands, so it cannot see or draw. It will need someone to do this on its behalf, but that person must strictly follow the paper’s instructions. Choose a student and give them the paper with the instructions.

2. Find a volunteer to compete against the paper. (Ideally, they shouldn’t be too skilled, or it might not be as interesting …)

3. Draw a tic-tac-toe grid on the board.

4. The student representing the paper should read the introductory text aloud (up to Move 1). It will soon become clear that the paper insists on making the first move. If students complain, explain that the paper must be smart if it demands to go first. Remind them that going second doesn’t mean they will lose — after all, they also play this game regularly, and no one gives up just because they aren’t the first to move.

5. The student with the paper reads aloud the instructions under Move 1 and follows them: place an X in a corner.

6. The student competing against the paper makes their move.

7. When it’s the paper’s turn, the student reads aloud what is written under Move 2 and follows the instruction. The teacher should carefully follow along to ensure the student executes the instructions correctly. For example, “opposite corner” means the diagonal opposite corner.

8. Continue the game. By the second or third move, some students may start to realize that the game is already lost. Point out that there is still no reason to surrender — perhaps the paper has just been lucky so far, or maybe it doesn’t realize it’s on the verge of winning and will completely mess up its next move.

9. The game will end either with a win for the paper or in a draw. If it’s a draw, point out that you never claimed the paper would always win — only that it would never lose. If the paper wins, announce that it has once again proven to be smarter than humans.

You can, of course, repeat the game as many times as you like.

Discussion after the game

Ask the students if they are now convinced that the paper is intelligent. At this point, they will usually all insist that it is not — despite the clear evidence of its abilities.

If someone argues that the paper didn’t play the game on its own but was helped by a human, clarify: no, the human provided only eyes and hands. The thinking was done by the paper.

If someone can play tic-tac-toe this well, they must be intelligent. If intelligence isn’t in the paper, then where is it? The students might suggest that the true intelligence belongs to the person who wrote the instructions on the paper.

Now you can explain that the paper is actually like a computer program: the student holding the paper was like a computer, blindly executing the instructions written in the program. Computers never do anything other than execute the instructions that someone has programmed into them. If the paper isn’t intelligent because it contains nothing but prewritten instructions, then a computer cannot be intelligent either, since it merely runs programs written by humans.

At this point, a sci-fi-minded student might argue that a computer would become intelligent once it can program itself — when the rules for playing perfectly are not written by someone else but learned by the computer on its own. This would be in line with the previous conclusion that the paper is not intelligent, but the person who wrote the instructions on the paper is. If so, a computer that would discover such rules could also be called intelligent.

The response to this is that, in such a case, a human would still write the learning program. A learning program is simply a set of rules that guide the computer in creating its own rules or instructions.

Is AI then even possible? A deeper inside into that is provided in the activity The Candy Computer.