Alice walked towards the mushroom on which the Caterpillar was sitting (or standing, though Alice) and smoking a hookah.

Something curious was happening with Alice too. Somehow, she was able to see the Caterpillar from the other side as well. “But I know very well”, said Alice to herself, “that it is not possible to see two different things from each eye.”

And soon, she started to see a number of different things as well. She saw some of the students of the Mouse, and she also saw a cat talking to another cat. Strangely, she also saw, very faintly, a woman. She tried her best to shift her focus to the Caterpillar. From the other side, Alice could see that he was sitting on the mushroom with his hookah in one hand and playing chess.

She could not explain why, but this made her happy.

She walked towards the Caterpillar and asked, “Mr. Caterpillar, sorry to disturb your game, but could you please help me find the Duchess?”.

The Caterpillar responded, “I do not know where the Duchess is, but I have some very important advice that will help you.” He was very old and had a very wise kind of voice — similar to Gandalf’s.

Alice nodded since she was desperately looking for anything that could help her.

The Caterpillar responded, “But first, you help me win this stupid game. I’m white, and I have to move.” Very soon, Alice forgot all about her troubles, and lost herself — or should I say found herself — in the boundless joy that Chess holds.

At first, it looked like a very scary situation for white. There was just a King and a Horse. So winning looked out of the question. And since Black had a Queen and a passed Pawn, it looked like Black would win with relative ease. However, the more she looked at it, Alice started getting convinced that it was a flat draw.

Both Alice and the Caterpillar agreed that the Horse has to take the Queen on g1. There simply was no other way to continue and not lose this. So, the Caterpillar played Nxg1!

Alice spent the next few hours on this. “Or has it been only a few minutes?”, she asked herself. It was getting more and more difficult for her to keep track of time and space.

After a lot of while, the Caterpillar started jumping on his spot at the center of the mushroom. It acted like a trampoline and launched him into the air. Alice was very afraid that the fall will hurt the Caterpillar, but as soon as he reached the topmost point on his parabolic trajectory, he spun a shell and converted into a Chrysalis. He landed on the ground with a thud and started talking again.

“Nxg1 was great… My opponent just resigned!”, said the Chrysalis.

Alice was too awestruck to talk, but managed a “but Mr. Chrysalis, who is your opponent?”.

“Can’t you see? It is the tree just in front of the mushroom. His name is Monte Carlo.”, replied the Chrysalis.

The Chrysalis went on, “He is very good at Chess. He searches through all the moves through his own very unique technique that he calls Monte Carlo Tree Search (MCTS). The only problem is, he takes ages to make a move. He started playing this match with my great-grandmother. And see, here I am, now a Chrysalis!”

Alice found this very curious indeed. She told the Chrysalis about reinforcement learning, and how some amazing scientists had trained RL agents to play Chess at a level far beyond what we had achieved till now, human or otherwise. She told him about Markov Decision Processes (MDPs) and how they were used to formulate the problem of Chess which was finally solved through an AI breakthrough called AlphaZero. Alice told him about the award-winning documentary on AlphaGo, its predecessor.

The Chrysalis, through the whole discussion, was listening very intently. He had one doubt though. He asked, “So, it is an assumption in an MDP that the current state is all you need to figure out the next state? You don’t need to know the previous states?”

Alice nodded. “Yes, and it makes a lot of computation easy. For example, in chess, you don’t need to know the opponent’s last move to find the best move. You just need to look at the board.”

Chrysalis replied, “That is not true. Chess is not an MDP. You can’t find the best move without knowing the previous move by the opponent. For instance, consider this arrangement.”, and he proceeded to show the following arrangement on the board:

“You see”, explained the Chrysalis, “Here, the best move really depends on what Black played. If they played e5, the best move is dxe5++ en passant, which is a mate in one.”

Alice looked at the board for a while, and she had to agree that he got a point.

“Anyway”, the Chrysalis continued, “I don’t think learning how to play chess is the best learning you can extract from this game. What is much, much more interesting and worth learning is the invention of the game. It is a beautifully crafted game, don’t you think? It did not start in the form that it is now. A lot of trial-and-error went into the evolution of the game when travelers took it to different places from its raw origins in India.”

Alice was very intrigued by the idea. She started thinking about how we could get an AI agent to invent chess. “If I knew”, she thought, “what exactly made chess so beautiful, then it’s easy — I would just need a generator that creates new games and a discriminator that judges their beauty. Maybe in order to measure the beauty of the game, the discriminator will need to learn how to play it well. And maybe the generator would someday create a game that is, I daresay, even more beautiful than chess.”

She also wondered whether these super-powerful models were actually better than humans. Or animals, a little voice inside her added. She thought, “While it is true that humans will never be able to beat computers at chess, can we really say a chess-playing agent really understands chess? For example, if you asked a human chess grandmaster to do a Smothered Mate or a back-rank mate on a (relatively weaker) opponent, they will most likely be able to adapt their game and successfully do it. But can an agent, without further training, stand up to a challenge like this one?”

Alice was just thinking about all of these things when suddenly she remembered that she had to find the Duchess, and that the Chrysalis had some advice for her. But just before she could open her mouth and ask, the outer shell of the Chrysalis started breaking, and within a few seconds out came a magnificent Butterfly.

Do you think it looked less like this and more like the Pokémon “Butterfree”?

Without wasting any more time, Alice asked, “Mr. Butterfly, you said you had a piece of advice for me for helping me find the Duchess.”

“Oh, I’m so sorry.”, replied the Butterfly. “I had really bad hearing as an old Caterpillar, and I heard The Chess instead of Duchess. But no worries! I can help you much better now. As a Butterfly, I have enormous control over the whole Wonderland.”

Alice asked, “How so?”

“Well, my species observed that as a Butterfly, whenever we flapped our wings, it changed the Wonderland in some small ways. Over the past millions of generations, we Butterflies have been observing and learning the patterns, and now we have gotten very good at changing some parts of the world according to our will by flapping our wings the correct way. Let me help you on your path to reach the Duchess. Farewell, dear.”

And with that, the Butterfly flapped its wings five times in some particular way, and then flew away, leaving Alice alone on a long path into the forest. After what seemed like hours (and also minutes), Alice reached the end of the path and found herself in front of a house.