An ungainly virtual life form created within Steve Grand’s new game.
The creator of a much-loved game for nurturing virtual creatures is developing a new playpen for more advanced beasts.
The self-taught British computer scientist Steve Grand still receives letters and e-mails from scientists and engineers thanking him for Creatures, a video game he designed in 1996 that inspired many people to pursue careers in artificial intelligence. He hopes that his latest project—a more advanced sandbox for nurturing virtual life forms—will be just as influential.
In the original game, players raised alien creatures known as Norns, big-eared, bug-eyed bipedal aliens that looked like a cross between an Ewok and a Furby. Players taught their Norns how to survive, explore, and breed (even if left alone, without player intervention, the creatures would gradually reproduce and evolve). Players could also teach their Norns language (by repeating the name of an object while the creature stared at it).
The game was a tremendous success—and not just in commercial terms. “Almost immediately, people devised Norn Genome Projects, breeding experiments and adoption agencies,” Grand says. “The marketing people told me to hide the science. ‘People aren’t interested in science,’ they said. They were wrong.”
Creatures made Grand well-known in the video game industry but in 1999 he left the medium to pursue AI in a different, and perhaps more scientifically controversial way: robotics. His creation, Lucy, was an android orangutan that he hoped would gradually learn to interact with its environment and communicate with human beings through language. The effort was written off by some academic AI experts, and Grand shut down the project before Lucy achieved such abilities.
Grand’s passion remained undiminished, and in 2011 the designer returned to the digital sphere with a new project, dubbed Grandroids. As with Creatures, Grand doesn’t consider Grandroids to be a video game in the traditional sense. “I think it’s fair to say it’s not a game at all,” he says. “There’s nothing you’re supposed to do; no way to win. It just is. It’s somewhere to visit and some fellow life forms to get to know. What you do with them is down to you.”
Grand says the new virtual beings in Grandroids are much more complex and lifelike than the ones in Creatures thanks to the increased computing power he has available to use, combined with “20 extra years considering the problem.”
The project, which has been funded by around 700 supporters on the crowdfunding site Kickstarter, doesn’t employ conventional animation to generate the walking and running of its creatures. Rather, their movements are driven by each creature’s own AI. “I’ve ‘wired’ dozens of virtual muscles directly to each creature’s ‘brain,’ ” he says. “These interact with a physics engine, so that movements and choices are as real as possible.”
This is just part of Grand’s unconventional approach to fostering AI. “Most AI work starts out with the mistaken assumption that the brain is a computer and thought is an algorithm,” he says. Grand is instead attempting to determine the behavior of his creatures through the arrangement and interactions of simple virtual neurons, enzymes, receptors, and genes. His aim is to replicate the way in which living organisms work. “Neurons send point-to-point signals, but many of the computations performed by living organisms are more spread out and diffuse,” he says.
Grand believes that he has found a way to imbue each creature with some form of imagination. “The Norns merely reacted to events in the world and learned passively how to react better in the future,” he says. “My new creatures have the beginnings of a proper mental life. I’m sure a lively debate will take place among philosophers of mind about these things, but I do think I finally know how to give my creatures the rudiments of an imagination.”
Despite these extravagant claims, Grand struggles to earn a living. He subsists on the financial support of two close friends and the Kickstarter donations he received at the start of the project. Nevertheless, he believes the work is not only worthwhile but necessary. “By creating alternative kinds of life, I hope to encourage people to ask deep and important questions about their own lives,” he says. “What does it mean to be alive, to be conscious, to be hurt, to be ethical? Artificial life can help to shed a little light on all these things. It’s the science of us, as seen in the reflections of aliens.”
© 2014 MIT Technology Review