In early middle school I found a new friend. His family had a computer. This was pretty much my first meeting with computers – what they do, how they look and what kind of noises they make. While all of these attributes have changed since then, some properties remain. We learned maths and language by playing games on the computer. What’s interesting is that we learned from experimenting and pattern generalization, not by being actively taught vocabulary or how to do math problems.
In artificial neural network theory, that is modern artificial intelligence for you, there is the concept of supervised learning. You give the network a set of input values and let it know what the appropriate, corresponding output values are [a training set]. The network will, if you’ve managed to do it correctly, and the stars are right, learn the pattern between input and output and be able to predict the correct output for input values it has not been trained for. Humans do this all the time, it’s called understanding things. Only when you understand something will you be able to generalize the knowledge and use it in new situations. At my friends’ place, the computer was supervising us implicitly, by letting us know when we’ve done right, and to some extent, when we’ve done wrong.
We didn’t have access to that many games back then, but we were fascinated by the machine, so we played whatever game we could get hold of. We really liked adventure games like King’s Quest and Space Quest. We would play them whole days at a time, exploring the game worlds and solving puzzles. Looking back, I realize these games were highly educational to us. As the games were in English, our secondary language, we had to work to understand the game text and phrase the instructions to the game correctly. I will forever remember the difference between “three” and “tree”. Space Quest III did not at all accept my instruction to “Kick three” in order to break an old tree in two and use it to get across a gorge, and it got us stuck for days trying to come up with something else to do to move the game along. Each area [screen] in the game would typically come with a description. We soon learned to type ‘look around’ on every new screen to learn what parts of the world we might be able to manipulate. This implied mapping new words to known images. I don’t remember ever using a dictionary while playing, we learned by trial and error and educated guesses on what’s what in each screen. All in all, these games did wonders for our vocabulary, resulting in me getting the top grade at end exams in English at upper secondary school. This is a kind of supervised learning. When we tried multiple solutions, we got feedback if they were right or wrong.
Now for something stranger. We found a math game. It had symbols like ‘a’ and ‘b’ and ‘x’ in it, which was totally alien to us at the time. Professor Weissman’s Algebra Tutorials it was called. We didn’t realize then that this was a program to test students’ skills in basic algebra. It had a number of mathematical problems you were to solve. If you failed, it would show you a stepwise solution. It would not explain anything, just show you each line of the calculation. For some reason, this intrigued us and we determined to beat all the challenges in this game. My friend and I would sit together and reason about what the correct solution might be, knowing nothing about the underlying mathematics. That’s how we learned basic algebraic manipulation of equations! I ended up studying engineering at university level and did my share of advanced calculus (maths).
I believe being two and knowing some English and maths already helped our learning significantly.
Thinking along the lines of the tendency to follow the path of least resistance (people are lazy), I believe if we had had other, equally fun, less challenging games, we would have played those instead. Is that the situation today, I wonder? With a plethora of highly entertaining games, how does a game that forces you to stop up and reason compare?