Playing Smart

ps.jpg

The front cover of Julian Togelius’ book, Playing Smart

Playing Smart: On Games, Intelligence and Artificial Intelligence Julian Togelius

The MIT Press, Cambridge, Mass, 2018

ISBN 978-0-262-03903-1

 Playing Smart by Julian Togelius is the latest addition to MIT Press’s Playful Thinking Series. Readers familiar with the series will have an idea of what to expect – namely engaging, thought provoking, and fairly brief books (or long essays) on topics related to contemporary games and culture. Togelius’s contribution to the series, which covers the topic of AI and Games (alongside a discussion of general intelligence) is all of these things. Expect, then, to be given plenty to think about.

Togelius, whose Artificial Intelligence and Games (co-written with Georgios N. Yannakakis) was published by Springer earlier this year, is well placed to write a book on games and AI, and his deep knowledge of the subject comes through clearly.  Togelius advances three claims: (i) Games are the future of AI; (ii) AI is the future of games, and (iii) games and AI for games help us understand intelligence - ‘Game design is a cognitive science; it studies thinking – human thinking and machine thinking’ (xv). This is a lot to cover in the book’s 160 pages and as such things move at quite a pace. The brevity of the volume notwithstanding – Togelius rises to the challenge of the low page count – the first two claims are handled convincingly with a deft summary of the ways in which games, first analogue and now digital, have been used to test and develop Artificial Intelligence. This is followed by a more speculative discussion of the potential ways in which AI might contribute to the development of games (and ultimately develop games): ‘Let’s step into the future….’ (48). For Togelius a key obstacle to the deployment of cutting-edge AI in games is found in the conservative nature of a games industry that is ‘confined by economic realities to be highly risk averse and rather shortsighted’ (52). But he remains, it seems, optimistic.

This discussion is framed by a discussion about Togelius’ third claim – namely that the study of AI will inform our understanding of human intelligence. Here the book is perhaps best described as thought provoking rather than conclusive. As Togelius notes, ‘it is still very much up for debate to what extent there exists such a thing as general intelligence’ (129) – the notion of general versus specific (AI) intelligence is raised as a fascinating challenge for developers and that’s pretty much where this book leaves us.

Alongside all of this, you can also expect to discover what Togelius thought about in the shower, whether or not he has a driving license, and the impact that losing his cats had on his work on AI. The book’s tone is, as Togelius suggests, ‘informal and relaxed, sometimes even playful’ (xv) but it is no less rigorous for that.

A book being thought provoking and engaging strikes me as a good thing, and Playing Smart manages to be both. It may not answer all of the questions that it raises, but it seems likely to provoke lots of different thinking amongst its readers. To my mind, the most exciting issue wasn’t so much the nature of intelligence, but the nature of games themselves. In Togelius’s imagined future, a fairly likely (and in some ways already here) future in which games are adapted and designed by AI that respond to player input, games as rule-based systems (whose outcomes are validated by the engagement with those rules) are seemingly replaced by more fluid experiences that function along rather different, shifting, lines.  For me, working out what that might look like is a rich reward for an afternoon's reading .

 

Paul Wake

 

NB - all quotations and page references refer to the advanced proof copy of the book and may change.