Milind Tambe, an associate professor of computer science, will be using science fiction as problem sets in a class on artificial intelligence for undergraduate programmers beginning in the fall, 2006 semester.

 

"Computer science is catching up with the ideas in these stories," says Tambe. "We are using science fiction as the spice for the main dish of teaching an important new area of our discipline."

 

While a number of universities use science fiction to introduce concepts in physics and other fields, Tambe believes his course is the first of its kind in computer science.

 

Tambe and third-year PhD candidate Emma Bowring worked together designing the class, CS 499, "Intelligent Agents and ScienceFiction." Bowring will be the teaching assistant for a class that "she made very significant contributions in creating," said Tambe.

 

The class will focus not on robots per se, but on their "minds," what are called in the field of artificial intelligence "agents." These are virtual robots, disembodied machine entities that can create strategies to achieve ends, and even negotiate with each other to cooperate while doing so.

"Science fiction provides three key benefits in this course," said Tambe. "First, it is a great motivator and it provides context, generating excitement about artificial intelligence topics in general, and agents and multiagent systems in particular.

"Second, science fiction also helps provide a perspective on how far we have come in our research, as well as current limitations, and future research challenges.

"Third, science fiction literature is a great vehicle for understanding the impact on society if agent-based computing truly succeeds."

Most of the texts will be standard scholarly references in the field of AI. But the assignments will also include science fiction films and tv shows, along with such famous stories as Asimov's "Runaround" — the 1942 tale that introduced his famous "Three Laws of Robotics."

In this story, set in 2015, astronauts on the planet Mercury send a robot named Speedy on a vital, but dangerous mission to bring back the element selenium. Instead of obeying, Speedy starts running in a circle around his destination. The reason, the humans discover, is the robot's calculation of required behavior conforming to the second law of robotic,: "A robot must obey orders given by a human," is in delicate equilibrium with its necessity to conform to the third law: "A robot must protect its own existence."

The humans manage tobreak the cycle by convincing Speedy that they are in mortal danger, which brings into play the top-priority first law: "A robot may not injure a human being or, through inaction, allow a human being to come to harm."

The syllabus asks students to analyze Speedy's thinking with what is called in "belief-desire-intention" or "BDI" logic, which formularizes persistent agent goals, with questions like "(a) Explain in BDI logic the commitment formed to save humans. (b) Is this commitment only invoked when a human is in danger or is it present under all circumstances?"

In more traditional academic course syllabus language, the course will cover "introduction to agents, elementary decision theory and reasoning under uncertainty, elementary game theory (includes Nash equilibria and prisoner's dilemma), teamwork and belief-desire-intention logics, emotions in agents."

Other science fiction source materials that will be discussed — and coded — by the class include Star Trek episodes on the alien distributed intelligence (one mind in many bodies) called the Borg; and on the emergence of emotions in Lt. Commander Data.

The USC Viterbi School and its Information Sciences Institute are leading centers of research in agents and artificial intelligence, and many of the non-science fiction texts the class will read are original papers by USC researchers including Stacy Marsella, David Pynadath, Jonathan Gratch, Gal Kaminka, and Tambe himself.

Tambe also hopes that some authors of science fiction sources to be analyzed and coded will visit the class.

"This will be a rigorous class in state of the art computer science," said Tambe, "but it will be one that I think will challenge students in an interesting way, one that they'll enjoy taking. I know I will enjoy teaching it."