My short story

Panayiotis Kamvysselis (pkamvyss@mit.edu)
http://web.mit.edu/pkamvyss/www/
1/15/1998

The following summarizes in prose my technical background and education to date. The second section lists some of my activities.

1. Academic background

Even though I had hacked on a Commodore 64 while my family was in Greece, my interest in computers began in earnest during high school in France, where my family had moved to when I was 15. Students were required to have a programmable calculator for use in their scientific classes and I was fortunate enough to acquire the HP-48SX, the best hand-held scientific programmable calculator at the time. Unsatisfied with the slow execution speed of code written in the calculator's high-level language, I spent many free moments between classes--and even some during the less exhilarating ones--to learn assembly for the Saturn, the calculator's micro-processor. Unfortunately the calculator is not intended to be programmed in such a low level by users. I discovered however that it was possible to encode a program as a graphical object and then execute it using an undocumented system call reserved for technicians. Now the way was open to the few hundred programs that I wrote during my high school years in France. I wrote: a full-featured 3D-function renderer; games such as a Tetris and a naval battle for two calculators communicating through their infrared ports; a text editor with custom characters and pictures; audio wave playback software so that the calculator said "bonjour" when turned on; a drawing program that used flickering to simulate shades of gray on the monochrome display; interrupt drivers that trapped keypad events such as "reset memory" and executed custom code; programs that allowed using the infrared port of the calculator to open some cars equipped with infrared-operated keys (radio transmitters eventually replaced them), etc. Seeing my enthusiasm, my teachers in France were very kind and donated a computer that I took home to interface with the calculator.

While a Freshman at Tufts University I got interested in robotics. Professor Haralambos Doumanidis hired me to create a 3D-scanner program for his "Scan Welding" process, which is a way of printing in three dimensions. Using C and the Macintosh toolbox I wrote a complete Macintosh application to integrate a laser, a camera and a turning table for the project. My work was subsequently used as the basis of several of Prof. Doumanidis' graduate students. I also wrote a programmable autonomous robot simulator that Professor Doumanidis used in one of his classes to test student code before execution on real robots.

During the summer 1995 I pursued my robotics interests in my first VI-A assignment at IBM's Watson Research Center. I worked in the medical robotics group headed by Dr. Russell Taylor. My task was to calibrate LARS, the laparoscopic robot system IBM had built in collaboration with Johns Hopkins medical school. I realized that the expensive calibration procedures used for other robots were unnecessary. Taking advantage of the "remote center of motion" design of the robot, my calibration method achieves sub-millimeter accuracy by using the robot itself as a measuring device rather than relying on expensive additional hardware. In addition, it is completely autonomous, quick, reliable and can be performed even by inexperienced users. My experiments with the calibrated robot further led me to the discovery of a major design flaw that was subsequently corrected in all of the eight manufactured robots. The calibration procedure is still used by the Johns Hopkins team for calibrating their LARS robots, which are presently used for liver cancer treatment.

In the summer of 1996 I worked for Dr. Kenneth Salisbury of the MIT AI laboratory, on haptic interfaces. I was interested in integrating visual and haptic perceptions on the computer to simulate the feeling of textures such as sandpaper using the Phantom haptic interface. The results were encouraging, but Dr. Salisbury left MIT and I chose to start working on something different for my M.Eng. thesis.

Professor Horn's machine vision class was instrumental in focusing my interest in robot vision. I enjoyed the mathematics involved and, more specifically, I was intrigued by the fact that vision, such a fundamental robotics problem, is still a long way from being totally understood. For the final project of the class I implemented in Java a shape-from-shading algorithm that recovers 3D shape from a single image of an object. The user can interactively select regions of interest for shape recovery and once the algorithm converges, the user can move the light source and the program re-computes the new shading in real time.

I recently took Professor Jordan's class in neural networks and became very interested in probabilistic approaches to robot vision and machine understanding. It seems much more promising to try to understand intelligence as a collection of neurons interacting with each other in non-deterministic ways rather than high-level procedural abstractions. It is hard to believe that we each have inside our head anything resembling an expert system. It is more plausible that, instead, we understand the world and in particular mechanical devices, in a highly neural fashion. With Paul Viola and Randall Davis of the MIT AI Lab we have begun work in computer understanding of mechanical devices in a probabilistic/neural framework. Initially, I have demonstrated an incremental estimation-maximization algorithm that correctly segments images obtained from the moving wheels of a mechanical clock, which will constitute the basis of my M.Eng. thesis. I would like to pursue this line of research by having the computer propose alternative clock designs.

Computers have historically played little role in design, an activity reserved for people. Thomas Stahovich in his Ph.D. thesis at the MIT AI Lab showed that it is possible for computers to abstract an engineering sketch to a multitude of different working engineering designs. His approach is based on abstracting the configuration space of the sketch in question and iteratively changing some parameters until a working simulation is obtained. I propose to start with a working mechanical device--for example a clock--and have the computer propose alternatives to its design by looking at it through a camera and by asking the user some questions. If a computer can do that using solely neural networks, it would have "understood" in a certain extent how the clock works. My proposal can be summarized in the following sentence: replace "search and simulate" by "learn and estimate" in machine understanding of mechanical devices.

MIT has resources that could be of enormous assistance in my task. Paul Viola is also interested in probabilistic understanding versus traditional AI. Randall Davis has a broad background in expert systems and his advice on what works and what doesn't would be invaluable. Berthold Horn is world renown for his expertise in vision systems and he could play a pivotal role in my thinking. The Learning and Vision group at the AI lab is made up of people with very similar interests. The MIT AI Lab as a whole has been my home for the past few years and the people there have always been wonderful in both criticizing and appreciating my work. Finally, Ford Corporation has expressed interest in investing in work at the lab that encompasses novel design techniques of mechanical devices as Prof. Davis suggested to me.

2. Extra-curricular activities

1. Athletics
   I have a very broad background in physical education. Having tried almost every sport that I could at MIT, which included the MIT 3 hours/week physical education requirement, I became member of the karate club, where I obtained the blue belt in two year, I played squash with partners for over three years, and have been swimming intensively for the past year. Every year I participate in a dancing demonstration of the Hellenic Club at MIT. My typical day for the past year or so includes at least three hours of physical activity: weight training, running/step master/swimming. I have been able to achieve a body composition of 4% fat. I measure 6' and weight 150lbs.
2. Technical/Professional societies
   I am in the process of becoming member of the IEEE professional society.
3. Honor Societies
   I was elected member of Tau Beta Pi and Eta Kappa Nu, the national electrical engineering honor society at the beginning of my Junior year. I have been vice-president for Mass Beta chapter of Tau Beta Pi for the calendar year 1996 and president for the calendar year 1997. I was responsible for the 1997 TBP career fair, which gave our chapter a net income of over $20,000.
4. Service organizations (on and off campus)
   This semester alone, I spent about 15 hours per week on average tutoring students in the electrical engineering and computer science of MIT, through a departmental community service program. I had 7 different tutees in 4 different subjects.
5. Public Service
   I implemented in 1997 the first online MIT survey that focused on the fundamental question of the +/- grades for the MIT transcript. Over 2,000 students used the survey over the web. This is a first for MIT because usually surveys get very low response rates. I am member of a team of students who are interested in helping focus feedback for all MIT subjects on a single convenient homepage, to make the process as painless as possible for students. We spend many hours weekly discussing ways of making the service better. I also volunteered to become the WWW chair of my dorm, and created an online movie database that solved many existing problems with missing videotapes.
6. Campus publications
   While president of the MIT Mass Beta chapter of TBP, I had the idea and co-founded a series of dinners called "TBP Leonardo da Vinci Dinners." These elegant and formal dinners held once a week starting this past semester brought together 20 MIT TBP members and about 5 faculty members per dinner. One of the faculty members spoke for about 30 minutes on topics of his choosing and discussion over dinner focused on that topic. The topics ranged from "the moon, planets, black holes and the universe," by Walter Lewin, professor of Physics, to "the role of the arts in the education of the engineer" by Alan Brody, provost for the Arts. The food was just as elegant and came every time from a different country. The success of the series was enormous. It made front page at the MIT "Tech" over three times, center page of the MIT "Tech Talk," was proposed to the President of MIT as a potential solution to the drinking problems that involved fraternities. More fundamentally, every single of over 50 MIT professors and 100 MIT TBP members who participated enjoyed the event very much. We will be continuing the series hopefully for many years to come. I have also been in various publications for dancing at the international student fair, for being in the Guinness Book of World Records (highest acceptance rate for a single family at MIT), and for other TBP activities that I supervised.
7. Artistic talents
   I spend a lot of time painting, drawing and sculpting. My work has been described as "inspired" and with "great potential." A short video I created called "Zero Gravity" was selected by the Arts department to figure among the demonstrations MIT has for visiting professors in the arts.