For some time now, digital artists have been making smart CG characters that star in computer games and feature films. But could this virtual AI be just as effective in the real world? Inventor David Hanson of Hanson Robotics in Dallas thought so. And his prediction was spot on.
The end result is Zeno, a physically constructed robot that uses advanced AI to reason and become smarter over time. Zeno, in fact, is the brainchild of Hanson and the namesake of Hanson's son. Zeno, which resembles a humanoid (albeit with a cartoonish rather than realistic appearance), is a prototype, first unveiled at WIRED NextFest last fall. The inventor hopes to produce the robot as a mass-market consumer product within the next year or so.
"The technology incorporated into Zeno has a huge potential for the consumer home-entertainment market," says Hanson, the company's founder.
While Zeno is new to the world, the company that created it has been operating for approximately five years. Known for developing the most expressive and intelligent conversational character robots today, Hanson Robotics has built a number of offerings that have been featured in Wired Magazine, Popular Science, The New York Times, and The Boston Globe, as well as on a range of television programs. Yet, to build this current-generation robot, Hanson chose to experiment with some technology that was unique to his world.
"I first encountered Massive at the Technology Summit in 2003, where Stephen Regelous [Massive's founder] and I were both nominated for World Technology awards," recalls Hanson. "Something clicked for us both as we realized how deep and common our vision for AI is, and how we could work together to build friendly robots for a better future."
Like Hanson, a groundbreaking developer in the robotics field, Regelous is a pioneering computer graphics software engineer from New Zealand who is best known as the creator of the Massive intelligent crowd-simulation system used in such films as The Lord of the Rings trilogy, Happy Feet, Ratatouille, and more. Together, the two visionaries partnered to design an extremely intelligent robot that can view a 3D mental image of its environment to determine and control physical action and reactions.
Intellectual and Physical Attributes
Massive's software originally was created to control intelligent self-animated crowd characters. The software enables artists to quickly create thousands of agents that think and act independently while responding to the environment around them. Since then, the software has moved into the foreground to animate characters closer to the camera.
"Massive is simply the best brain for automating lifelike characters," says Hanson, a former sculptor who worked for Universal Studios and Walt Disney Imagineering prior to transitioning into Disney Technical Development, where he headed robotics and materials projects.
In just three and a half months, Hanson Robotics developed a way for Zeno to interface with the Massive software, thanks to assistance from two Massive artists, Gregory Wilton and Mark Thielen. However, no tweaks were required to the AI software itself.
Prior to using Massive, Hanson Robotics had utilized its own proprietary animation engine inside its robots. With that system, all the interactive animation rules were hard-coded, programmed in C, C#, or Drools. "That was not an easy way to author or edit lifelike behavior," Hanson notes. "Massive makes the animation process much easier and more intuitive, using the tools that are the standard in animation."
The vision and decision-making capabilities in the Massive software give Zeno the ability to navigate, make facial expressions, and move its body based on what the robot sees within its physical environment. Hanson also created a character engine with speech recognition and conversational AI for language reasoning so that Zeno can recognize and remember both speech and faces, and interact accordingly. Thus, Zeno is fully conversational and responds appropriately; Zeno also can keep track of where people's faces are and will turn around so it can make eye contact.
The robot's body was envisioned by leading Japanese robot designer Tomotaka Takahashi and is reminiscent of the beloved Japanese anime television character Astroboy. Physically, Zeno is 17 inches tall and weighs just a few pounds. The robot is highly articulated, with nearly 30 built-in servos (specialized motors) in its legs, torso, arms, and face, and it runs on lithium polymer batteries. Moreover, the robot is able to display emotions on its smooth-skinned, flexible, expressive face, and can perform stunts with its agile and self-balancing body. For instance, Zeno can lie down, rise to a standing position, gesture with its arms, smile, make eye contact, and open and close its eyes and mouth.
It is in this area where Hanson's expertise in sculpting comes into play. The inventor regards each robot as a four-dimensional sculpture: Every action, expression, and movement is sculpted uniquely by the interaction between the mechanics underneath and the polymer on the surface of the robot's face.
These physical actions are driven wirelessly by a standard home PC through Hanson Robotics' own software. However, plug-ins pertaining to Massive's API had to be developed that allowed input and output values to be passed between the Massive software and Hanson Robotics' software so that the brain of a Massive agent could process the information received and subsequently provide the instruction for the control of the robot.
Zeno's brain is distributed among a number of internal processors located inside the robot, on the host computer wirelessly networked to the robot, and on a bank of servers on the Web. According to Hanson, this architecture allows the low-cost robot (whose price tag is envisioned in the $200 range) to be as smart as the best research robots in existence.
Subsequently, Massive runs on the PC or server side but eventually could be embedded into the robot. Massive serves as the physical brain, controlling the animation, physical reflexes, and most of the emotions and 3D reasoning. Massive shares the reasoning tasks with Hanson's Character Engine software, which handles the language, personality, conversational, and perceptual capabilities of the robot.
According to Hanson, Zeno's predefined movements and actions have unlimited permutations, and he will get smarter and more aware over time. The goal, he says, is to bring it to market as a children's toy, able to tutor, express, and teach a variety of subjects.
"With very little modifications to our software, [Hanson Robotics] is controlling the physical actions and reactions of a robot in the real world, which has exciting implications for Massive in a variety of markets outside the animation industry," says Diane Holland, Massive CEO.
All it takes is some smart thinking.
Karen Moltenbrey is the chief editor of Computer Graphics World.