The Rice University Computer Science alumnus and inventor of Lycos only went to college to get a good job. Michael “Fuzzy” Mauldin (Sid Rich, 1981) said, “That was my goal and my parental directive. In high school, I was good at three things: math, physics, and debate, and that suggested becoming a mathematician, a physicist, or a lawyer.”
Instead, he earned a B.A. in Computer Science. Mauldin was introduced to computers and learned to program in BASIC during a summer seminar following his junior year. Then he signed up for a FORTRAN class at his local community college and applied to six universities where he could major in Math and take more CS courses.
“Programming in FORTRAN at the time meant writing everything down on paper before punching out cards to feed into the mainframe. If you weren’t careful, you could get a syntax error instead of results,” said Mauldin.
“When I got to Rice, I took CS classes and Math classes. I almost failed Calculus because I was having more fun –and was too busy– in my compiler class. At that point, I had no future agenda, I just got caught up in the idea of computers talking with humans in English, like the movie Space Odyssey: 2001 and ELIZA, the natural-language computer processing program published in 1964.”
In fact, the problem of getting humans to interact with computers without learning a special language would drive Mauldin’s academic career. At Rice, Mauldin began exploring Linguistics courses to better understand how he could blend natural language (English) with typing to program a computer. Punch cards were already being replaced by keyboard commands, but programming languages remained more symbolic than speakable.
“I took first and second year Linguistics classes that ended up falling under Rice’s distribution requirements for both Linguistics and Philosophy, so I was also getting an introduction to Philosophy. It was a foundational part of my education, to sit in courses so closely related to each other that you didn’t even know which of the two disciplines you are in.”
In his senior year at Rice, Mauldin’s choices were to get a job or go to graduate school. He interviewed with a local engineering company but could not identify any interesting roles that would allow him to continue working with computers and natural language, so he applied to graduate schools.
Mauldin was accepted into the CS Ph.D. programs at Carnegie Mellon University and Stanford University, and he chose CMU. “I wanted to work on computers understanding English,” he said, “and Jaime Carbonell was the only one brave enough to take on someone as eclectic as me. I was a problem student because I had good ideas but was hard to manage; Jaime was a perfect match for me.”
Mauldin retired before reaching age 40, after inventing both Lycos and Julia–an early chatterbot. But he readily admits he didn’t succeed in defending his Ph.D. dissertation the first time.
“I had to remedial my defense. I’d given eight job talks and received four good offers. But when I gave my thesis defense at CMU, the thesis committee ‘wanted additional data about the absolute recall performance of my retrieval system.'”
It would take nine more months of doing surveys then having others look at the results and text to determine by hand which documents were relevant to which queries. Although he had four good job offers, he had nine months of work ahead if he wanted to complete his Ph.D.
He said, “That was the hardest and best decision of my life. I said ‘no’ to the job offers and stayed at CMU to get my degree. I have friends who, in similar circumstances, planned to finish their degree later but they never did. It’s hard to find time to finish your degree when you have more interesting things – things that pay your bills – to do.”
With his new Ph.D. in hand, Mauldin was once again on the job market but just ten months after he’d turned down four offers, he did not get any.
“Artificial Intelligence (AI) goes through seasons, just like the oil and gas industry. Machine Learning (ML), now called deep learning, is currently in its summer zenith. When I successfully defended my Ph.D., AI was in its winter. There was a huge crash in industry investment fund availability, so I stayed at CMU as a research faculty member, working on existing projects that overlapped my own interests.”
He coined the term ‘chatterbot’ around 1994 to describe computers participating in automated conversations. Mauldin also invented, Julia, a chatterbot utilizing AI to conduct a conversation with humans.
By then, even CMU was struggling to provide continuous salaries to its researchers. Mauldin’s group had won two grants that would begin in September, but his current funding would run out in April.
“By now, Jaime was my boss. He said, ‘You have four months when we’ll have to pay you [with department funding]. There’s this thing called the World Wide Web. See what you can do with it.’
“Now I’d just finished my Ph.D. in information retrieval and the web was full of people pouring content in, but no one could discover where that content was located. My thesis was about using AI and knowledge to solve really tricky problems in disambiguation. It’s a lot of work. And I started wondering if I could write a search engine without AI.”
Mauldin began working on the project for CMU in April 1994 and released his first search engine in July. Within four months, he had a model ready for his department to use.
“Inventing new tech applications is kind of like surfing. To some people, it’s easy. Most people, you put them on a board in the waves and they’ll just drown. If you’ve been around boards all the time and a big wave comes, you might surf it, survive it, and make it look easy. I was in the right place at the right time, with the right training: the advent of the web, while I was at loose ends at CMU, and knew a lot about information retrieval.”
The Lycos catalog, available on CD, was similar to a library catalog but with links instead of call numbers.
“We gave you links and you could go online with AOL and look at the information on the linked page. In 1995, we introduced an offline search product.”
Before the Lycos catalog, the only way to find information on websites was through educated guesses. If a user wanted to look at specific research being done at MIT, they had to look for the MIT website, and hopefully find the right department webpage, and then the professor’s webpage. There was no directory, much less a Yahoo or Google search box. But a graduate student at the University of Washington, Brian Pinkerton, had just released WebCrawler, a full text search engine using an external database.
“Brian had sold WebCrawler to AOL for $1M. I looked at what he had done and had some different ideas –similar to data fusion– and some ideas on citation analysis, which we called ‘popularity’ and those ideas helped us create the biggest catalog of the Internet. All that was patented in 1995, and CMU sold Lycos Spider Technology to CMG, Inc. the same year.”
Mauldin said he is “seriously retired” and spends his time ranching and exploring the wilds of Utah in his Jeep. He also designs fighting robots.
“Everything about the Internet is more complicated than when I learned it. Now I’m more concerned about speed control for my robot. Should I be considering the move from brushed to brushless motors?”
When it comes to advice for current CS students, Mauldin has only a few words.
He said, “Work on what interests you and make the choices that allow you to do that work. Eating is nice, so sometimes you have to compromise. Otherwise, find the change you want to make in the world and work to make that change.”