Playing games is as old as humanity. Throughout history, people have used games to compete with and relate to each other in one way or another. Late Canadian philosopher Bernard Suits defined it as “the voluntary attempt to overcome unnecessary obstacles.” However, if the obstacles are unnecessary, then why would we want to overcome them? Author Oliver Roeder studied artificial intelligence at Harvard University and holds degrees in economics and game theory. In his recent book Seven Games: A Human History, he delves into the background of some of the world’s most enduring games — checkers, chess, Go, backgammon, Scrabble, poker and bridge. To find out more, Speak Up contacted Roeder. We began by asking him why we play games.
Oliver Roeder (American accent): I think the obvious reason is that games are fun, right? Games are pleasurable ways to spend time and as humans, we often have time to kill. We all had a lot of time to kill a few years ago, in the height of the pandemic, and indeed we saw more games being played at the height of the pandemic than probably at any other moment in human history, because we like to spend time this way and they are a very great excuse to come together with our fellow human. I think there are more subtle and interesting reasons, too, one of which is that games are a form of practice. Games crystallise small elements of the real world. For example, chess crystallises planning ahead or poker crystallises hidden information and deception or backgammon a kind of account dealing with the randomness of the world. So games capture these elements and, by playing these games, we can practise these little parts of the world and take the lessons we learn back with us to the real world.
A SENSE OF AGENCY
Games bring us together, bring us pleasure and teach us life lessons, but Roeder explains that games also give us a sense of agency.
Oliver Roeder: Games are the art form that captures human agency. Games put us in these positions that we might not have access to in our real lives. For example, there are many, many games that put me in the role of, say... a general leading an army into battle. Whereas no one in their right mind is going to put me in charge of an army in the real world, in a game, I’m able to live some version of that life. So games can transport us and give us access to modes of agency that we wouldn’t otherwise have. We crave decision-making and we crave having an impact and, if that’s denied elsewhere, games will always be available to provide an outlet for that desire.
DAYS OF PLAY
In a world with an increasing amount of leisure time, games take on greater importance. Bernard Suits went so far as to predict a utopian future, where all of our material needs are met by machines and where we play games all day.
Oliver Roeder: Suits argues that this future utopia is coming and the way that we best prepare for that is to foster institutions that support game-playing and to basically get ready by setting up the structure where we can all play games in the future. Is this a little fantastical? Yes, of course it is, but I do agree that games play this very central role. I don’t foresee utopia in the near future, unfortunately, but I think the argument still holds that fostering games and sports and these sort of institutions is incredibly important for humanity.
Games have driven the development of artificial intelligence. Programmers use a number of games to teach computers to learn and improve their basic coding.
Oliver Roeder: If games are practice for humans, games have also been practice for computers. Like, the earliest computers we had, seventy-eighty years ago, played chess and checkers. That’s kind of how their creators tested them. Computers and humans quote-unquote ‘think’ and play games in very, very different ways. The kind of starkest examples of this are the chess machines in the 90s, most famously Deep Blue, which beat Garry Kasparov, at the time the world chess champion, in 1997. And Deep Blue was good because it was extremely fast and it could search millions and millions of chess positions every second and approach chess as like this brute force optimisation problem.
IMITATING THE BRAIN
Recent AI projects, like AlphaGo by Google company DeepMind, work in a different way, using neural networks that are modelled after the human brain. AlphaGo conquered Go in 2016, but why do developers make machines play games in the first place?
Oliver Roeder: Because they thought machines would go on to do things in the real world after they had conquered chess. And, for a long time, that wasn’t true. All Deep Blue ever did in its whole entire life was play chess. But these more modern neural networks, so-called deep learning techniques, have showed more promise off the Go board. Look at some of the Deep Mind projects like protein folding, which is important for drug discovery and that kind of thing.
Artificial intelligence is particularly good at solved games, where the best moves are a mathematical certainty.
Oliver Roeder: That’s true of checkers, that’s true of tic-tac-toe, that’s true of a game called Connect Four, it’s true of a few other games. It is not true about chess, even though computers are much, much, much stronger than humans at chess these days. Chess is not solved. Just because the computer knows the right answer doesn’t mean it’s not interesting for humans to strive at the game. Even if they’re playing suboptimally, that’s fine, it’s still fun. It can still be interesting and the concept of ‘striving play’, the play of the beginner at a game is still really important and can be really enjoyable.
With games of chance, like poker and backgammon, humans will always have a chance to win against a machine. What’s more, they can use AI to learn how to play more efficiently to boost their chances.
Oliver Roeder: Every time AI comes in and conquers a new game — whether it’s checkers, chess, Go, poker — the elite players go through these stages of grief. There’s anger and denial and ultimately acceptance. And, ultimately, the elite players of these games realise, “Well, this computer is really damn good, we should try to learn something from it.” So, basically, the way that top players in these games use AI now is to train and to learn.
Neural network AIs are self-learning programmes, which makes them very good but often beyond human understanding.
Oliver Roeder: The issue with a lot of these neural network techniques is, we can see them perform very, very well, but we don’t know exactly how. It’s what’s often called a ‘black box problem’. So this neural network takes in all this data, trains itself, builds these connections within its system and can play, but these sort of connections and whatever its quote-unquote ‘intuition’ is is not accessible to us. In other words, the computer can’t explain why it’s doing what it’s doing — it just does it. The same is true of AlphaGo in Go or modern poker-playing machines and stuff. So they’re really, really good. The issue is, they can’t teach.
THE HYBRID MODEL
Ultimately, play makes us human. Chess grandmaster Garry Kasparov pictured a future in which the best chess-playing entity would be a human and a computer working side by side.
Oliver Roeder: The idea is you would get the best of both worlds. You would get the raw emotionless calculating power of the machine and the wisdom and intuition of the human, and, if you compare those, if you pair those together, you get the best games-playing entity you can imagine. I think this is probably going to be true in domains beyond games, too. I think it’s already true. Like, how many of us do our jobs with a computer by our side for the entire day? I mean I hate to give it even more publicity, but the whole Chat GPT thing might be really useful, but it’s only useful with a human guiding it and working with it. It’s not science fiction; I think it’s already happening.