Oral-History:Maria Gini
About Maria Gini
Maria Gini is a professor in the Department of Computer Science and Engineering at the University of Minnesota. Before joining the University of Minnesota she was a research associate at the Politecnico of Milan, Italy, and a visiting research associate in the Artificial Intelligence Laboratory at Stanford University. Her research focuses on methods to distribute intelligence among robots or software agents. Her major contributions include algorithms for multirobot systems, robot navigation, planning with incomplete information, and negotiation for agents. She is the chair of ACM SIGART, a member of the AAAI Executive Council, and a member of the board of the Intelligent Autonomous Systems Society. She is on the editorial board of Autonomous Robots, Integrated Computer-Aided Engineering, and other journals, and she was the chair for the 2006 Distributed Autonomous Robotics Systems Conference.
About the Interview
MARIA GINI: An Interview Conducted by Selma Šabanović, IEEE History Center, January 22 2015
Interview #806 for Indiana University and the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.
Copyright Statement
This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to Indiana University and to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.
Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user. Inquiries concerning the original video recording should be sent to Professor Selma Šabanović, selmas@indiana.edu.
It is recommended that this oral history be cited as follows:
Maria Gini, an oral history conducted in 2015 by Selma Šabanović, Indiana University, Bloomington Indiana, for Indiana University and the IEEE.
Interview
Interviewee: Maria Gini
Interviewer: Selma Šabanović
Date: January 22 2015
Location: College Station, TX
Early life, education, interest in robotics
Šabanović:
Your name and where and when you were born.
Gini:
Maria Gini. I was born in Italy. I did all my studies in Italy, and I came to the United States first as a postdoc. I went to Stanford, Artificial Intelligence Lab, a long, long time ago, 1976, and then I went back to Italy and then I started my job at the University of Minnesota in 1982, and the reason why I got to Minnesota, after all the places I could have gone, is because my husband is also faculty there. So we decided to get married, so I go to Minnesota and then I got a job, and I've been there ever since, which means I obviously like being there, otherwise I would not have stayed there.
Šabanović:
Could you tell us a little bit about your early years and how you got interested in robotics?
Gini:
So in Italy-- first, one thing that is a bit unusual, I have a twin sister and we studied together. So in Italy we studied physics because there was no computer science degree at the time-- there is a computer science degree now-- and some computer science was done in engineering, but I did not get interested much in engineering. They're all boys there, and so I didn't really like the idea. So we studied physics and done some computer science, a little bit of beginning with some ideas of artificial intelligence there. I went with my twin sister to Stanford when we had the fellowship from Italy, and we have lots of papers, we work together for many, many, many years, which is a little bit unusual, and lots of people when they saw the two last names the same assume it's husband a wife, so it was always a surprise. Say, "Oh, oh. It's a she." Yeah, it's a sister. <chuckles> So. So, we got interested in general in artificial intelligence, which is where we will talk <inaudible> robotics. I had a teacher in college-- again, we were studying physics. We were looking at cybernetics, which is kind of the term more common in Europe, and she gave us some papers to read, and I still remember a paper, which is a paper by a biologist in Chile, which is Maturana, and his paper, "What the Eye of the Frog Tells the Brain of the Frog," and what I thought was very kind of inspiring is the fact that you can understand the process, which is kind of complex, and sort of describe the process in a very precise way.
So they were coming up with ways of saying, "That's what it does, and that's the way we can model it," and I thought I had no idea that this was doable and was very, very interested in the idea. Then I started-- once I finished my degree, I started working as a kind of research assistant, research associate in the engineering department with a professor who had been at Stanford in the AI lab for a year or two years, and came back to Italy all excited about artificial intelligence, started a research group at the Politecnico di Milano, which is one of the top engineering schools in Italy, on artificial intelligence. And so we kind of started working with-- I started working with him. My sister was also working together. So we stayed together until I got married. We worked together for a long, long time. So we started doing different things in artificial intelligence, and he was telling us all the times the good experience he had-- I mean, California. This was, again, early, middle '70s, so California was the place to be, crazy and all those other things, and in Italy we were kind of much behind, so he was keeping on saying how good it was, so we decided to try to go to Stanford, so we got fellowships from Italy for both of us, and because Marco Somalvico is my, again, supervisor, knew Professor John McCarthy who was the director of the AI lab, so we contacted him and we are accepted to go to the AI lab. So very little knowledge of English, twin sisters, never really been too much-- in fact, we'd been in the U.S. just for a short vacation-- get on a plane, get off in Stanford, had to figure out how to find a house, the car-- I mean, all this other stuff-- and how to reach the AI lab. The AI lab at the time used to be up in the hills on Arastradero Road-- I mean, if people know the campus, the Stanford campus-- and so we had to kind of get up there, so we bought a bicycle and we're biking up every day and all this stuff, and there we started working in robotics with Tom Binford, who was one of the-- in fact, he was the person doing robotics there at the time.
So this was '76, '77, so it was a time when the main things in robotics going on, at least at Stanford, but also in other places, was trying to come up with programming languages for the robots. So Stanford had those two arms, the blue arm and the yellow arm, and had this programming language called AL for Assembly Language, which had already been developed at Stanford, and they were using it to program manipulations <inaudible>. It was a robotics arm. So the work we done there, again with my sister-- so it was kind of-- because AL was a compiled language, so you had to write a program and then there was-- run the program on the PDP-10 and then there was a PDP-11, it was controlling the robots. I mean, very complicated. So you make a mistake, you had to spend a lot of time. So we had this idea. Coming from the AI community, I was used to LISP as a programming language, which is interactive, said why not make an interactive system to do the same thing? So we kind of developed a system which is called POINTY, which was then used for a number of years there at Stanford, wrote some papers, and the idea was basically to use the same syntax as AL but make-- again, this was a long time ago, so there was no real use interfaces because there was no real graphics, right? I mean, this is past history <chuckles> but-- and so but Stanford had-- they were using TVs as terminals, and they had all this very complex system with disk space to map in the TVs, so basically it could write character. And so we design a very simple interface so that you can see the different frames <inaudible> variables that were defining the positions, the orientations of the robot, and you can see them and then you can visualize the important parts of your program, and at the same time you can issue individual commands to the robots.
So the reason why this system is called POINTY is because it started with an idea that Russ Taylor, who was working at the same place beforehand, which is-- and Raphael Finkel was also working there. So the idea they had was you needed to have precise coordinates for the objects that you want to pick up, and now as people, it's very difficult for us to find coordinates. So the idea they had is, "Let's put a pointer in the hand of the robot. We calibrate the pointer with respect to the robot, so we point to a point and then we say, 'This is corner. This is a cup,'" or whatever. So we used the same idea, and still again-- but POINTY needed really to be interactive-- right?-- because you want it to move around and pick up the position, and so what we did, we integrated this pointing with the programming language, so that then the coordinates you're getting were becoming frames and your definitions in your program, and that's kind of the work we did the year we were there, and at the time, again, there was a lot of interest in programming languages. Tomas Lozano-Perez at MIT was working on his language. There was a group in France. Then Russ Taylor was working with IBM at the time, was working on another language. There was a lot of interest in program languages that then kind of died. So if you look from the history point of view-- right?-- a lot of the robotics started in computer science in the programming languages community, and people were really trying to design specialized languages for robots, even the Unimate, where Bruce Shimano-- he was also at Stanford there at the same time-- he implemented, designed this language called VAL. Again, there was all sort of different syntax but the same idea, like you create a specialized language, allows you to specify motions for the robots and a little bit of sensing actions.
So there was not too much sensing yet at the time. There were cameras but not much more. All this kind of effort after I would say the middle '80s kind of died, I think in part because on one side, I think from the commercial point of view, it was not easy for people to use those programming languages, even though VAL is relatively simple, but still you kind of need to know how to program, and the people using the robots were not really computer scientists. So the company didn't really adopt those, and in academia AL was very-- I think was-- I'm convinced it was the best language. I mean, I worked on it, so obviously-- but it was a very good language, but it was not really portable because it worked for the Stanford arms, and then-- and again, there was this kind of interface between the PDP-10 that was doing all the processing of the language compilation, and then you had to download the programs on the PDP-11 which is controlling the robots. So each group had different hardware, and I would say the problem is still here today. Right? There is an attempt to now-- I mean, there is a lot of results <inaudible> like ROS, is an example, that works across multiple platforms. But at the time, that was not the case.
So you couldn't say-- now, in fact, when I went back to Milano to Politecnico, I couldn't use AL. I didn't have the robot. I didn't have the-- I mean, I could run the software but there were no robots, and really nobody else really could use things. So it was very hard to share those results, which I think is partly the reason why they kind of disappeared. Again, the companies didn't pick them up. The hardware wasn't there. There were lots of reasons, and now what I think is interesting, in the last few years things are coming up and there is a part of programming languages which is called domain-specific languages, like DSL, and there are people now coming back with the same idea, like you want a language specific for your domain, but is not really independent. It's more like an interface, right? It gets mapped with regular language. And there are efforts-- I mean, I had one of my students who had been working on this and design a DSL for robotics-- for robots, can navigate around. So in science, a lot of things come and go, and again, I'm kind of interested and I would like to see if this time the DSL is going to have more success than the earlier languages, because I still think the languages at the time were very high-level, had a lot-- I mean, considering also the state of the art of programming languages at the time were very advanced, and it's been kind of going downhill in a sense. You have to go to low-level details and all these things. So now hopefully things are coming back, but we'll have to see.
Stanford
Šabanović:
Would you say ROS and another <inaudible>?
Gini:
Yeah. Yes. Yes. Right. Because at least ROS now is becoming kind of a standard platform that provides a bunch of libraries and functionalities. So now almost everyone wants to have software which is ROS-compatible. So it at least provides kind of a basis for sharing software. But it's just happening in the last few years. I mean, it took a long, long time before getting to this point.
Šabanović:
Were there any other projects that you worked on while you were at Stanford?
Gini:
No, at Stanford I really worked all the time on this project, because it took-- I mean, first, I was not a computer scientist. I had my degree in physics. I'd done some courses. I knew a little bit of programming. This was also so-- such a long time ago. There were not too many people that really knew how to write programs. So part of the time that I spend is to learn the programming language that was used at Stanford, which was just another thing you use only there, that was used to implement-- AL-- that was used to implement POINTY, and then trying-- and then we did a bunch of experiments. We developed some software for a camera system. It was at a time when there were few early camera systems, were not as difficult to use and not as expensive. They were still very slow, and in fact there's a time when the software started in which you had software that can recognize blobs in images, and so we used that for some experiments in trying to pick up different objects with the arm, and it was slow going. I mean, I remember many nights <chuckles> trying to stay here and get these programs to work <chuckles>, and the robots always had some problems. I mean, same problems are still true today but they were much worse. There was only one robot arm. There were two arms, the blue and the yellow. The yellow never worked, so we had only the blue arm, so we had also to share the arm with other students. So again, scheduling different times and tried to do things. But it was really fun.
Šabanović:
Did you work with any of the other students there?
Gini:
Yeah, yeah. I was working-- there was a student-- I think he was getting his PhD at Stanford-- I think it was in mechanical engineering-- Shahid Mujtaba, M-U-J-T-A-B-A-- and he still-- I have lost track of him, but I visited him a few years ago and he's still in the Bay Area. He used to work I think for Hewlett-Packard, if I remember. So not really in robotics. So he really helped us with the interface with the PDP-11, the assembly language, which I was not too, too comfortable with, and the other person who was also a student there at the time was Ron Goldman. He was working also on the AL system, and I haven't seen him-- he stayed in the Bay Area. He's not in academia. I'm not sure what he's been doing. So those are the two people we were working close together, and then there was-- Vic Scheinman was a student there, and Bruce Shimano was across the hall, and Tom Binford was around there. So there were all those people in the robotics group, and then there were other faculty. Saul Hermana [ph?] was there teaching, so I took his classes, and Richard Waldinger, who <inaudible>. He's kind of more AI side. I mean, again, office across so we could see each other. <chuckles> I mean, it was such an exciting place because there were a lot of things going on, and there were still very few places in the world where you could see-- and John McCarthy was around there, and Hans Moravec was a student there at the time. Rodney Brooks was a student at the time, so I met with him, and David Lowery [ph?]. I mean, they were all students in the Tom Binford group. So again, it was a very large group of people. They're all famous, right? All around the world doing cool stuff. <chuckles> So it was really a fun time to be there.
Šabanović:
After Stanford, you went back to Italy?
Gini:
I went back to Italy, yeah, because I had the-- I mean, common things to many international people, I had the J1 visa from Italy, so I had to go back for two years, because I was funded from Italy, and then it was-- so we decided to-- so my husband was a student at Stanford, but he was a student in the numerical analysis group, and we kind of met and he liked Italian food. I mean-- sorry, I don't want to say anything bad, but that's how things start. <chuckles> And he was finishing his PhD, and then we kind of decided to get married. He finished in 1981, so then he got a job in Minnesota, and I had no idea, and so he calls me and he says, "Well, I have a job in Minnesota. They might hire you. You want to come? You want to get married?" All these in the same thing, right? <chuckles> And I said, "Sure." I had no idea. Often I tell the students, "In life, sometimes you have to jump into things." Right? "If you are too cautious--" I mean, maybe I was just crazy. "But if you are too cautious you miss things, right?" So I said, "Sure, why not?" Right? Again, no idea. No idea it was a cold place, for instance. But it's a very nice place. I mean, Minnesota is a good place, and the-- I think people value education maybe more than in other parts of the U.S. There's kind of a more well-educated workforce. The students are-- in general, the undergraduates, they are-- I don't know how much-- typical Midwestern. Very solid, very polite, most of them, really doing their job, good ethics-- I mean, all this stuff that you don't find everywhere else. I kind of learned after being different places. And so it's really a pleasure to work there.
Minnesota
Šabanović:
What kind of environment did you find in Minnesota when you went there?
Gini:
So when I went there-- so the department was more. It was like 14 faculty, no women. So I was the first one. And in Italy there were sort of similar problems. Like in engineering, where I was in at the Politecnico, there were no women in a group, my advisor, but there were two other women faculty kind of in the same sort of computer science division. So even though there were not too many women, it was still okay. In physics, there were not too many women, it was okay. So I never really thought there was a problem with women. So I got to Minnesota, I'm the only one. Initially I didn't care. I mean, I said, "Sure, why not?" Then I start slowly observing. "Where are the women here?" And one thing was kind of-- it's not to do with me-- I think it's to do with me being a woman, but ended up working well. So the department was growing. There was not too much space. So the offices were sprinkled a little bit around, so they found they had an office-- it was kind of a very large room with-- used to be I think a dean's office or something, so it was a very large room. There was a small bathroom and there was a women's powder room in the back.
So what they decided to do, my husband and his secretary for another program got the big part of the room, and my office was the powder room behind the bathrooms. Were not used-- right?-- because they were kind of private. But it was a very tiny office, and I didn't care much. I mean, I can work under a bridge. Whatever, I don't care. I mean, I'm not looking for signs of power. So I didn't think much about it at the time. Some of the computer science associate companies in the department were trying to convince the legislature we needed a building. So one day I'm sitting there doing my stuff and the governor of the stuff, Rudy Perpich at the time, comes into my office, followed by all the usual-- and they were showing how bad the situation it. "Look, we have to put faculty in the women's powder room. We don't have space." Well, we ended up getting a new building. I'm sure there were other reasons. But at the same time-- again, at the time I didn't think about it, but I don't think they would have put a man there. So I mean, again, just to be honest. I'm always willing to accept, assume that people have good wills, good intentions, but thinking backwards, I say, "Maybe I should have said, 'No, I don't want this office. I want a real office.'" But at the end, everything worked. Now the department is 40 faculty, so much, much bigger. Now we have four women faculty, one more coming in fall, which is better, but still if you look at the ratio, we're still not where we should be. There's no-- <chuckles>-- I mean, there's nothing really-- <laughter>
Gini:
I don't think there are-- there's no explicit bias against women. I mean, I can't say-- I work very well with my colleagues. Everybody's friendly, welcoming, and so I think it's more the implicit bias that people have. I mean, we have interviewed different women, made offers. Sometimes they don't come because they get offers so many places, but still it would be really nice to have more women, because we know that you need women faculty to get women students. I mean, it's all kind of a chain. So the situation isn't good. I mean, we'd really like to get it much better, and I've been doing a lot of stuff to try to get more women in general in computer science.
Šabanović:
What kind of things do you think would make a difference?
Gini:
I mean, at the end of the day, we need the numbers, right? There are studies that show that if you have about 30 percent women in a group, the dynamics change. Right? The question is how to get there, right? Until we get there, I don't think it's-- so the difficult part, how do we get there? But until we get there, it's going to be a struggle, because it's true. I mean, same is for other minorities. Like I complain we don't have enough women, but I look around-- I mean, in my department, there is an African American faculty who is a junior faculty. He is the only one. I've been there since '82. Never one. Even for talks, colloquia, I think maybe 10 years ago I was the person to invite an African American faculty to give a talk. I mean, again, I think of women because obviously-- but there are other groups even more underrepresented. I mean, I had a Native American student who got his PhD last year with me. But again, it's the first one I've seen in 30 years.
So clearly this is not sustainable, right? But how to get there, and for the other groups I think is even worse. Again, if you are the only one, there's no way you will feel welcome and so on. You really need to bring the numbers up. But because the numbers are small, people don't come. Right? It's a vicious cycle, and it's really, really difficult. One of the things that seems to help is to try to-- at least for women-- reach them out earlier. Right? I mean, there are a lot of studies that show if you bring robotics or so to elementary school, the girls are as interested, as engaged as the boys. I mean, there's absolutely no difference. The difference starts a little bit in middle school, when the girls decide that it's not cool to be high-tech, right? And so if you reach out in high school-- I mean, it's good, but it's already a little bit too late. The question is if you start in elementary school, it takes a long time before they come to college, and some of them get lost still.
So I know there are a lot of programs-- in Minnesota there are people I know that do things-- I mean, even some of my students go out to an elementary school, they do the first league, Lego league, they do Lego-- I mean, there's a lot of activities. There's a science museum for kids, hands on. I mean, there's a lot of stuff, but the pipeline seems to be still very, very leaky, right? Even though we have reached more of the young girls, at some point-- and one thing that I think is even more scary, we complain in academia but there are all these NSF studies-- they've been out for a year or so-- that show that in the workforce-- we're talking about computer science, not robotics in particular. If you look at the workforce in mid-career, the number of women who quit their jobs-- much, much higher than the number of men who quit the job, which means even if we get enough of them, they get there, they get to mid-career, then they go. So there is still clear issues there, right? Because you're not saying just produce more students and then everything will work, because they keep on getting lost along the pipeline. So it's very difficult. So.
Šabanović:
Were there any other people who were doing robotics at the time that you went there?
Gini:
No. No. There was one faculty doing computer vision. There was nobody else doing robotics. And then-- in fact, nobody else doing AI. So he was doing AI in computer vision. Then maybe 10 years later or so we hired a faculty doing robotics. Now we have a good robotics group with four or five faculty. It depends exactly how we classify people. So now in fact I think we are the largest place in the college doing robotics. So they're all in computer science, while in other schools they tend to be more in mechanical engineering. There are a few in mechanical engineering, there are a few in electrical engineering but more on the theory side, more on control theory, but the ones that do-- there's a couple in aerospace-- but we have the largest contingent of faculty doing robotics, which is a little bit unusual.
Šabanović:
What kind of projects did you start working on when you got to Minnesota?
Gini:
I was still working for a while on the programming languages, and the first project I worked on, first PhD student, I was interested in doing kind of error detection and error recovery in assembly programs. Again, at the time it was more manipulation and assembly. So what we're doing is try to extend the programs with kind of finite state machines that could detect, and when you detected an error had some kind of procedures for generating a patch for the error. And at the time I worked, there was a professor in mechanical engineering who had-- was it IBM arm? I can't even remember the brand. This was a long, long time ago. So I had the physical robot that I was working with in his lab, and so that was the first of the projects I was working on.
Šabanović:
What was his name?
Gini:
The professor was Max Donath. He is still at the U, but a few years he started-- he moved-- I mean, he was doing robotics, and now he's working on transportation problems. So he's doing things like automatic guidance for truck drivers or for buses. So he's been a director of the Transportation Center for maybe 15, 20 years. So he's no longer really a robotics person.
Šabanović:
What kind of projects did you do with the robots?
Gini:
Again, with that robot, the main thing we have done is, again, this-- we implemented a system, we did experiments to recover from a bunch of errors in an automated way. I think it was too early-- I mean, I tend to like-- my work tends to be really a different year [ph?]. I don't like things where other people have already done stuff. And so I think at the time-- I mean, the detection of error and recovery is a very, very tough problem, and there's not been a lot of workout, so we really-- we were very, very early trying to-- this idea that you can get-- giving the program, in a sense-- get the program extended automatically so that it can detect errors and fix them, and obviously it was too ambitious. I mean, it worked well in the specific examples we did, but it was very difficult then to broaden it and extend it and make it much more general. So then I think in the-- at least in the AI community, people started no longer working on manipulator arms and they started doing mobile robotics. So that was the next things I've done. I bought at the time-- the first mobile robots that you could buy for a reasonable price was the Lab Mate. I'm sure you have seen it, right? So we bought our Lab Mate. We put all our sensors and all our computers and tons of wires and stuff. So we started working on the program. So we've done navigation, obstacle avoidance, navigation in environments like with moving people. We've done things like finding a door and closing a door because with only the sonar sensors, there's all the noise and errors and all this kind of stuff. So we've done, again, robotics programming but on mobile robots, which that's what became popular, right? Until a few years ago. Now manipulation is coming back, even in the computer science community, but for many years only navigation, so.
Šabanović:
What are some of the projects or results that you were most proud of?
Gini:
Well, I think, again, the program system I had developed at Stanford-- again, even though it was very old, I think it was really cool and very new. I mean, I think it was a nice idea and it was, again, a combination of I came-- because I was not a computer scientist, so again, I didn't think about compilers. I didn't think about-- and I used LISP because I learned LISP, and I loved it. And so came in a sense from a different community that was using these interactive ways of programming, and so it was a combination of those ideas that I think produced something which is very cool. The work I've done in error detection and recovery I think was very interesting, but again didn't-- and then at some point I got an NSF grant, did some work, but then it was really difficult to extend it so kind of stopped there. The navigation work, the navigation with moving obstacles was quite early. I mean, it was done, again, a long, long time ago, and we had done it with the Lab Mate, so with real experiments. Then we done some work with parallel arms for motion planning. Because I did the-- I went back to Stanford twice, two years, separate years, for sabbatical, and so the first time I went back for sabbatical was at a time when Jean-Claude Latombe-- I worked with him at the time-- came up with- - had started-- had not yet written his motion planning book but was writing the book at the time and he came up with these different algorithms for motion planning. And so I thought the algorithms were very cool, even though, again, computation complexity was very high. So when I got back, we decided to try parallelizing those algorithms, because they're very easy to parallelize. So had student, had another faculty to do parallel algorithms, so we did a lot of simulation work. We didn't do any real work with real robots.
So we did a bunch of that for a while and we, again, produced a bunch of simulations and different algorithms. So then I got bored. Now, other people-- in fact, Nancy-- I mean, one reason why I met-- I knew Nancy is because she also started to realize, she say, "Those are the embarrassing parallel problems." Which is true. I mean, you just run the same stuff many places. Whoever comes up with the answer first tells you, right? So very, very simple. But there was nobody else really doing parallel algorithms for motion planning at the time. And then, again, I kind of got bored and I said-- change and do other things, and I started working more in multi-agent area, so not robotics. So we done kind of agent-based system in which agents can submit bids for tasks. So we started looking at task location, which is something I still do today, but not in robotics, so there's no physical location but it's only things you have to do, but we're looking at tasks like, let's say, construction task or so on in which there were temporal constraints. And again, that's very new. Nobody was doing those things.
And so we developed a bunch of algorithms to develop a system-- again, prototype and all those things-- for how to run auctions, again, with those temporal constraints. That run for a few years, and then we extend it to different kind of things, and then I've done-- more recently we have done algorithms for the trading agent competition, supply chain management, and so they are more machine learning, prediction methods. We have been in a competition for many years. We got third place one year, and then we said, "Okay, this is getting boring." So we kind of dropped out. But this has been also a lot of work, and in fact I have-- one of my PhD students won-- we won a couple of years ago an INFORMS award, and his work-- his idea was how you predict. So the idea came out of the supply chain management competition but really is applied to business. So now if you have-- you see sales, you see demand, so can you predict what will happen next, which is the difficult part with <inaudible>. And so he came up with a method based on mixture of ____________ Gaussians and so forth making predictions, which is very good and, again, got a number of publications, we got this award, and he's now a faculty in the business school at Rotterdam and just got promoted for professor very, very recently. I went to his promotion ceremony last summer because, wow, this is-- they do things in a very cool way, not like here. "I got promoted." Okay. Oh, they say, "Oh, congratulations. You are done." There, there's a big ceremony, everybody's invited. The person gives a speech. There's an academic procession. I mean, all this stuff which is kind of fun. <chuckles>
Šabanović:
And they bang a--
Gini:
Yes, exactly, exactly. Yes. Yes. Yeah.
Šabanović:
They make the PhDs lay down on the floor and--
Gini:
Yes. Yeah. No, no, no. No, they don't do that. <laughs> But they do that even for PhD defenses. I mean, I was at a defense of a PhD student and, again, I was surprised. I mean, here, for us PhD defense, "Okay, what's next?" <chuckles> So there, it's very serious business. Yeah, yeah. So, again, this was kind of more recent work, and then more recently I've been doing-- I'm doing a lot of work in the-- I kind of like the competitions-- not too much of the competitions, but because it's kind of benchmark, like allows to compare my results with other people, and I think even though I dislike when I don't do as well as others, at the same time I think it's important for science. So I've been working on the RoboCup, search and rescue-- the agents competition. We competed once two years ago. No, twice. But otherwise use as a tool, we use as a benchmark, and one of the nice things about it is you have the results from the competition, so we can run our program and then compare. We don't know exactly how the others manage to win, but you can see the results that they got. And so, again, it's one of the things I always insist with my students is you have to compare, and not just compare with a stupid method but you have to compare with the best method.
And so I've been doing a lot of work in the area. We've done things on team formation and how, depending on how you form the teams-- this is a situation which is very-- I think it's a very interesting competition because there's a lot of uncertainty and a lot of lack of knowledge, and there's also limited time and limited communication. So it's really difficult. All the kind of things that in general you need for your algorithms-- lots of communication-- I need to know where things are-- and so there you don't, and so you have to kind of find ways around. And so, again, we've done work in forming different types of teams and see how much they help, and we've done work in pieces that I presented here briefly, which is-- one of the issues that comes up in the search and rescues as a general problem is there are some tasks that have a cost that grows with time. Typically fires, in the RoboCup. But you can think in general-- invasive species-- there are a lot of other real-world phenomena in which if you don't stop the spread, it spreads, and then becomes unstoppable. Think about the Ebola virus, just as a different example, but it's a very similar issue. So, and the interesting part is that in the computer science community, if I look at all the work on task location, all the tasks have a fixed cost, right? And then you just decide, "Well, okay." So we haven't seen-- I mean, there are very, very few exceptions. So we kind of started looking to this issue again. If I have some kind of growth function and if I have-- I know the function, I can estimate a function, how do I do my allocation to try to optimize the situation? So again, this is sort of recent-- very recent work. I still have to do much more. It's still a beginning. So it's still interesting to me. <chuckles>
Collaboration
Šabanović:
Who have been some of your collaborators?
Gini:
So in general, I like to-- I tend to be-- I mean, I think, if I kind of go back and look-- in high school in Italy you can study different types. I mean, you can take different types of high school. So the study I done were kind of classical studies, in which I didn't do-- I mean, I did more science and math and physics than any high school does in the United States, but compared to the scientific studies, it was much lighter, but I did a lot of philosophy study and all these things, and I think that kind of gave me more-- I'm more interested, curious about things outside my field, which I think, again, comes from the education. So I like always to have a co-advisor for my students, because as a faculty I don't have a lot of time to do things myself. But the students, they do the work and then they collaborate. So I had co-advisors in psychology. I had recently a student who did a study on how people represent space and time in games. So he looked at the tower defense games. And so the co-advisor was a psychology person because we're doing a lot of human subject studies. I've had other co-advisors in math, or in statistics. So typically I don't do a lot of work myself directly, but through the students I really interact-- other faculty, psychology, I've done a lot of work with.
Šabanović:
Are there any particular ones that you’ve worked with more often?
Gini:
That’s a good question. I mean, I tend to change again because I tend to be very driven by students. My approach, which not all the students like, is allowing them to figure out what they want to do. I don’t give them a project. I don’t say, “Do this.” There are many ways of doing PhDs and that’s kinda my style. I tell them up front and so, when they come up with something, then I’ll help them connect. I have a student right now. She’s just started working on this project, so with the faculty in the child development who’s working with the very young autistic children. And so now we’re looking at-- is try to use the robots <inaudible> now is the humanoids to help diagnose attention deficit disorders in very young children. No experiments have been done yet. It’s just the hypothesis, just some theories. Again, this is a faculty in the Child Development Institute. It’s a very strong institute in Minnesota and he has all the connections. And so hopefully they’re going to start experiment with kids. And, again, I didn’t know anything about autism. I mean, I kind of read things in school but the student’s kind of interest-- she’s a she and you can imagine that women tend to like more these kinda projects. But really, in a sense, she came up with this on her own. I didn’t tell her, “Look at this.” We talked, discussed many ideas, then we got this connection. I say, “Okay, I help you find any connection,” and now things are working. We are hoping to get a proposal out and try to get funding. So, again, those things, to me, I don’t plan much in advance. I tend to be more irresponsible maybe. I don’t know, jump into things.
Šabanović:
Have you focused on any point on particular applications. You mentioned this is on autism or...
Gini:
This is a first kind of autism or medical application. I’ve done, again, some works of the search and rescue, like this robotics and <inaudible>. Because also we have an NSF Industry Center around search and rescue. So I said, “Oh, yeah, those programs are cool and I can do my task or location in that context.” Again, earlier I was a little more doing manufacturing or manipulation tasks but I haven’t done those in a long time. And a lot of navigation in unstructured environments. Again, I don’t like in general. You simplify an environment so much and then you come up with your nice and clean algorithms. I kind of like there are parts of the field and so, again, we’re looking, “Well, if I don’t know where things are”-- right now we’re doing-- a lotta student are doing a exploration, like with many robots if a buildings could be partially collapsed. And so-- and try to-- which we don’t know the map. We don’t know anything about the building and try to have the small robots to go around as a team and explore and stay within communication range of each other so they can all come out at the end. So the idea is you have-- in a building, you drop all your bucket of robots at the entrance, and then they self-organize. They go around and find things. So we have a few algorithms. We’re still trying to get some more experimental work. But, again, I like this, when things are not known. It’s not clear. I don’t like, in general, coming up with an optimal solution, clean things up. I like rough stuff. It’s kind of rough <inaudible>part in general so that’s more my style of work.
AI
Šabanović:
So as a computer scientist is there any interest in A.I.? What would you say robotics-- I mean, does robotics in a sense bring the rough stuff into the area?
Gini:
I think <inaudible> A.I. there’s a lot of data. I mean, I think clearly in a sense it’s more the style of work or style of reason, or the interests. There are people are more interested in coming up with very precise algorithms or proofs, some which are impor-- I mean, I don’t deny their importance but it’s not the style that I do. I like to do experimental work. So even if it’s not robotics, and we did the-- again, this agent that was submitting bids for auctions, we implemented a system. So we didn’t just prove the properties of Soviet design of the system and implement it with some users using it. Design user interface, design all the algorithms with all these experiments. So it’s more, again, proving things experimentally rather than theoretically. That’s kind of the most of my work. And it’s a lot of computer science. I mean, I think, even though traditionally computer science used to be more, again, algorithms and proofs, I think is more and more interest in-- there’s a lot of interest right now, for instance, in Markov random trees, or random trees, or Markov decision processes, or what’s called MCT, Markov Chain Tree. So used for games. I mean, there is a lot of interest in coming up with answers through computation, and computations are not directly exactly what you want to do but are more you explore the space of choices. In fact, I have a student I was working on the game of “Go” which I don’t know how to play but, again, he’s an expert. So I say, “Okay.” We kinda look at the all the signs. And, again, same thing in part we’re looking at using user knowledge and now we’re trying to see running on pilot computers. Can we improve, come up with better moves just because we do a better random search? So I think computer science, again, with the random search coming into the picture is more willing to accept that empirical results are valid results. You don’t need to prove theorems all the times, and that’s kinda what I like personally also. So, yeah.
Šabanović:
What do you see as some of the future challenges, both for you and for these fields in general, robotics or A.I.?
Gini:
Well, robotics, I think, the challenge really-- I think there’s still a lot of open issues and there’s a dream of making these real intelligent robots. And I think we’re still very far because the robots still-- I mean, there are more results in robots going on the web and finding knowledge and so. But, honestly, when you look at, they’re still narrow. You pick up a narrow domain. Cooking, like the work of Michael Betts [ph?] and his group, which is very cool. I’ve seen demos and so. I mean, you can cook here omelette and all this stuff, but I’m sure that if it want to do-- cook a pasta, you have to come up with your knowledge and rules and other things. So, I mean, there are more a proof of the concept is doable. But, going from an instance, or a small set, to a really broad things, like what people can do, I think there’s still a lot of big gap has to be filled. And I don’t know that adjusting or expanding and say, “Okay, now we make the robots small, capable of go on the web and finding information.” And so I don’t know that that’s would really be enough or not. I mean, I hope so but I don’t know. I mean, seem to me there’s still something in the human intelligence that drives more with the exploration and is not clear how we can kind of, in a sense, encode that into the programs or so. I mean, if we look at how kids learn, which is the always fundamental thing and there’s a lot of exploration. And I know people-- and I know the group in Europe, they use the icob [ph?] and you kind of manipulate stuff, explore and try to learn so there’re good results. But can it really be expanded to general things?
Again, each time I see a narrow topic-- which is the only thing you can do in research. You can’t solve a big problem in one step. But how would we be able to expand to more general things? That’s I’m not sure. I mean, I’m sure we will get there at some point. I’m not pessimistic but I think it still take a long time. On the other side, I think with the new emphasis on robotics, is <inaudible> speaks of this all the time. I mean, the robots doesn’t need to do everything. You can ask for help. So it’s more these robots and people mixing together. And I think that’s also interesting and that’s the first things that will work because, again, you are not trying to make a robot that can do everything but you’re kind of guiding. And once you figure out exactly how to get the humans and the robots to interact, what I suspect will end up happening in a commercial development is what happens with computers. If you think about with computers, you go to a airport to check in or something or other and there’s a computer. “Oh, I cannot do this. The computer doesn’t let me do.” I mean, in lots of cases, we end up-- I mean, the commercial developments say, “Okay, there’s this stuff I don’t know how to do. It’s not allowed.” And I suspect with robots I think there will be similar things. There will be things in which, if we get to the point where we can buy robots and get the robots to go around in your house and do things, there will be a lot of stuff we’ll have to-- I think we will have to adapt to robots more than the robots will have to adapt to us, in a sense like computers. I mean, we can do lots of stuff with computers but, in a sense we have-- in many case we have to adapt to what the computer allows me to do as opposed to, say, this is the way I want to do a thing. Computer do the stuff for me.
And I think it was a tradeoff between, again, how much we’re willing to give things up because of convenience or because of other things and watch really-- we should insist and say, “No. These things are to be done the way I want it.” Because, again, think about-- I mean, lot of simple thigs, like every time you do automation. So I’ve been coming back-- I go to Italy often. So coming back now to the United States, most of the airports now have these computers where you fill in your form. The form used to be written on a piece of paper. Now you fill in the computer form. So, when I come back to the U.S., I know exactly what the rules are so I tend to follow the rules. I really don’t mind. So I never bring back meat or vegetables and so but I always bring back the special parmesan cheese, which you dine Italy. It’s very nice. I bring back some pasta, special things. I mean, and I know exactly what is allowed. So, in the old system, every time I come back with my husband, we get on a form, we say, “Food,” and we write on the back and say, “Spaghetti,” I don’t know. “Parmesan cheese, pantone [ph?]” when we come back after Christmas, whatever. Right there and you give to the officer when you get out and read. Say, “Oh, no sausage?” “No sausage. I never bring sausage. I never bring meat. I never bring vegetables. I know what the rules are.” Say, “Okay, go.” Now, with the automated system, it doesn’t work anymore, right. <Inaudible> you say, “Yes, I have some food,” it prints a big X on your form or a big bar on your form and you’re forced to go to a agricultural officer.
And then they look at and then they decide to explain what it is and then they decide whether they want the x-ray or not. I mean, it’s not a big deal but, in a sense, again, is, when people are the loop, they can make decisions often on things-- it’s their judgment. I mean, of course, they could say, “Well, we’ll if we are letting the officer judge, maybe you look very nice and friendly and sincere and you’re not.” I mean, they have the dogs to check things and so-- I mean, there is a potential for human error but, at the same time, there’s the-- I think there’s more-- I mean, you can make decisions from more intelligent. Let’s put that way. And, if you put in the system, obviously the system cannot judge, and so they system say, “Okay, send to the person.” Again, simple things, but automation very often brings us annoyance. It’s because it takes away the ability of the individual to make decisions and exceptions, so. And I think there was an article on the “New York Times Magazines” a couple of weeks ago, which-- I gave a talk to high school kids a couple of weeks ago and it was just came out the day before. And this article-- I don’t know if you have seen it-- and it’s kind of starts saying there is an unknown name of a person. She’s at home. She’s in pain. She has this robot help her and calls the robot and say, “Give me some pain medication.” Now there’s a problem, because the wireless is down today in her house and so the robot cannot give this pain medication, without approval from a supervisor. Because the wireless system is down, it cannot do anything.
So what should robots do? It’s kinda to bring sort of the ethical dilemma. I mean, do we allow robots to make decisions, if so to what point? And so-- and in that sense, I think it’s again is a similar example. So what they’re saying there, “Well, you need the robots with common sense or capable of feeling emotions.” But to me the issue’s really the issue of autonomy. If you have a robot that makes decisions, what’s the level of autonomy? The other example that they mentioned is the classical example. A car, you’re Google car, and you run into intersection and there is a kid in the middle of intersection, what do you do? I mean, if you are a person, split decision, decide whatever. If it’s a car, what is the car going to do? And there are-- a lot of this has been out in the community for a long time and the question there is really a difficult question because do you want your robot to make decisions that are ethical or not? I mean, we do this all the time, judge, make mistakes, and so. And I think this is still, to me-- I don’t think the question is really make the robot ethical. I mean, you can make a robot ethical, give rules and so, but I don’t think we act with raw source. That is a lot of decisions that we make in a split second. So if somebody asked me, “What’s your rule?,” maybe I’ll say, “I’ll ignore the kid and run over the kid,” or whatever, or “I change my direction.”
But in real life, maybe I get scared and in fact I killed a kid because I wanted to save my-- I don’t know what I going to do when you get to this situation unless we experiment and then we don’t know. So what could I put, even if I could write a program for the robot? What could I put? And that’s we still-- I don’t know. I don’t have a good answer. I don’t think the, again, making the robots emotional and more us is-- I don’t think it’s the answer. But I don’t know there’s an answer but, as we go more and more to automation, we need to figure out what to do. And that’s-- that, I think, is a kind of scary part in a sense. I don’t know. Maybe somebody will figure it out. I mean, I know there’s a lot of debate in the A.I. community on lots of things. I’m discussing them with my A.I. class. I’m teaching undergraduate class on A.I. so first assignment I gave to them is read some of these-- the recent things that’ve been in the news for the last couple of months. Read what arguments are discussed and then we can kinda to-- at least to understand that there are issues. I don’t have answers. I don’t think anybody has answers, and robotics clearly, I think, is worse than computer science. Because, again, if you think about your car, which is going to drive on its own-- I mean, the fact that robots have a physical body, can do physical actions is worse than a computer. I mean, computer can kill off your internet, destroy your programs so destroy whatever-- or said bad things about you or send out you-- I mean, can do bad things, but still the physical actions that the robots can do can be more damaging than just what the computer can do. So I think it’s a little more scary.
And the robotics, the autonomous car driving, I think is the next-- I mean, it’s coming. There’s no doubt. And it’s a more scary things. I mean, the other things have been a lot of debate in the A.I. community. I don’t know. I haven’t heard much in the robotics community on whether robots should be allowed to kill. Again, I haven’t really-- maybe the A.I. people like more to debate, have more philosophers so they’re-- there’s a <inaudible> I think a couple of years ago. Each car was <inaudible> other people, lots of other people. Again, say, what are the limits and there is-- I think that’s another important thing because right now, despite the fact that the drones are controlled by people-- I mean, people make mistakes but I’m-- I don’t know. I am convinced that there are drones, the fire, they drop bombs on their own. I may be wrong. Again, I’m not an expert. I never really try to read but I’m sure this is already happening, either in, again, intentional or by mistake. And nobody’s really discussed it and any of this issue of killing people-- again, this debate is at a A.I. conference. There is this Geneva convention. I mean, serious stuff.
I mean, it’s kinda under the radar because nobody want really to talk about. But it’s something that I think should come up and it should be specific rules. Because, if technology can do whatever we want, but there are things in which we can put limits. I mean, there can be laws and <inaudible> set the limits. And I think I agree with the people that say should never have a robot that fires, that kills a person. I mean, there should be a ways a person makes a decision. It could be wrong, but at least it’s a person. Because I don’t think you can come up with-- despite all this sophisticated programing, with good rules for the robot to decide. And that’s-- again, I have not-- I tend to go more to the A.I. conferences and, again, in the A.I. community’s there’s a lot of debate on this. I don’t know if there’s as much in the robotics community. I think as a voting, small things, in which will be really nice if the community could get together and really push and come up with regulations that controlled it but...
Šabanović:
That’s the area I work in.
Gini:
Oh, you work in-- oh, so you know much more than I do. See I am ignorant but there are things I find fascinating. It doesn’t mean I know much about. But okay. Okay, yeah.
Šabanović:
<Inaudible> “New York Times” article <inaudible> on that.
Gini:
Oh, okay. Oh, okay. Yeah, yeah. I mean, again, I never thought much about. There was this panel at Hichkai [ph?] two years ago and there were a number of people talking about different things. And I say, “Wow, okay. I never really talked about but now, soon to be big problem and soon to be likely to be an obvious solution.” I mean, <inaudible> people don’t like to separate, right. Soon to be like clear cut. So you can say, “Not allowed. End of the story.” But there’s no political will, I think, to do it so. Yeah, I’d love to talk with you.
Šabanović:
<inaudible>
Gini:
Yeah. Yes, yes, absolutely. Yes, yes, yeah.
Funding
Šabanović:
Just to kind of catch this before I forget. We were also curious, where has some of your funding come from? Where do you get it?
Gini:
Most of our funding is from the National Science Foundation. We had, with a good project a few years ago, maybe 10 years ago, we got the DARPA grant, which in fact is-- we end up designing a very small robot which is called the Scout, which is also commercial robot. And the main idea there was a way to design a robots for exploration. So it’s very small like a soda can. There’s a camera. It’s remote control. This was funded by DARPA. That’s the only DARPA grant I ever had. Otherwise it’s all National Science Foundation. Some of my students have fellowships, Fulbright’s or other things. Some work as T.A.s <inaudible> you do what you want. You are a T.A., you do your research as you want. Otherwise, they have to kinda force you to do things. So, but otherwise-- yeah, and a few other smaller local grants, yeah.
Šabanović:
You mentioned education a few times while you were talking about it. One of the questions we like to wrap up with is what is some of the advice that you might give to people who are students or young people who are interested in getting into robotics?
Gini:
Well, in general, I used to tell all the students, “Study robotics but don’t expect a job in robotics.” Till a few years ago, there were really no jobs. So there were a lot of undergraduates who were very interested and so I say, “Study because you learn to deal with the real hardware, which most computer scientists don’t do much about. You learn to solve interesting problems so it gives you a lot of skills. But don’t expect to get a job in robotics.” Now, I started changing because now there are lots of jobs in robotics. So, in general, what I would say is that, for me is you have to follow your passion. You have to figure out, in a realistic way, how to combine what you love, what can feed you in the long-term. So, I mean, I know a lot of undergraduates just follow the passion and have no clue they’ll have no jobs in the end of the line. But I think you can combine things. So combine what you love with something that has some commercial potential, but follow what you want. I mean, what I always tell my students is, “You want to get a job that you love to go to work every day.” I mean, not everybody’s so lucky and I understand some of us are lucky. Because we do what we like to do basically. Nothing robot can really do but still, if you do a job just because you know we’re gonna get a lotta jobs and make money you hate, you don’t have any sort of a life. If they love robotics, again, that’s-- if they don’t get a job in robotics, there a lot of other jobs. I mean, lots of real-time systems, programming, depending on the aspects that they work on. There are lots of opportunities. Again, not maybe exactly the narrow things they are thinking about. So I had an undergraduate-- most undergraduates like to write games for game companies. And then they got very, very bright some years ago and they worked on this search and rescue. So it’s kinda doing the simulation. He got a job with a game company in San Francisco and, when I talked with their recruiter, she was saying, “Oh, a lot of the students that we interview are just really good at games. They really don’t have the backgrounds. They don’t really understand operating systems. They don’t really understand how to do memory management.” And so and I told her, “You have to look for real computer scientists, not game developer students coming out from those schools. Because the skills that they need are real solid, central computer science skills: algorithms, again, real-time systems <inaudible>.”
Šabanović:
What you were saying about <inaudible>.
Gini:
So what I’m telling them, in general I say, “If you want-- let’s say you want to develop games, say. Obviously, you can play and develop a few things, just examples. But don’t focus on game development. Focus on more the fundamental concept in computer science. Very good programming skills but, again, the theoretical skills: algorithms, complexity, data structures, real-time systems. And try to be the best in what you are doing.” Because each job now is kinda specialized. It requires specific skills and subset. And fundamental knowledge of computer science, and be very good at programming open a lot of doors. And if you focus too much on the end application, often you miss. Right now, one thing I’m a little concern-- I mean, it’s not much robotics but a lot of women now are in HCI because HCI, human computer interaction, is a little more people friendly. You do work with people. You look at real problems. You think you have a bigger impact on the world, and a lot of HCI. But I’m concerned I see some women that are really-- technically are not very strong and they go into HCI because they think it’s a good field. Now, it’s very difficult now, I know, to find jobs in HCI if you’re not very strong. And so I’m always concerned by people that get-- and the games are the same thing. You cannot see the end product so you say, “Oh, I can do it,” and you don’t understand that whatever you want to do, you have really to do it very, very well. I mean, if you want really to get in a field and get good jobs, you have to be the best as you can. I mean, you can’t just be mediocre. And there’re a lot of people that kinda, again, move around and then they find themselves in trouble ‘cause they don’t find the job that they dream. And that’s always a bad thing. So, again, try to focus on a few things fundamental and do them really well. Then lots of doors open. And people, again, try to go too far too soon, look at the end product instead of building the skills. And robotics, the same thing. It’s not just that you can build your own robots and do things. I mean, you have to build a lot of technical skills. That those are the ones that the companies that want to give you jobs.
Šabanović:
Well, thank you.
Gini:
Okay, sure.
Šabanović:
Well, I have-- can I ask a couple follow up questions? Mm-hm.. yeah!
Gini:
Sure, sure, sure.
Šabanović:
So maybe it’s two parts but I’ll start with one part. What was your sister’s name and is she still involved in robotics?
Gini:
Oh, sure. So my sister name is Giuseppina and sh-- in _____ is called “pina” and Giuseppina’s one of those Italian names that the Americans always misspell. <off topic conversation>
Gini:
So she’s working in Italy. She’s still at the Polytechnico. She’s married. She has two children. So we kinda split when we got married. She’s still-- she’s working A.I. and some robotics. I don’t talk with her much about her research anymore. I mean, when we get together, we talk about other stuff, but she’s done-- Polytechnico is a top school for the students who are very good. She’s done a number of things. They were building robots, writing software, but she also does a lot of-- her husband is a chemist and works for a research institute in Italy. So they do a lot of study of chemicals, impact of chemicals on environment. And so she does some work also with her husband, do some kinda machine learning, apply multiple chemical discovery and things. But, yeah, she’s still active but we haven’t published papers in a long, long time together, so.
Robotics in Italy
Šabanović:
Okay. My other question was you said there wasn’t a lot of robotics when you first came to Stanford and left. How have you seen robotics developing in Italy specifically over that time...
Gini:
Sure, that’s a good question, sure.
Šabanović:
...and the role of women in robotics in Italy?
Gini:
So in Italy-- again, when I got back to Italy, at some point-- again, robotics started in different countries in different places. So in Italy there’s a company, Olivetti, which is kind of dead now but it used to be a top computer company before they kind of shut down. And so, at some point, I don’t know exactly how it happened but there was couple of people from Olivetti who started developing a robot. It’s called Sigma and it was a robot with manipulator arm. It was for assembly because they were assembling chips or other things. And so there’s two kind of engineers high-level in the company somehow contacted our group in Milano and <speaking Italian> and so we started doing some collaboration. And then, I can’t remember. It was in the early-- middle-- very early '80s or the end of the '70s. So we started a robotics society just called Siri [ph?] which is _____ Ialiana Robotica Industry Outlet, I think is the name. And so-- and there were other groups. So Olivetti was the main one. Then there was Cumow [ph?] which is a subsidiary of Fiat and they’ve been doing-- they’re still in business and do robots for welding and for more car manufacturing. And there were a few other companies in Italy and then there were a bunch of people at different universities. So we kinda started this society of robotics and had some meetings and a few conferences and so. The society still-- I think, it’s still alive today and so kind of started from there. At the time, there were not too, too many other groups. There was a group in Pisa, there was a group in Roma, there was a group in Genoa. I think we were in Polytechnico Milano. Then there was a group in Polytechnico of Torino. Those are the main ones. And then a few other people doing robotics. Now, there’re a lot of robotics work in Italy, even in smaller universities. I mean, I still know a lot of the people and they do very different things. I mean, there’s a group in Pisa, Dario-- what’s his last name? <speaker restarts answer>
Gini:
Yes, Paulo Dario, yes. Yeah. Yes, so in Pisa and they’re a very large group and there’s a group in Roma and there’s a Daniel Inardi [ph?]. In fact, there is one of the person who started this-- I mean, is part of this robot search. I mean, one of the first maps in this robot search is a map of Filino [ph?] which is this place in Umbria, which is in Italy and Daniel Inardi provided it. So they still-- that group does a lot of work. There’s a group in Napoli. They do a lot of control theory. There’s a group in Genoa still very strong model on the cognitive aspects and they do the icob and have a bunch of other project. There’s a group in Padua. A very good friend of mine <inaudible> start-up company and now this is a Verona, is Bologna. Everywhere there are groups of robotics. So now it’s grown a lot. In terms of women, I don’t-- I mean, there are a few women I know in Italy do robotics, not too, too many. And, in general, in computer science, not too, too many. The ones I know-- I mean, I know a few. I think the difference may be with U.S. and Italy. I think in the U.S. is a cultural problem, I mean, very clearly. In Italy, I think maybe because people tend to be a little more social and at the same time, I think, in Italy you split your personal life and your job-- are very, very split. In the U.S. they’re all mixed together. I mean, your friends at home are also the people you work with.
And so, in Italy it’s relatively rare. I mean, you can work with people every day and never go to their house, never know anything about them. So people tend to keep the two worlds a little more split. So I think from that point of view-- and, again, people tend to be a little more used to work in groups and chat and have fun and all this stuff together. And so I think women feel, in general, a little less singled out or isolated, because, again, there’s more a sense of groups than there are here. So, even though, again, there are problems. There are not too many people but I never had these impressions that I have here, that even lots of students have here and really find them uncomfortable in situations. And I never had that in Italy. Things may’ve changed. Again, when I was in Polytechnico, again, there were couple of women faculty there. There’s still some women faculty there. Again, it’s an engineering school. I think the ratio there women to men is higher there. And, in many-- one interesting things in many Mediterranean countries, the ratio of women to men is better than in Northern countries in Europe, which is very counterintuitive. And somebody told me-- and I often I thought about-- and I was in the Netherlands, in <inaudible> for sabbatical a few years ago and so I discovered there-- I was working with the women faculty there and I discovered there the Netherlands is kinda the worst in Europe in terms of ratio women to men. Why? You are in a country where there is a lot of social support, social services.
So why do women don’t go to jobs. And, again, if you go to Norway, Sweden, and so, it’s the same thing. So there’s something very strange that I really don’t fully understand. I was talking last night with a person from Spain, from Barcelona, and she was te-- sort of asking her about women and she would say, “Things have gotten much worse, even in Spain.” Spain used to be another place with lots of women. Manuela Veloso’s from Portugal. There’re a lot of other women. There are two women from Spain here. There are only two from Europe. So used to have good representation. She was telling me that things have changed, again, and in part she said because they moved some of the computer science degrees into engineering schools. And it’s the same thing happen in the U.S. This kind of one of the manufacturers is known is when computer science moved from liberal arts to engineering schools, which is happen in many schools. That’s dropped the number of women.
So I have to talk with her a little more because she was saying basically are very, very few women. There are no new jobs in Spain because of the economic situation. So it’s hard to say whether the few women are around, whether they will get jobs or not. So the jobs market there is kind of-- doesn’t exist. But she was saying it’s a fewer women and the women that they had <inaudible> then get jobs somewhere else in Europe or in the U.S. because, again, there are no jobs in Spain. But I was very surprised, again, to hear because traditionally all the Mediterranean countries for whatever reasons, maybe the warm climate-- who knows? I have no idea why-- have many more women than other places. Turkey still has a lot of women. Greece has good numbers and the former Yugoslavia used to have. Again, as you move north, they drop, which is really, again, really strange and nobody really has any specific explanation. Because, again, if you look from the society point of view, the support structure, yeah, exactly. So it’s weird.
Šabanović:
<inaudible>
Gini:
I don’t know.
Šabanović:
Yeah, well with Germany it’s so strong that they incentivize for women not to go back into the workforce after they have children. But I don’t know that they <inaudible>.
Gini:
Yeah, right, right. No, this-- yeah, yeah. The rest of a-- I man, Netherlands, again, in part some people say-- which may be true, they say, when women have the support-- because, again, you can have-- you don’t have to work, many still prefer not to work. I mean, it’s true that lots of women prefer to stay home with their children and, in some cases, you cannot because you don’t make enough money so you’re kind of forced to go to work. Which I think is more in some Mediterranean countries or again it’s not as much money so you really have to work. So it could be that it’s a real choice that people make. I know in Germany, again, when people take time off, it’s very hard to get back into the workforce. But so it’s kind of interesting to figure out what does it take, again, to get women to be interested in being in the workforce. And, again, it’s not necessarily social services, again, which is the first thing that people think about. So I-- again, it’s very complicated.
Šabanović:
Lots of open questions.
Gini:
Absolutely, yeah. That’s <inaudible>.
Cybernetics
Šabanović:
One more follow-up. You mentioned that one of your first inspirations was a cybernetics paper from <inaudible> but was that a cybernetics class? What role did cybernetics have in <inaudible> the designs <inaudible>.
Gini:
It was kind of a class on the-- because at the time, again, there was no-- I mean there were a few schools doing informatics. I think there was one in Pisa and other two. There was nothing in Milano and so they-- in physics, the professor I was working with, he was very interested-- then he moved to informatics when they opened informatics. So he was very interested in teaching a little bit more computer science if you want. Again, this is long time ago. And so there was a-- it was a kind of a seminar class or a special class that we did, in which we had a lot of readings. I don’t remember exactly how I signed up or how I ended up there. And we were doing basically auto <inaudible> theory, which again was not covered in any of the regular classes. And this kind of-- and the faculty who was teaching the class was very interested in these kind of issues so she- - so that-- I would not say she was an A.I. person but it’s just-- again, this, we’re talking about the early '70s. It was kinda bringing in, again, some of these idea from cybernetics that were very popular at the time and kinda bringing them in. But, again, there was nobody doing A.I. at the time there. So I think sometimes <inaudible>-- so I also tell the students, “You know, sometimes do something strange or unusual.” I mean, you never know. It’s what our V.P. for research call serendipity. That was the best word. I think it’s the best word everywhere. But it’s true that the randomness, even in real life, you run into things or happen to discover something. And, if you’re always very goal driven, very focused, you miss those potential opportunities. So yeah. Yeah.
Šabanović:
Okay.
Gini:
Okay?
Šabanović:
Thank you! Thank you very much!
Gini:
Okay, sure.
