Dr. David Grossman

>Dr. Grossman, a fellow of the IEEE for contributions to Robotics,

>spent 25 years at IBM in research and management in robotics and

>artificial intelligence punctuated by three sabbatical years at

>Stanford. More recently he co-founded LiveCapital.com, an Internet

>startup company that provides instant online financing for small


Dr. Grossman began his talk by defining a robot is something that is "surprisingly animated for the time and place where it was found." For example, a washing machine would have been a very surprising invention 100 years ago, however mundane they are today.

He showed all of the most spectacular ones from history and mythology, starting with an Egyptian statue that would speak at dawn, and continuing through Greek doors that would open when a fire was lit, and many surprising and entertaining machines and hoaxes from the middle ages and the 19th century. He covered things in chronological order, so that the Golem of Jewish tradition was long before Mary Shelly's Frankenstein, which was not long before a duck that was animated by gears and cams, a remarkable piece for the technology of the time.

Twentieth Century stuff was covered in much more detail, with more of an eyewitness slant. He said that the most successful way to automate something was to begin by defining the task in the most elemental way, so that the tasks to be done can be separated out and the machine to do them will be simple. As an example, he pointed out that people failed for hundreds of years to make a flying machine because they were trying to make wings that flapped like a birds. It was only after the concept of a fixed wing came along that success became possible.

Dr. Grossman does not think that robots have been responsible for the loss of many human jobs. He thinks rather that jobs that robots are doing are mostly ones that humans don't want or can't do, for example doing the masks for semiconductor chip circuit layers. It's not possible to get a human with a paintbrush to draw those lines.

Looking to the future, Dr. Grossman thinks that the problem of getting a computer to understand natural language processing will take at least 50 years to solve. While speech recognition is a solved problem at this point, understanding is so context dependent, not just at the sentence level, but also at the idea level, that he does not see how a computer can be made to do it in the near future.

Tian Harter