Goren Gordon | Curiosity Across Disciplines

Computing Across Disciplines – ACM FCA

Goren Gordon directs the Curiosity Lab at Tel Aviv University. His background and approach is truly multi-disciplinary: Goren holds six academic degrees, including two PhDs, but none of them are strictly in computer science. We talk about his journey from quantum physics to social robotics, how he chooses which ideas to pursue, and how he advises his current students as they navigate interdisciplinary research. He also shares some great summer reading recommendations.

Bio

Goren has a BMSc (bachelor in medical sciences), MSc in physics and MBA from Tel-Aviv University, Israel. In Weizmann Institute of Science he finished one PhD under the supervision of Prof. Gershon Kurizki, on dynamical quantum decoherence control, a way to facilitate the emergence of quantum computers; and another PhD under the supervision of Prof. Ehud Ahissar, where he developed and extended mathematical models of curiosity, analyzed animal’s behaviors with it and implemented it in robots that learned about their own body, similar to infants. He did his postdoc with Prof. Cynthia Breazeal at the Personal Robots Group in the Media Lab, MIT, where he studied how curious robots interact with curious children. He discovered that curiosity can be “contagious”, i.e. children playing with a curios robot became more curious themselves. Goren now heads the Curiosity Lab in Tel-Aviv University, where he develops state-of-the-art computational models of curiosity; quantitative assessment tools for curiosity; and curious social robots that learn about other agents in their environment, all by themselves. Goren is also interested in scientific education and he has a Teaching Certificate from MIT. He develops quantum computer games that teach quantum physics to children via play as well as gives popular lectures on quantum physics, the brain and inter-disciplinary thinking.

Transcript

Andrew Miller: So first of all, I want to thank you so much for talking with us today, and the first question I’m asking all the guests on this podcast is, actually for most guests I’m asking, what first got you interested in computing? But I think for you, it might be a better question to ask what first got you interested in science in general and research in general?

Goren Gordon: So actually, I started really early and everything that was in the beginning was actually in science fiction books. I can, all my life since I remember, like from the age of 12 or 13, I had these dual passions. One of them is for neuroscience and the brain, and that was from “Fantastic Voyage Two” of Asimov, which everybody knows “Fantastic Voyage” one when people shrink and they go into the body, but he actually didn’t like that book. He wrote “Fantastic Voyage Two,” and they shrunk and went inside the brain. That’s an amazing book that started my passion about understanding the brain and how we think and what goes on there.

Goren Gordon: The other way is I read “The Emperor’s New Mind,” by Roger Penrose, which it’s a very heavy book but it really explains well quantum physics. It’s just so weird and counterintuitive that I had to understand them, so throughout my entire research career I always was interested in both of them, and I always, it was very hard for me to decide which one I loved more, so I researched both of them most of the time.

Goren Gordon: For me, one of the major goals is to actually make science fiction nonfiction, because there are so many ideas and so many crazy stuff out there that slowly becomes reality. It was very important for me to understand what’s going on and what are the barriers to actually make it work? That’s what drove my passion throughout the years.

Andrew Miller: Looking at your journey so far, I can really see those two trends. You call yourself someone with a triple double, right?

Goren Gordon: Yeah.

Andrew Miller: You have two undergraduates, one in medicine and one in physics, two master’s, one in physics and one in business, and then two PhDs, one in physics and neuroscience. Did I get that right?

Goren Gordon: Yeah, I think that’s right. What people don’t know, that I did all of them in parallel. People usually ask me, “Why did you quit physics and went to do neuroscience?” But almost throughout my entire career, I did two degrees in the same time. Each one of them took me a little bit longer, but I literally did my first two undergrads together, and also my two master’s together, and the PhD, I finished my first PhD when I already finished all the courses for the neuroscience. I did all the courses for the neuroscience PhD while I was doing the quantum physics PhD. Throughout my entire career, I always was interested in both of them, so it’s not like I jumped from one to another, but it was like really combining them or studying both of them.

Andrew Miller: Yeah, one question that I think a lot of people would have is how do you sustain motivation to pursue degrees in such seemingly disparate areas? Did you try and connect them together, like did you try and do it in one project that would work for both?

Goren Gordon: Yeah, so actually when I finished my first PhD in quantum physics, I tried to turn to what the research field called quantum biology, which is trying to find quantum phenomena within the biological system and also the brain. I looked for this connection, but then after doing all the math I realized that I don’t really believe that it works, although it’s a blooming field now.

Goren Gordon: Actually now in my lab, to my amazing surprise and wonder, I actually have two projects that’s combined both neuroscience or machine learning actually, and quantum physics. That’s like a total surprise for me. I didn’t think I would do it again, but I actually came back to quantum physics recently and it’s just wonderful. There’s again, another research field, amazing research field, it’s called quantum cognition which actually attempts to use quantum physics to explain irrational behavior of people. It’s just an amazing thing, because you’re using quantum math and everything that I love about quantum physics, and to actually understand psychology of people, which is for me again, it’s just amazing to be able to somehow combine them both in a research project, which is so cool.

Andrew Miller: Yeah, I mean you sort of touched on this just now. I’m interested to hear, so what is it about physics for example that you like, but what does physics not give you that you need to get from another discipline, like neuroscience?

Goren Gordon: Yeah, so the reason I actually left quantum physics is because for me, it was too unrelated to the day to day world. You know, I solved equations and I understand how you can make things levitate, and how measurement can create amazing phenomena and stuff like that, but in the end you don’t see it, you can’t touch it, it doesn’t do anything in the world.

Goren Gordon: But when you do neuroscience, then you can actually start to understand like the human behavior, and trying to understand that. Also, when I went into neuroscience I in parallel also went to robotics, which is totally in this world and moving in the world, and the actual physical embodiment of all these ideas. I didn’t just went to neuroscience in the sense of understanding the brain, but also trying to recreate it. For me, it was a huge jump from quantum physics to engineering and robotics, and so then you need to do an experiment that fails because something breaks, as opposed to just running equations and stuff like that. It was a big jump, but it was very interesting. That was what was missing for me in theoretical quantum physics, that I wanted to do something in the real, day to day world.

Andrew Miller: Your research at the moment is very robotics-driven. One of the things that I think is interesting for our listeners is that you have all of these degrees, and none of them are officially in computer science, right?

Goren Gordon: Yeah, I have to say something. It’s funny, so I have several degrees, but actually what I’m doing now is like a multidisciplinary approach for four disciplines which I don’t know anything about, which is computer science, engineering, which I don’t have a degree in engineering, developmental psychology and education. That’s like four things that I’m doing now which I have no clue about, and it’s so much fun, it’s unbelievable. I’m learning all the time. I feel like a freshman again, and I love the feeling, although my students I’m not sure like it, because I have no idea what to say to them, because I don’t know anything.

Goren Gordon: From interest perspective, again, everything is totally new for me, which is amazing. In my neuroscience PhD, actually it is computation and neuroscience, so I did all the machine learning and reinforcement learning, and then supervised the neural networks and the [inaudible 00:07:27], so I did all of these computer science things, but without the degree. I took several computer science courses, but it was more oriented to neuroscience.

Goren Gordon: In my post-doc at MIT in the media lab, I did robotics, mainly like really hard-core social robotics, operating robots and doing experiments with robots and people. Now in my lab, I’m trying to combine somehow all of it together, so I have projects which are purely computational, trying to extend some models, and I have projects that are purely engineering, trying to build new robots, and also everything in between, so both social robotics and new algorithms for robots.

Andrew Miller: Let’s definitely talk about the work that you’re doing now. You run the Curiosity Lab. I’m interested to hear about the kind of research you and your students do, maybe if there’s a project that you think would be interesting for us to hear about?

Goren Gordon: Yeah, so I want to say first is that my … If someone asked me, put a gun to my head, “Which discipline are you?” and I had to choose, so I am actually a physicist, even though I’m not doing physics now. Because I think the physics perspective actually gave me this really amazing analytical viewpoint of everything. I feel extremely comfortable with equations and dynamical systems and stuff like that, and this relates to the question because the Curiosity Lab, although curiosity in most people’s mind is actually something psychological, you know that should be in the psychology department?

Goren Gordon: But I’m actually coming to it from a machine learning perspective, and I’m trying to make like equations of curiosity. Everything I’m doing, and some of my projects are actually pure psychology, are actually driven by trying to approach this psychological question from a quantitative model perspective. For example, some of my projects are actually about curiosity assessment of people, so I want to be able to assess whether a person is more curious or less curious, but most of the assessment tools are again pure psychological. You watch a person, you count how many questions he asks.

Goren Gordon: Now, we have also another blooming field called artificial curiosity, or intrinsic motivation, or even developmental robotics, which we actually have equations of a curious agent. So, what is a curious agent? What I’m doing in my lab is actually I’m using these equations to define the optimal curious agents, and then I compare what people do to that optimal agent. I can actually quantify people’s behavior by all of these machine learning algorithms, which is a totally new approach to how to do all of these assessment tools. This is one big part of my lab now.

Goren Gordon: Another part is the social robots piece, and it’s a continuation of what I did in my post-doc at MIT, where we used social robots to actually interact with children. What I did there is actually again a little bit less computational, but more using robots for psychological experiments, is to actually see whether children can become more curious if they interact with a curious robot. The children played with the robot, and the robot behaved like a curious child. It said, “Oh, I love to learn, let’s try new things, feel free to make mistakes” and stuff like that.

Goren Gordon: Then, we measured the curiosity of the child after the interaction. We showed that children are actually more curious after the interaction with a curious robot than after interacting with a non-curious robot. Now, we’re taking this work one step further where we’re actually applying all the machine learning algorithms of curiosity inside the robots. Now, the robot is literally curious and he’s literally learning with the child, because he doesn’t know in the beginning, so he learns with the child.

Goren Gordon: Then, there is still this interaction between the curious child and the curious robot, which I think is really cool because we’re trying to mirror what goes on in the child’s brain, inside the robot using all of these machine learning algorithms. Once we have that, we can have this what we call social companion for the child, when they have a robot that actually grows with the child, and learns with the child, so a long-term interaction.

Andrew Miller: That’s fascinating, I mean you know from a computer scientist’s perspective it’s sort of like you’re working across the whole stack, right? You’re sort of looking at the algorithmic underpinnings, and then trying to hit both sides of the human/robot interaction sort of all within one lab, with this them of curiosity. One thing that occurs to me is, so I’m reaching from one discipline to another, and that other discipline I’m treating as my application area. But in all of this work, you’re sort of crossing disciplines all the time. Does it feel that way to you? Does it feel like, you know, we have this core and then we’re bringing in other people or we’re learning new specialties as we need them?

Goren Gordon: My biggest challenge when I opened the lab and also when I started to work at MIT is that I’m not a psychologist. While I’m trying to think all the time about how we think and how we learn using all these equations, there is different methodology for psychology research, and even more so to write psychological papers. I had to learn this basic jargon of this whole discipline, which is totally different from what we’re using in computer science and engineering, which is very hard.

Goren Gordon: I surrounded myself with psychologists. It’s weird for me, because I went to MIT and some of my best colleagues from MIT are actually psychologists. Because this is something I was really lacking, even though I started working with it, so I read a lot of psychology books and papers and I surrounded myself with developmental psychologists and assessment specialists, because it was really a different, I won’t say way of thinking, because I’m actually bringing my way of thinking to psychology, but it’s just a different way how to do experiments.

Goren Gordon: The pure methodology is different, but I don’t feel I’m crossing any boundaries, because it’s all mixed up now, because I’m using equations to describe human behavior, and I’m doing equations for assessment. My robot is actually happy when it’s learning because it gets rewards. It’s not like I have this thing and I’m going to other disciplines. It’s like everything is mixed up.

Goren Gordon: I’ll give another example. When I’m writing algorithms for my robot, I’m not looking for the optimal algorithm, because what I am interested in is the human mind. I’m looking for algorithms that was what we call biologically plausible. I’m trying really hard not to invert matrices in my algorithms, because I don’t believe the brain inverts matrices. Everything is about neural networks and learning and stuff like that, because I’m not trying to solve a problem optimally. I’m trying to really recreate how we think. Again, I don’t feel I’m crossing boundaries, but everything is like a mishmash of everything.

Andrew Miller: Right, because my view of it from the outside was definitely this idea that you have this really big hat stand, and you sort of put on a disciplinary hat and go talk to the psychologists, or you put on a hard hat and go talk to the engineers or something, but it doesn’t feel that way?

Goren Gordon: No, so it’s actually weird because one of my degrees people actually forget I have is actually, I have a master’s in business administration, which is totally unrelated to anything but it actually taught me a lot about organizational psychology, and a lot of things that were totally weird for me. It gave me a really solid background to actually be able to communicate with psychologists and sociologists, because actually this is the biggest barrier usually, is that engineers and computer scientists just, it’s very hard to talk to psychologists because they don’t understand. How come you can do an experiment when you don’t have an algorithm or an equation? How can you do all these, I don’t know, qualitative studies?

Goren Gordon: It’s very hard for a computer scientist to understand these topics, but I actually did this master’s in business administration, and I took so much from, I don’t know, the marketing course, because it’s so weird. It’s like people doing totally irrational things for no reason whatsoever, but it’s real. It’s a really, really different way of thinking, and it really helps me when I’m now, for example, delving into irrational behavior. It gave me the background to be able to talk to all of these people, which is I think very, very important. It probably is the highest barrier for multidisciplinary research, is the language barrier and the actual way of thinking about all of these things. It’s very, very hard.

Andrew Miller: Yeah, and another one that I think keeps coming up when I’m talking to people about this stuff is just there’s other ways in which the silos make themselves present. One of those, for example, is just the language in a publication. You can sort of, when I’m reading a paper to review, I can tell if this is someone coming from another discipline. You don’t have those social in-group cues. What are some of the mechanics of how you go about structuring these projects so that you can still do what you want, but perhaps allow the collaborators to say, “Oh, actually this is how you would say this here?” or “Here’s how we word this”?

Goren Gordon: I actually surround myself or try to surround myself with specialists from the discipline that I don’t know a lot. Now, for example, I have again a lot of assessment projects, which again it’s totally, I’m doing assessments with robots and algorithms and machine learning, but again the methodology and the jargon is so different. I just take a specialist and pick her brain until we have this common language, and how to look at the data and how to write the paper and stuff like that, because there’s no way to learn it by itself.

Goren Gordon: It’s like again, you need the full bachelor’s degree and master’s in order to be able to write the paper in that discipline by yourself, and I don’t have that, so I’m just collaborating with specialists in that specific topic that I need in order to be able to write it. There’s no way I can write like a paper in psychology now alone, because I mean, I can write a paper, but it won’t get accepted. It’s a very tough job, but actually I really do feel like I’m benefitting from it, because they bring such a different view and such a different way of looking at things, that again I’m integrating it into my way of thinking also on the brain and all the algorithm stuff, so it actually works.

Andrew Miller: The other challenge I imagine for this kind of work is, at the very beginning, I imagine that you’re someone who just has a ton of research ideas. I’m wondering, what’s your process for, “Okay, this is something that I’m going to pursue,” or “This is actually too far,” or “This would require too many leaps”?

Goren Gordon: Yeah, so at one point in my life, it was actually I think, I don’t know, like probably five or six years ago, I realized, much to my dismay, that I will not be able to pursue all my ideas. It was a very shocking and despairing thought. Actually, the thing I hate and despite the most is unfulfilled potential, so if I have this idea that I will never do, then I hate to keep it to myself. I think it’s horrible and a mistake, so I actually opened a blog. It was on for just one year, and I called it The Research Fountain. Every week, I put a research idea there, so all the things I wanted, like I think were amazing but there was no chance in hell I’m going to do it. Some of them are like full-page, full-blown ideas. Some of them are like crazy start-ups, and some of them are totally bizarre.

Goren Gordon: I had there like a new way for fusion using nanotube luminescence, I don’t know, like crazy stuff, and another way, like a different shape for an umbrella. I don’t know, crazy stuff, and it was just an outlet. It was actually before I got my lab, so it was just like an outlet for all the ideas, because I was still like a PhD student. It was, you know, you don’t have the freedom when you’re a PhD student.

Goren Gordon: I kept two ideas, which are huge and I’m not going to talk about them, because I have like two huge ideas which I still hold in my heart that I’m going to pursue them in the near future, but all the other ones I just wrote and hoped for the best that someone would pick them up. Since I started my own lab, it’s actually easier. Just, I have an idea so I take in a student and, “Here, take a PhD.” My biggest challenge in the last two years was actually to stop doing that, because my lab grew and grew and grew, and now it’s too big.

Goren Gordon: My biggest challenge now is to actually learn to say no to people who want to do research and then pursue some of my ideas. But yeah, it actually is a challenge, because again you always, from every research there are actually more questions than answers, and you can’t pursue all of them. It’s actually very, very hard, even more so in such a multidisciplinary, because what you don’t know is actually bigger so you always want to look to other places.

Goren Gordon: In the boundaries of these disciplines, there’s actually much more unknown things than in the heart of psychology and the heart of computer science. In the boundary between them, there are so many unknown things that you want to try them all. In the end, currently my guiding principle is that I’m the Curiosity Lab, and I’m doing only curiosity things. If I can’t relate it directly to curiosity, I probably won’t pursue it even though it’s very, very sad. That’s the only thing that keeps me sane and be able to do all my job.

Andrew Miller: Yeah, and so you’re in this interesting phase. It sounds like you’re three years into this current position, about right?

Goren Gordon: Mm-hmm (affirmative), yes.

Andrew Miller: You’re now, I’m assuming, starting to have students graduate.

Goren Gordon: Yes.

Andrew Miller: It also must be weird that all of your students are getting engineering degrees, despite working in this sort of mixing pot.

Goren Gordon: So it’s crazy, but actually most of my students are not from engineering. I’m trying to collect students with at least one discipline that I don’t know, so I have a mechanical engineer. I have a math and philosophy. I have education, so I’m trying to collect a lot of students that bring in their own knowledge to this lab. I’m in engineering, but it’s an important issue to raise because in many universities, you don’t have this multidisciplinary faculty. In the end, you get a position at a specific faculty, so if I’m doing engineering, computer science, psychology and education, the question is, where do you go? Because there is no department that does all of this.

Goren Gordon: Again, it’s a very hard decision for all these multidisciplinary researchers. For me, I chose a place that actually has all of these. I mean, the industrial engineering department, which is an amazing department because it actually has computer scientists, psychologists and engineers. Each one is doing one or two of these things, but I have here people to talk about algorithms, and I have here people to talk about human factors, and I have people who actually, like we have a robotics education lab. This place actually is very good for me, because I’m surrounded by, for each of my disciplines that I’m doing, I have people to talk to. This is actually an amazing place, and this was the only place that actually had all of this together.

Andrew Miller: It’s interesting to hear, so you’re a few years out, sort of the trauma has faded from your most recent job search, right?

Goren Gordon: Yeah.

Andrew Miller: That must have been a really challenging problem to satisfied, right, to get all of those different factors?

Goren Gordon: Yeah, it was very hard. Again, it’s very hard because again you need to have also like students that will be able to cope with what you’re doing. For example, my biggest challenge now is that I need students that actually know how to program, know how to deal with robots, and know psychology and stuff like that, because this is the research. The learning curve in my lab is very hard, and it takes a lot of time for students to actually pick up all of these things, because there are no HRI courses, human/robot interaction courses, in my university.

Goren Gordon: They don’t have any background. They’re either programming or knowing a little bit of psychology, but it’s very hard to bring them up to speed with everything I’m doing, because every research area requires at least two or three disciplines. It’s not easy, but again, this is why I selected this department, because surprisingly, in the industrial engineering department, they actually learn how to program. They learn how to operate robots, and they learn a lot of psychological courses, so they actually are coming more prepared than any other department, which is very cool.

Andrew Miller: Scaling back even further, I’m wondering now reflecting on sort of your interdisciplinary research journey, are there things that you would do differently? Are there things that you would, advice that you would give to people who are just starting out?

Goren Gordon: Yeah, so actually I do have advice which is like a very, very strict one, which is uncustomary to people. If someone is interested in a truly multidisciplinary approach, like computers and psychology or computers and biology or there are a lot of combinations today, I really recommend to start from the more analytical one. I mean, some people, never go from biology to computers, or from psychology to computers. That’s just too hard and too weird.

Goren Gordon: I mean, start from the most analytical and most demanding, like undergrad, and from there you can go wherever you want. I mean, if you have a bachelor’s in computer science, you can go to a master’s in biology or bioinformatics or stuff like that, but the other way around doesn’t work. I’m extremely happy that I did my physics bachelor really early, because it gave me the really analytical thinking and the mathematical aptitude to do all the other stuff. Again, I’m doing psychology now, but I’m literally solving equations, dynamical system equations, which is the way to go.

Goren Gordon: If you’re interested in a multidisciplinary, start from the harder one. I’m still ambivalent regarding this double major program that you have, you know, computer and psychology in the same bachelor degree. I’m a little bit ambivalent, because you do need to know each one of them really thoroughly. I don’t know, but I think it’s better to have like a computer science degree with several psychology courses, and if you’re interested in the cross-disciplinary, then when you start research, then switch. This is my view from my long years. I mean, I’m really happy that I have this really solid physics background before I started going all the other directions.

Andrew Miller: It’s really interesting seeing that it’s not like you have to do 10 degrees. The point is to be strategic about, “Okay, I’m going to have a grounding in this, but I want to also add this into the mix.”

Goren Gordon: Yeah, to understand, yeah I know, so you don’t need to have six degrees to do computational psychology. I really do believe, you know a lot of students come and ask me about my advice, what to do first, what to do? I really think the bachelor’s should be highly analytical, either physics, computer science, maybe math. I don’t like math, but okay, but afterwards, I mean if someone doesn’t care what they learn and they just want to learn everything, then my optimal trajectory is like physics bachelor, biology master’s, and PhD what you’re actually interested in.

Goren Gordon: Because you can do computer science, you can do bioinformatics, you can do whatever you want, but I think the physics and biology are like, again, if you learn physics, it’s actually not hard to go to computer science later on. Because again, in physics, you program all the time and you also have this analytical thinking, this way of learning equations and algorithms and stuff like that because again, the other way around doesn’t work. If you do a bachelor’s in biology, you can’t do master’s in physics, just too hard.

Andrew Miller: Is there anything else that you feel we should touch on, or things that I didn’t ask you about that I said I would?

Goren Gordon: Well, I have all the clichés, but it’s amazing how many things are in the boundaries. From my perspective, all the social sciences and humanities will greatly benefit from a migration from computer science to these disciplines. I mean, there’s so many things I think we can benefit from this interaction, and for all the computer scientists, don’t be afraid to thoroughly interact and learn from the psychologists and sociologists and anthropologists. They have a totally different view, but it’s complementary. You can really try to integrate everything they know to your scheme of how things work, and I think that’s really just fun to go and expand. As the head of the Curiosity Lab, never stop learning, ever. This is the most important, and actually there are studies to show that if you continue learning and are curious, then your whole life is better, so just, you know.

Andrew Miller: Well, that’s a great way to end I think, so thank you so much for joining me.

Goren Gordon: Thank you very much for having me.

The post Goren Gordon | Curiosity Across Disciplines appeared first on ACM FCA.