At the PAX East conference last year, a young man approached the microphone during the Q&A with Mike Krahulik and Jerry Holkins, creators of the popular Penny Arcade webcomic.
Instead of asking a question, he bellowed, “Welcome to ACTION CASTLE! You are in a small cottage. There is a fishing pole here. Exits are out.”
An awkward pause, followed by some giggling from the audience. “Is it our turn to say something?” said Mike.
“I don’t understand ‘is it our turn to say something,’” said the young man.
Instantly, Mike and Jerry understood, along with everyone in the audience born before 1978.
“Go out!” said Jerry.
“You go out. You’re on the garden path. There is a rosebush here. There is a cottage here. Exits are north, south, and in.”
The game was afoot.
They were playing Action Castle, the first of a series of live-action games based on classic text adventures from the late ’70s and early ’80s. Game designer Jared Sorensen calls the series Parsely, named after the text parsers that convert player input into something a computer can understand.
In Parsely games, the computer is replaced entirely by a human armed with a simple map and loose outline of the adventure. No hardware and no code; just people talking to people.
It’s a clever solution to complex problems that have plagued game designers for decades. How do we understand the player’s intent? Can we make AI characters act human, instead of like idiot robots? Is it possible to handle every edge case the player thinks of without working on this game for the next 10 years?
Making computers think and react like us is hard. So instead of making software more human, some game developers are trying to make humans more like software.
It’s a similar approach used by Amazon for Mechanical Turk — their motto is “artificial artificial intelligence.” By layering an API over an anonymous human workforce, developers can solve problems that are best tackled by humans, but without the messiness of actual human communication.
Projects like Soylent add another layer of abstraction, invisibly embedding Mechanical Turk in Microsoft Word to crowdsource tedious tasks like proofreading and summarizing paragraphs of text. The effect feels weirdly magical, like technology that beamed in from the future.
In the gaming world, this substitution usually feels less like magic and more like robotic performance art. These performers are software-inspired actors — people pretending they’re videogames.