Artificial Intelligence and the Job Market

When you use a search engine, typing in key words results in instantaneous results, but refining those results and providing the user with more accurate information is the task of many engineers, including Google's Dr. Anthony Francis. “My official title on my business cards is Renaissance Engineer – Search Quality. If you want something less prosaic, then I'm a member of the technical staff, or in Googlespeak, simply an Engineer,” he said.

Dr. Francis described his current role as “making a computer's memory better based on context.” He builds systems that try to understand how people search the Internet so that their results reflect the most relevant information. He became interested in AI through his interest in science fiction, including the robot stories of Isaac Asimov, the computers on Star Trek, and Hal 9000 from 2001. However, it wasn't until college that he decided to make a career out of his interests in AI.

One area Dr. Francis concentrated on is contextual information retrieval. He described it as, “Humans can remember an amazing amount of information and yet we are not distracted by all the other things that we know. If I'm talking to you about Star Wars and say ‘Ford,’ you'll think of Harrison Ford, but if we were talking about Mustangs, you would think of Ford Motor Company without skipping a beat. My goal was to make computers do the same thing.”

Dr. Francis explained various AI applications and how they are achieved, including “design intelligence,” which is done by studying the algorithms of how a human or animal responds and then tailoring computer programs to replicate those responses. This study of algorithms is called cognitive science, an entire field of AI that has produced very interesting results. Understanding algorithms is important because it allows engineers to attempt to predict human behavior and responses. This idea, according to Dr. Francis, is where design intelligence becomes AI.

“But [design intelligence] only goes so far. As computer problems get larger and more complex, the task of studying algorithms becomes too difficult.” This is where “machine learning,” or exposing a computer to evidence and letting it come up with its own answers, is applied. The difficult aspect of machine learning is that computers cannot work on their own—a human must first determine relevant questions and build a framework for capturing predictable answers. Dr. Francis offered an example of how machine learning is used for AI. “The robot pet I worked on at Yamaha is a good example,” he said. “The engineers who built it gave it a half dozen senses, but to make it learn who liked it we had to teach it to recognize people by sight and to interpret a touch on the head as a friendly pat. Then the pet behaved in a ‘natural’ way. That's machine learning and design intelligence working together. Machine learning is huge at Google, and at any large company that has to analyze a lot of data, a process called data mining.”

Another area of AI revolves around designing machines that can make choices. The military uses AI planning and logistics by providing computers with details and then using those same computers to help determine complicated plans to move heavy equipment and troops in an effective way.

As for preparing for a career, Dr. Francis explained that a background in computer science is important, but understanding the common programming languages can be even more useful. “Computer languages are constantly in flux,” he said. “One of the first computer languages used for AI (indeed, one of the first computer languages in wide use) is Lisp. Lisp stands for LISt Processing and focuses on giving a computer programmer powerful tools to manipulate symbols. Lisp is still in use, and many hardcore programmers argue that you need to know it to understand the techniques it used. It's very, very easy to do complicated AI algorithms in Lisp. However, it is a specialized language and you need to know a mainstream language as well.”