Skip to content
Regina Barzilay: Deep Learning for Cancer Diagnosis and Treatment | Lex Fridman Podcast #40
This video isn’t embeddableWatch on YouTube →
Lex Fridman PodcastLex Fridman Podcast

Regina Barzilay: Deep Learning for Cancer Diagnosis and Treatment | Lex Fridman Podcast #40

Lex Fridman and Regina Barzilay on mIT’s Regina Barzilay on Deep Learning, Cancer, and Life’s Purpose.

Lex FridmanhostRegina Barzilayguest
Sep 23, 20191h 17mWatch on YouTube ↗

EVERY SPOKEN WORD

  1. 0:00 – 1:12

    Setting the stage: From NLP to deep learning in oncology

    1. LF

      The following is a conversation with Regina Barzilay. She's a professor at MIT and a world-class researcher in natural language processing and applications of deep learning to chemistry and oncology or the use of deep learning for early diagnosis, prevention, and treatment of cancer. She's also been recognized for teaching of several successful AI-related courses at MIT including the popular Introduction to Machine Learning course. This is the Artificial Intelligence podcast. If you enjoy it, subscribe on YouTube, give it five stars on iTunes, support it on Patreon or simply connect with me on Twitter @lexfridman, spelled F-R-I-D-M-A-N. And now here's my conversation with Regina Barzilay. In an interview you've mentioned that if there's one course you would take, it would be a literature course with a friend of yours, uh, that a friend of yours teaches. Just out of curiosity, 'cause I couldn't find anything on it, are there books or ideas that had profound impact on your life journey, books and ideas perhaps outside of computer science and the technical fields?

  2. 1:12 – 5:02

    Books that shaped her worldview: cancer history and immigration stories

    1. RB

      I think because I'm spending a lot of my time at MIT and previously in other institutions where I was a student, I have limited ability to interact with people so a lot of what I know about the world actually comes from books, uh, and there were quite a number of books that had profound impact on me and how I view the world. Let me just give you, mm, one example of such a book. I've, um, maybe a year ago read a book called The Emperor of All Maladies. It's a book about, um, it's kind of a history of science book on how the treatments and drugs for cancer were developed, and that book, despite the fact that I am in the business of science, really opened my eyes on how imprecise and imperfect the discovery process is, a- and how imperfect our current solutions, uh, and what makes science succeed and be implemented and sometimes it's actually not the strengths of the idea but devotion of the person who wants to see it implemented. So, this is one of the books that, you know, at least for the last year quite changed the way I'm thinking about scientific process just from the historical perspective and what do I need to do to make my ideas really implemented. Let me give you an example of a book which is not a kind of, uh, which is a fiction book, is a book called Americanah and this is a book about a young female student who comes from Africa to study in the United States and it describes her past, uh, you know, within her studies and, uh, her life transformation that, you know, i- in a new country and kind of adaptation to a new culture.

    2. LF

      Mm-hmm.

    3. RB

      And when I read this book, I saw myself in many different points of it, uh, but- but it also (laughs) kind of gave me the- the lens on different events and some events that I never actually paid attention. One of the funny stories in this book is how she, uh, arrives, uh, to- to her c- new college and she starts speaking in English and she had this beautiful British a- accent because that's how she was educated-

    4. LF

      Mm.

    5. RB

      ... uh, in her country, uh, this is not my case, uh, and then she notices that the person who talks to her, you know, talks to her in a very funny way, in a very slow way, and she's thinking that this woman is disabled, uh, in- in... and she's also trying to kind of to accommodate her and then after a while when she finishes her discussion with this officer from her college, she sees how she interacts with the other students, with American students, and she (laughs) discovers that actually, uh, she talked to her this way because she thought that she doesn't understand English-

    6. LF

      Mm.

    7. RB

      ... and, uh, I thought, "Wow, this is a funny experience," and (laughs) l- literally within few weeks, I went to- to LA to a conference and I asked somebody in- in the airport, you know, how to find, like, the, a cab or something and then I noticed that this person is talking in a very strange way and my first thought was that this person have some, you know, pronunciation issues or something and I'm trying to talk very slowly to him and I was with another professor, Ernst Fränkel, and he's like laughing (laughs) because it's funny that I don't get that the guy is talking in- in this way because he thinks that I cannot speak. So, it was really kind of mirroring experience and it let me think a lot about my own experiences moving, you know, from different countries. So, I think that books play a big role in my understanding of the world.

  3. 5:02 – 7:00

    Why ideas don’t win alone: personalities, adoption, and scientific “dark ages”

    1. LF

      On the- on the science question, you mentioned that it made you discover that personalities of human beings are more important than perhaps ideas. Is that what I heard?

    2. RB

      It's not necessarily that they are more important than ideas, but I think that ideas on their own are not sufficient and, um, many times, at least at the local horizon, it's the personalities and their devotion to the ideas is really that locally changes the landscape. Now if- if you're looking at AI, like let's say 30 years ago, you know, dark ages of AI or whatever word is symbolic, times you can use any word, you know there were some people l- now- now we're looking at a lot of that work and we're kind of thinking this was not really, uh, maybe a relevant work, but you can see that some people managed to take it and to make it so shiny and, uh, dominate the, you know, the academic world and make it to be the standard. If you look at the area of natural language processing, uh-... it is well-known fact that the reason the statistics in NLP took such a long time to became- to become mainstream, because there were quite a number of personalities which didn't believe in this idea and then stop research progress in this area. So I do not think that, you know, kind of asymptotically maybe personalities matters, but I think, uh, locally-

    3. LF

      (laughs)

    4. RB

      ... it does make quite a bit of impact and it's-

    5. LF

      Beautifully put, okay. (laughs)

    6. RB

      ... generally, uh, you know, speed ups, speeds up the rate of adoption of the new ideas.

    7. LF

      Yeah, and, uh, and the other interesting question is in, uh, the early days of particular discipline, I think you mentioned, uh, i- in- in that book was is ultimately a book of cancer. Oh-

    8. RB

      It's called the, uh, The Emperor of All Maladies.

    9. LF

      Yeah, the... yeah. And those maladies included the trying to- the medi- medicine? Was it centered around medicine mostly?

  4. 7:00 – 8:38

    Cancer treatment history: accidental origins and sobering trial-and-error

    1. RB

      So, so it was actually, uh, centered on, you know, how people thought of curing cancer, like i- like for me, it was really a discovery how people... what was the science of chemistry behind drug development, that it actually grew up out of d- dying, like coloring industry, that people-

    2. LF

      Hm.

    3. RB

      ... who developed chemistry in 19th century in Germany and Britain to do, you know, the, the really new dyes, uh, they looked at the molecules and identified that they do certain things to cells, and from there the process started and, you know, like historically saying, "Yeah, this is fascinating that they managed to make the connection and-"

    4. LF

      Mm-hmm.

    5. RB

      "... look under the microscope and do all this discovery." But as you continue reading about it and you, uh, read about how ch- chemotherapy drugs, which were developed in Boston, and, uh, some of them were developed and, um, um, Farber, Dr. Farber from Dana-Farber, you know, how the experiments were done, uh, that, you know, there was some miscalculation, let's put it this way, and they tried it on the patients and then just, uh, and those were children with leukemia and they died. And then they tried another modification. You look at the process, how imperfect is this process and, you know, like if we're again loo- looking back like 60 years ago, 70 years ago, you can kind of understand it, but some of the stories in this book which were really shocking to me were really happening, you know, maybe decades ago, and we still don't have a vehicle to do it much more fast and effective and, you know, scientific the way I'm thinking computer science scientific.

  5. 8:38 – 11:33

    Do we need mechanistic understanding? Computer science vs biology mindsets

    1. LF

      So from the perspective of computer science, you've gotten a chance to work the application to cancer and to medicine in general. From a perspective of an engineer and a computer scientist, how far along are we from understanding the human body, biology, of being able to manipulate it in a way we can cure some of the maladies, some of the diseases?

    2. RB

      So this is very interesting question and if you're thinking as a computer scientist about this problem, uh, I think one of the reasons that we succeeded in the areas we as a computer scientist succeeded is because we don't have... we are not trying to understand in some ways, like if you're thinking about like e-commerce, Amazon, Amazon doesn't really understand you and that's why it recommends you certain books or certain products, correct? And, uh, in, um, you know, traditionally when people were thinking about marketing, you know, they divided the population to different kind of subgroups, identify the features of this subgroup and come up with a strategy which is specific to that subgroup. Uh, if you're looking about recommendation system, they're not claiming that they're understanding somebody, they're just managing to, from the patterns of your behavior, to recommend you a product. Now, if you look at the traditional biology, and obviously I wouldn't by... say that I, uh, am at any w- way, you know, educated in this field, but, you know, what I see, there is really a lot of emphasis on mechanistic understanding, and it was very surprising to me coming from computer science how much emphasis is on this understanding. And e- given the complexity of the system, maybe the deterministic full understanding of these processes is, you know, beyond our capacity, and the same with the computer science when we do recognition, when doing recommendation in many other areas, i- it's just probabilistic matching process and in some way maybe, uh, in certain cases we shouldn't even attempt to understand or we can attempt to understand but in parallel we can actually do this kind of matchings that would, uh, help us to find cure or to do early diagnostics and so on.

    3. LF

      Hm.

    4. RB

      And I know that in these communities it's really important to understand, but I'm sometimes wondering (laughs) what exactly does it mean to understand here?

    5. LF

      Well, there's stuff that works and, uh, but that can be, like you said, separate from this deep human desire to uncover the mysteries of the universe, of, uh, of science, of the way the body works, the way the mind works. It's the dream of symbolic AI, of being able to reduce human knowledge into, into logic and be able to play with that logic in a way that's very explainable and u- understandable for us humans. I mean, that's a beautiful dream. So I, I understand it,

  6. 11:33 – 18:02

    A personal turning point: Regina’s 2014 breast cancer diagnosis

    1. LF

      but it seems that what seems to work today, and we'll talk about it more, is as much as possible reduce stuff into data, reduce whatever problem you're interested in to data and try to apply (laughs) statistical methods, apply machine learning to that. On, on a personal note, you were diagnosed with breast cancer in 2014. What did facing your mortality make you think about? How did it change you?

    2. RB

      You know, this is a great question and, uh, the thing that I was interviewed many times and nobody actually asked me this question. I think I was-... 43 at a time, and it's the first time I realized in my life that I may die. Uh, and I never thought about it before. And, yeah, and there is a long time since you're diagnosed until you actually know what you have and how severe is your disease. For me, it was like maybe two and a half months. And I didn't know where I am, uh, during this time because I was getting different tests and one would say it's bad and I would say, "No, it is not." So until I knew where I am, I really was thinking about all the different possible outcomes.

    3. LF

      Were you imagining the worst or were you trying to be optimistic, or what-

    4. RB

      It would be the really, uh, uh, I, I don't remember, you know, what was my thinking. It was really a mixture with many components at the time.

    5. LF

      (laughs) Wow.

    6. RB

      Uh, speaking, you know, uh, in our terms. And one thing that I remember, uh, and, and you know, every test comes and then you're saying, "Oh, it could be this or it might not be this," and you're hopeful and then you're desperate. So it's like it, there is a whole, you know, slew of emotions that goes through you. But what I remember is that when I, uh, came back to MIT, I was kind of going through, the whole times through the treatment to MIT but my brain was not really there. But when I came back really, finished my treatment and I was here teaching and everything, you know, I look back at what my group was doing, what the other groups was doing and I thought it's trivialities. It's like people are building their careers on improving some parser on two, 3% or whatever. And I was, it's like seriously, I did a work on how to decipher Ugaritic, like a language that nobody speak and, and whatever, like what is significance? When all of a sudden, you know, I walked out of MIT which is, you know, when people really do care, you know, what happened to your ICLR paper and (laughs) you know-

    7. LF

      Hm.

    8. RB

      ... what is your next publication, uh, to ACL. To the world where people, you know, people... You, you see a lot of sufferings that I'm kind of totally shielded on it on daily basis, and it's like the first time I've seen like real life and real suffering. And I was thinking, "Why are we trying to improve the parser or deal with some trivialities when we have capacity to really make change?" And it was really challenging to me because on one hand, you know, I have my graduate students who really want to do their papers and their work, and they want to continue to do what they were doing, which was great. And then it was me who really (laughs) kind of re-evaluated what is important. And also, at that point, because I had to take some break, I, uh, look back into like my years in science and I was thinking, you know, like 10 years ago this was the biggest thing. I don't know, topic models, there we have like millions of papers on topic models and variation on topics models, now it's totally like irrelevant. And you, you, you start looking at this, you know, what do you perceive as important at different point of time? And how, you know, it's ca- fades over time. And, uh, since we have limited time, all of us have limited time on Earth, it's really important to prioritize things that really matter to you, maybe matter to you at that particular point. But it's important to take some time and understand what matters to you, which may not necessarily be the same as what matters to the rest of your scientific community and pursue that vision.

    9. LF

      So though that moment... D- Did it make you cognizant, you mentioned suffering, of just the general amount of suffering in the world? Is that what you're referring to? So as opposed to topic models and specific detail, problems in NLP, did, did you start to think about other people who have been diagnosed with cancer?

    10. RB

      Well, I-

    11. LF

      Is that the way you saw the w- started to see the world, perhaps?

    12. RB

      Oh, absolutely, and it actually (laughs) creates because, uh, like, uh, for instance, you know, there is parts of the treatment where you need to go to the hospital every day and you see, you know, the community of peoples that you see and many of them are much worse than I, I was at the time. And you all of a sudden see it all. And people who are happier someday just because they feel better and for people who are in our normal realm, you take it totally for granted that you feel well, that if you decide to go running, you can go running, and, uh, you can... You know, you're pretty much free to do whatever you want with your body. Like, I, I saw like a community, my community became those people. And, um, I, I remember one of my friends, Dina Katabi took me to Prudential to buy me a gift for my birthday, and it was like the first time in months that I went to kind of to see other people and I was like, "Wow, first of all, these people are, you know, they're happy and they're laughing and they're very different from these other my people." And secondly thinking, "Aren't they totally crazy? They're like laughing and wasting their money on, on some stupid, uh, gifts." And, you know, they may die. They already may have cancer.

    13. LF

      Yeah.

    14. RB

      And, and they don't understand it. So you can really see how, (laughs) how the mind changes, that you can see that... You know, before that you can ask, "Didn't you know that you're gonna die?" Of course I knew, but it was (laughs) kind of a theoretical notion, it wasn't something which was concrete. And at that point when you really see it and see how little means sometimes the system has to help them, you really feel that we need to take a lot of our brilliance that we have here at MIT and translate it into something useful.

    15. LF

      Yeah. And useful can have a lot of definitions, but of course alleviating suffering, alleviating trying to cure cancer is a beautiful mission. So I of course know the theo- theoretically the notion of cancer-

    16. RB

      Mm-hmm.

  7. 18:02 – 21:43

    Curing cancer vs predicting it early: where AI can change outcomes fastest

    1. LF

      ... but just reading more and more about it, uh, 1.7 million new cancer cases in the United States every year, 600,000 cancer-related deaths every year.... so this has a huge impact, United States, globally. W- when, uh, broadly, before we talk about how machine learning, how MIT can help, when do you think we, as a civilization, will cure cancer? How hard of a problem is it from everything you've learned from it recently?

    2. RB

      I cannot really assess it. What I do believe will happen with the advancement in machine learning, is that a lot of, uh, types of cancer we will be able to predict way early, and more effectively utilize existing treatments. Uh, I think, I hope at least, that with all the advancements in AI and drug discovery, we would be able to much faster find relevant molecules. What I'm not sure about is how long it will take the medical establishment and regulatory bodies to kind of catch up and to implement it. And I think this is a very big piece of puzzle that is currently not addressed.

    3. LF

      That's the really interesting question. So first, a small detail that I think the answer is yes, but is cancer one of, uh, o- one of the diseases that when detected earlier, that's a significantly improves the outcomes?

    4. RB

      (sighs)

    5. LF

      It... so like... 'cause we'll talk about there's the cure, and then there is detection, and I think where machine learning can really help is earlier detection. So does detection help with-

    6. RB

      Detection is crucial. For instance, the vast majority of pancreatic cancer patients are detected at the stage that they are incurable. That's why they have such a, you know, terrible survival rate. I- it's like just few percent over five years. Uh, is pretty much today their sentence. But if you can discover this disease early, there are mechanisms to treat it. And in fact, I know c- a number of people who were diagnosed and saved just because they had food poisoning. They had terrible food poisoning, they went, uh, to ER, they got scan-

    7. LF

      Mm-hmm.

    8. RB

      ... there were early signs on the scan and that what saved their lives. But this wasn't really an accidental case. So as we become better, uh, we would be able to help to many more people, uh, that have, uh, you know, that are likely to develop diseases. And I just want to say that as I got more into this field, I realized that, you know, cancer is of course a terrible disease, but there are (laughs) really the whole slew of terrible diseases out there, uh, like neurodegenerative diseases and others. Uh, so we of course, a lot of us are fixated on cancer just because it's so prevalent in our society and you see these people, but there are a lot of patients with neurodegenerative diseases-

    9. LF

      Mm-hmm.

    10. RB

      ... and the kind of aging diseases that we still don't have a good solution for. And, uh, we, y- you know, and I felt as a computer scientist, we kind of decided that it's other people's job to treat these diseases, uh, because it's like traditionally people in biology or in chemistry or MDs are the ones who are thinking about it. A- and after kind of start paying attention, I think that it's really, uh, wrong assumption and we all need (laughs) to join the battle.

  8. 21:43 – 23:26

    Machine learning for cancer risk and diagnosis: combining weak signals at scale

    1. LF

      So how... it seems like in cancer specifically, that there's a lot of ways that machine learning can help. So what's, what's the role of machine learning in the diagnosis of cancer?

    2. RB

      So for many cancers today, we really don't know what is your likelihood to get cancer and for the vast majority of patients, especially on the younger patients, it really comes as a surprise. Like for instance, for breast cancer, 80% of the patients are first in their families, it's like me-

    3. LF

      Hmm.

    4. RB

      ... and I never saw that I had any increased risk because, you know, nobody had it in my family and for some reason in my head it was kind of a inherited disease, but even if I would pay attention the, the models that currently, it's very simplistic statistical models that are currently used and in clinical practice they really don't give you an answer so you don't know. And, uh, the same true for pancreatic cancer, the same true for, uh, non-smoking lung cancer and many others. Uh, so what machine learning can do here is utilize all this data-

    5. LF

      Mm-hmm.

    6. RB

      ... to tell us early, uh, who is likely to be susceptible and d- using all the information that is already there, be it imaging, uh, be it your other tests, and you know eventually, uh, liquid biopsies and others where the signal itself is not sufficiently strong for human eye to do good discrimination because the signal may be weak but by combining many sources, a machine which is trained on large volumes of data can really, uh, detect it early and that's what we've seen with breast cancer and people are reporting it in other diseases as well.

  9. 23:26 – 34:21

    The data bottleneck: access barriers, privacy, and patient-driven data donation

    1. LF

      That really boils down to data, right? And, and the different kinds of sources of data and you mentioned regulatory challenges, so w- what are the challenges in gathering large datasets in this space?

    2. RB

      Again, another great question. So it took me after I decided that I want to work on it two years to get access to data. Uh-

    3. LF

      Any data? Like, any significant dataset?

    4. RB

      Any significant amount, like right now in this country, there is no publicly available dataset of modern mammograms that you can just go on your computer, sign a document and, and get it. It just doesn't exist. I, I mean obviously every hospital has its own collection of mammograms, there are data that come out... that came out of clinical trials but we're talking about you as a computer scientist who just want to run-... his or her model and see how it works. This data, like ImageNet, doesn't exist, and, uh, the im- there is an e- uh, set which is called like Florida Dataset, which is a film mammogram from '90s, which is totally not representative of the current developments. Whatever you're learning on them doesn't scale up. This is the only resource that is available, and, uh, today there are many agencies that govern access to data. Like, the hospital holds your data and the hospital decides whether they would give it to the researcher to work with this data or not.

    5. LF

      An individual hospital? So like a-

    6. RB

      Uh, yeah, I mean, the hospital may, um, you know, assuming that you do research collaboration, you can submit... You know, there is appro- approval process guided by IRB, and you, if you go through all the processes, you can eventually get access to the data but if, y- you yourself know our AI community, (laughs) there are not that many people who actually ever got access to data because it's very challenging process, uh-

    7. LF

      And, uh, sorry, just in a quick comment, e- MGH or any kind of hospital, are they scanning the data? Are- are they digitally storing it usually?

    8. RB

      Oh, it is already digitally stored. You don't need to do any extra processing steps. It's already there in the right format. Is that- that right now, there are a lot of issues that govern access to the data, because the hospital is legally responsible for- for the data, and, you know, they have a lot to lose if they give the data to the wrong person, but they may not have a lot to gain if they give it... Uh, as a hospital, as a legal org- entity, um, is giving it to you, and the way, you know, what I would imagine happening in the future is the same thing that happens when you're getting your driving license, you can decide whether you want to donate your organs. You can imagine that whenever a person goes to the hospital, they... It should be easy for them to donate their data for research and it can be different kind of, do they only give you test results or only mammogra- or only imaging data or the whole medical record? Uh, because at the end, we all will benefit from all this insights, and it's not like you can say, "I want to keep my data private, but I would really love to get it f- you know, from other people, because other people are thinking the same way." So if there is a mechanism to do this, uh, donation and- and the patient has an ability to say how they want to use their data for research, uh, it would be really a game changer.

    9. LF

      People, when they think about this problem, there's, uh, it depends on the population, depends on the demographics, but there's some privacy concerns. G- generally when... Not- not just medical data, just any kind of data. It's what you said, "My data, it should belong kinda to me. I'm worried how it's going to be misused." How do, h- how do we alleviate those concerns? Um, because that seems like a problem that needs to be... That problem of trust, of transparency, needs to be solved before we build large datasets that help detect cancer, help save those very people in their- in the future.

    10. RB

      So I think there are two things that could be done. There is a technical, uh, solutions and there are societal solutions. So on the technical end, we today have ability to improve disambiguation, you know, i- like for instance, for imaging, it's e- easy. For, you know, for imaging it you can do it pretty well.

    11. LF

      What's disambiguation?

    12. RB

      Disam- uh, sorry. Disambiguation.

    13. LF

      (laughs)

    14. RB

      Removing deidentification, removing the names-

    15. LF

      Ah.

    16. RB

      ... of the people.

    17. LF

      Mm-hmm.

    18. RB

      There are other data, like if it is a rotex, you cannot really, uh, achieve 99.9%, but- but there are all these techniques that actually some of them are developed at MIT, how you can do learning on the encoded data, where you locally encode the image, you train, uh, a network which only works on the encoded, uh, on encoded images, and then you send the outcome back to the hospital, and you can open it up. So those are the technical solution. There are a lot of people who are working in this space, where the learning happens in the encoded form. I, we are still early-

    19. LF

      That's fascinating.

    20. RB

      ... uh, but this is a interesting research area where I think we'll make more progress. There is a lot of work in natural language processing community, how to do de- deidentification better.

    21. LF

      Yeah.

    22. RB

      But even today, there are already a lot of data which can be deidentified perfectly, like your test data, for instance. Correct? Uh, where you can just, you know who- the name of the patient, you just want to extract the part with the numbers. The big problem here is, again, hospitals don't see much incentive to give this data away on one hand, and then there is general concern. Now, when I'm talking about societal benefits and about the education, the public needs to understand anything, um, that there are situation, and I still remember myself when I really needed an answer. I had to make a choice, and there was no information to make a choice. You're just guessing, and y- at th- that moment you feel that your life is at stake, uh, but you just don't have information to- to- to make the choice, and many times when I give talks, um, I get emails from women who say, "You know, I'm in this situation, can you please run statistic and see what are the outcomes?" We get, uh, almost every week a mammogram that comes by mail (laughs) to my office at MIT.

    23. LF

      (laughs)

    24. RB

      I'm serious.

    25. LF

      That's amazing.

    26. RB

      Uh, that, uh, people ask to run because they need to make, you know, life changing decisions, and, uh, of course, you know, I'm not planning to open a clinic here but, uh, w- we do run and give them the results for their doctors. But-... the point that I'm trying to make, that we all, at some point, or our loved ones, will be in this situation where you need information to make the best choice, and if this information is not available, you would feel vulnerable and unprotected. And then the question is, you know, what do I care more? Because in the end everything's a trade-off, correct?

    27. LF

      Yeah, exactly. Just, uh, out of curiosity, what... it seems like one possible solution, I'd like to see what you think of it, based on what you just said, based on wanting to know answers for when, when you're yourself in that situation. Is it possible for patients to own their data, as opposed to the hospitals owning their data?

    28. RB

      (sighs)

    29. LF

      Of course, theoretically, I guess patients own their data, but can you walk out there with a USB stick containing everything, or upload it to the cloud, where a company... You know, I- I remember Microsoft had a service, th- like I try, I was really excited about, and Google Health was there. I tried to give... I- I was excited about it. Basically, companies helping you upload your data to the cloud so that you can move from hospital to hospital, from doctor to doctor. Do you see a promise of that kind of possibility?

    30. RB

      I absolutely think this is, you know, the right way to, to exchange the data. I don't know now who's the biggest player in this field, but I can clearly see that even, you know, for ... even for totally selfish health reasons, when you are going to a new facility, and many of us are sent to some specialized treatment, they don't easily have access to your data. And, uh, today, you know, (laughs) women who want to send a mammogram need to go the hospital, find some small office which gives them the CD and they ship as a CD. So you can imagine we're looking at the kind of decades-old mechanism of data exchange. So I definitely think this is an area where hopefully all the right regulatory and technical forces will align, and we will see it actually implemented.

  10. 34:21 – 40:51

    Why better algorithms don’t automatically change care: regulation and incentives

    1. RB

      I think that in many cases when even people do have data, we still don't know what exactly do you need to demonstrate to- to change the standard of care?

    2. LF

      Ah.

    3. RB

      Uh, like let me give you example related to my breast cancer research. So traditional- in traditional breast cancer risk assessment, uh, there is something called density which determines the likelihood of a woman to get cancer, and this pretty much says how much white do you see on the mammogram. The whiter it is, the more likely the tissue is dense. And, um, the idea behind density, it's not a bad idea. In 1967, a radiologist called Wolf decided to look back at women who were diagnosed and, uh, see what is special in their images. Can we look back and say that they're likely to develop? So he come up with some patterns, and it was the best that his human eye can, uh, you know, can identify, then it was kind of formalized and coded into four categories, and that what we are using today. And, uh, uh, today this density assessment is actually federal law from 2019, the- uh, approved by President Trump and for the previous FDA commissioner, where women are supposed to be advised by their providers if they have high density, putting them into higher-risk category, and, um, in some states you can actually get supplementary screening paid by your insurance because you're in this category. Now, you can say, "How much science do we have behind it?" Whatever, biological science or epidemiological evidence. So it turns out that between 40 and 50% of women have dense breasts. So about 40% (laughs) of patients are coming out of their screening and somebody tells them, "You are in high risk."

    4. LF

      Mm-hmm. High risk.

    5. RB

      Now, what exactly does it mean if you, as half of the population, high risk? ??? A- ah, maybe I'm not. You know, or what do I really need to do with it? Because-... the system doesn't provide me a lot of the solutions because there are so many people like me, we cannot really provide very expensive solutions for them. And the reason this whole density became this big deal, it's actually advocated by the patients who felt very unprotected because many women went, did the mammograms which were normal, and then it turns out that they already had cancer-

    6. LF

      Mm-hmm.

    7. RB

      ... quite developed cancer. So they didn't have a way to know who is really at risk, and what is the likelihood that when the doctor tells you you're okay, you are not okay. Well, at the time, and it was, you know, 15 years ago, this maybe was the best piece of science that, uh, we had, and, uh, it took (laughs) you know, quite, uh, 15, 16 years to make it federal law. But now that this is, this is a standard, now with the deep learning model, we can so much more accurately predict who is going to develop breast cancer just because you are trained on a logical thing, and instead of describing how much white and what kind of white, machine can systematically identify the patterns, which was the original idea behind the sort of the radiologist machine is can do it much more systematically and predict the risk when you train the machine to look at the image and to say the risk in one to five years. Now, you can ask me how long it will take-

    8. LF

      (laughs)

    9. RB

      ... to substitute this density which is broadly used across the country and really is not helping t- to bring these new models. And I would say it's not a matter of the algorithm. Algorithm's already orders of magnitude better than what is currently in practice. I think it's really the question, who do you need to convince? How many hospitals do you need to run the experiment? What... You know, all this mechanism of adoption, and how do you explain to patients and to women across the country that this is really a better measure? And again, I don't think it's an AI question. We can work more and make the algorithm even better, but I don't think that this is a current, uh, you know, the barrier. The barrier is really this other piece that for some reason is not really explored. It's like anthropological piece. And, uh, coming back to your question about books, there is a book that I'm reading, uh, it's called American Sickness by Elizabeth Rosenthal, and I got this book from my clinical collaborator, Dr. Connie Lehman, and I thought I know everything that I need to know about American health system, but you know, every page doesn't fail to surprise me. And I think that there is a lot of interesting and really deep lessons for people like us from computer science who are coming into this field, to really understand how complex is the system of incentives-

    10. LF

      Mm-hmm.

    11. RB

      ... in, in the system to understand how you really need to play to drive adoption.

    12. LF

      You just said it's complex, but i- if we're trying to simplify it, who do you think most likely would be successful if we push on this group of people? Is it the doctors? Is it the hospitals? Is it the governments or policymakers? Is it the individual patients, consumers? Who needs to be inspired to most likely lead to, uh, adoption? Or is there no simple answer? (laughs)

    13. RB

      There's no simple answer, but I think there is a lot of good people in medical system who do want, uh, you know, to make a change, and I think a lot of power will come from us as the consumers because we all are consumers or future consumers of healthcare services, and I think we can do so much more in explaining the potential and noting the hype terms and not saying that we now cured all Alzheimer's, and, you know, I'm really sick of reading these kind of articles which make these claims. But really to show with some examples what this implementation does and how it changes the care, because I can't imagine, doesn't matter what kind of politician it is, you know, we all are susceptible to these diseases. There is no one who is free. Uh, and eventually, you know, we all are humans and we are looking for way to alleviate the suffering, and, and this is one possible way where we currently are under-utilizing which I think can help.

  11. 40:51 – 50:11

    Beyond diagnosis: drug design as a frontier for ML innovation

    1. LF

      So, it sounds like the biggest problems are outside of AI in terms of the biggest impact at this point, but, uh, are there any open problems in the application of ML to oncology in general? So improving the detection or any other creative methods, whether it's on the detection, segmentations of the vision perception side, or some other clever, uh, inference? Uh, yeah, what in, what in general in your, in your view are the open problems in this space?

    2. RB

      Yeah, I just want to mention that beside detection, another area where I am kind of quite active and I think it's really an increasingly important area in healthcare is drug design.

    3. LF

      Absolutely.

    4. RB

      Uh, because, uh, you know, uh, eh, it's fine if you detect something early but you still need to get, uh, you know, to, to get drugs and, uh, new drugs for these conditions, and today all of the drug design, ML is nonexistent there. We don't have any drug that was developed by the ML model, or even not developed but at least they knew that ML model plays some significant role. Uh, I think, uh, this area with all the new ability to generate molecules with desired properties to do in silico screening, uh, is really a big open area. To be totally honest with you, you know, when we are doing diagnostics and imaging, primarily taking the ideas that were developed for other areas and you're applying them with some adaptation, the area of, you know-Drug design is really technically interesting and exciting area. You need to work a lot with graphs and, uh, uh, capture various 3D properties. There are lots and lots of, uh, opportunities to be technically creative, and, um, I think there are a lot of open questions in this area. W- w- you know, we're already getting a lot of successes even, you know, with the kind of the first generation of these models, but there is much more new creative things that you can do. And what's very nice to see is that actually the, you know, the- the more powerful, the more interesting models actually do do better, so there is a place to- to innovate in machine learning in this area. Uh, and some of these techniques are really unique to, let's say, to, you know, graph generation and other things. So, uh...

    5. LF

      Wha- what, uh... Just to interrupt really quick, I'm sorry. Uh, graph generation or graphs, is drug discovery in general, what's, uh, how do you discover a drug? Is this chemistry? Is this trying to predict different chemical reactions or is it, uh, some kind of... W- what do graphs even represent in this space? (laughs)

    6. RB

      Ah, oh sorry. Ah, sorry. So...

    7. LF

      And what's a drug? (laughs)

    8. RB

      As they say. Okay. So let's say you're thinking there are many different types of drugs, but let's say you're gonna talk about small molecules because uh, uh, I think today the majority of drugs are small molecules. So small molecule is a graph, the molecule is just where the node, uh, in the graph is an atom and then you have-

    9. LF

      Yes.

    10. RB

      ... the bond, so it's really a graph representation if you're looking at it in 2D, correct?

    11. LF

      Yeah.

    12. RB

      You can do it 3D, but let's say we're, let's keep it simple and stick in 2D. Um, so pretty much, uh, my understanding today how it is done at scale in the companies, uh, you, without machine learning. You have high-throughput screening, so you know that you are interested to get certain biological activity of the compound, so you scan a lot of compounds, uh, like, uh, maybe hundreds of thousands, some really big number of compounds. You identify some compounds which have the right activity and then at this point, you know, the chemists come and they're trying to now to optimize this original hit to different properties that you want it to be maybe soluble, you want, uh, to decrease toxicity, you want, uh, to decrease the side effects. So then a-

    13. LF

      Are those, uh... Sorry, again, sorry to interrupt. Are, are, uh, can that be done in simulation or just by looking at the molecules or do you need to actually run reactions in real labs with lab coats and stuff?

    14. RB

      So, so there is... So when you do high-throughput screening, you really do screening. It's in the lab, but it's- it's really the lab screening. You screen the molecules, correct? You don't really-

    15. LF

      I don't know what screening is. (laughs) So what-

    16. RB

      Uh, the screening is just check them for certain property, usually.

    17. LF

      Like, in the physical space?

    18. RB

      The-

    19. LF

      In the physical world?

    20. RB

      Yeah.

    21. LF

      Like, actually there's a machine probably that is doing some... That- that-

    22. RB

      Yeah.

    23. LF

      ... that's actually running the reaction?

    24. RB

      That's actually running the reactions, yeah.

    25. LF

      Yeah. Wow.

    26. RB

      So, so, so there is a process where you can run, and that's why it's called high-throughput, that, you know, it become cheaper, uh, and faster to do it on very big number of molecules. You run the screening, you identify potential, um, you know, potential good starts, and then where the chemists come in who, you know, have done it many times and then they can try to look at it and say, "How can it change the molecule to get the desired, uh, profile in terms of all other properties?" So maybe how do we make it more bioactive and so on. And there, you know, the creativity of the chemist really, um, is one that determines the success of this design, uh, th- because they, again, they have a lot of domain knowledge of, you know, what works, how do you decrease toxicity and so on. And that's what they do. So all the drugs that are currently, you know, in the, uh, F- FDA approved drugs or even drugs that are in clinical trials, they are designed using these, uh, domain experts which-

    27. LF

      Yeah.

    28. RB

      ... uh, goes through this combinatorial space of molecules on graphs or whatever, and find the right one or adjust it to be the right ones.

    29. LF

      Sounds like the- the- the breast density heuristic from '67.

    30. RB

      It- it's not-

  12. 50:11 – 57:15

    Her NLP journey: from rule-based systems to data-driven translation—and brittleness

    1. LF

      ... I think you do ha- you have done a lot of, uh, really great research in NLP, natural language processing. Uh, ph- can you tell me your journey through NLP? What ideas, problems, approaches, were you, uh, working on, were you fascinated with, did you explore before this magic of deep learning reemerged and after?

    2. RB

      So, when I started from my work in NLP, it was in '97. (laughs) This was very interesting time. It was exactly the time that I came to ACL. At the time, I could barely understand English, uh, but it was (laughs) exactly like the transition point, because half of the p- papers were really, you know, rule-based approaches, where people took more kind of heavy linguistic approaches for small domains and tried to, um, build up from there, and then there were the first generation of papers which were corpus-based papers and they were very simple in our terms, when you collect some statistics and do prediction based on them. But I found it really fascinating that, you know, one community can think so very differently a- about, uh, you know, about the problem. And, uh, I remember (laughs) my first paper that I wrote, it didn't have a single formula, it didn't have evaluation, it just had examples of outputs, and this was the standard of the, of the field at the time, uh, in some ways. I mean, people, uh, maybe just started emphasizing the empirical evaluation, but for many applications, like summarization, you, you just show some examples of outputs. And then increasingly you can see that how the statistical approaches dominated the field and we've seen, you know, increased performance across many basic tasks. The sad part of the story, maybe, is that if you look again through this journey, we see that the role of linguistics in some ways greatly diminishes, and I think that you really need to look through the whole proceeding to, to, to find one or two papers which make some interesting linguistic references. It's really big-

    3. LF

      You mean today, yeah.

    4. RB

      Today. Today. This was definitely-

    5. LF

      So things like syntactic trees, just even basically against our conversation about human understanding of language, uh, which is I guess what linguistics would be, structured, hierarchically represe- representing language in a way that's humanly explainable, understandable, is, is missing today.

    6. RB

      I don't know if it is, what is explainable and understandable. In the end, you know, we perform functions, and it's okay to have machine which performs a function. Like, when you're thinking about your calculator, correct? Your calculator can cal- do calculation very different from you who do the calculation, but it's very effective and it- and this is fine. If we can achieve certain tasks with high accuracy, it doesn't necessarily mean that it has to understand it the same way as we understand. In some ways it's even naive to request, because you have so many other sources of information, uh, that are absent when you are training your system. So, it's okay, as a dream, I said, and I will tell you one application that's just really fascinating. In '97 when I came to ACL, there were some papers on machine translation. They were like primitive, like people were trying... Really, really simple. And the feeling, my feeling was that, you know, to make real machine translation system, it's like to fly on the moon and build a house there and a garden and live happily ever after. I mean-

    7. LF

      Yeah.

    8. RB

      ... it's like impossible. I never could imagine that within, you know, 10 years we would already see the system working, and, and now, you know, nobody is even surprised to, to, to utilize this system on daily basis. So, this was like a huge, huge progress, things that people for a very long time tried to solve using other mechanisms and they were unable to solve it. That's why, coming back to your question about biology, that, you know, f- of, in linguistics people try to go this way and try to write the, the syntactic trees and try to obstruct it and to find the right representation, uh, and, you know, they, they couldn't get very far with this understanding, while these models using, you know, other-... sources, actually cable to make su- a- a lot of progress. Now, I'm not naïve to think that we are in this paradise space in NLP, and I'm sure as you know, that when we slightly change the domain and when we decrease the amount of training, it can do like really bizarre and funny thing. But I think it's just a matter of improving generalization capacity, which is, uh, just a technical question.

    9. LF

      (laughs) Wow. (laughs) So that's, that's the question, uh, how much of language understanding, uh, can be solved with deep neural networks? In your intuition, I mean, it's u- unknown I suppose, but as we start to creep towards-

    10. RB

      Mm-hmm.

    11. LF

      ... romantic notions of, uh, the spirit of the Turing test and conversation and dialogue, and something that maybe to, uh, to me or to us silly humans feels like it needs real understanding, how much, uh, can that be achieved with these, uh, neural networks or statistical methods?

    12. RB

      So, I guess I am very much driven by the, by the outcomes. Can we-

    13. LF

      Mm-hmm.

    14. RB

      ... achieve the performance which will be satisfactory for- for us for different tasks? Now if you'll again look at machine transition system, which are, uh, you know, trained on large amounts of data, they really can do a remarkable job relatively to where they've been a few years ago. And if you, you know, if you project into the future, if it will be the same speed of improvement, you know, this is great. Now, does it bother me that it's not doing the same translation as we are doing? Now, if you go to cognitive science, we still don't really understand what we are doing. Uh, I mean, there are a lot of theories, and there is obviously a lot of progress and studying, but our understanding what exactly goes on, you know, in our brains when we process language is still not crystal clear and precise that we can translate it into machines. What does bother me is that, um, you know, again, that machines can be extremely brittle when you go out of your comfort zone of the... and when it is a distributional shift between training and testing, and it have been years and years. Every year when I teach NLP class, you know, I show them some examples of translation from some newspaper in Hebrew or whatever, it was perfect. And then they have a recipe that Tomie Yakalo 　 system sent me a while ago and it was written in Finnish, of Karelian pies, and it's just a terrible translation. You cannot understand anything what it does. It's not like some syntactic mistakes, it's just terrible.

    15. LF

      Yeah.

    16. RB

      And year after year, I try it into Google Translate, and year after year, it does this terrible work because I guess, you know, the recipes are not big part of their-

    17. LF

      No.

    18. RB

      ... training repertoire.

  13. 57:15 – 1:05:42

    Turing test realism: ELIZA, human belief, and what “intelligence” might mean

    1. LF

      (laughs) So but, in terms of outcomes, that- that's a really clean, good way to look at it. I guess the question I was asking is, uh, do you think... Imagine a future... Uh, do you think, uh, the current approaches can pass the Turing test in the way, the- in the best possible formulation of the Turing test, which is, would you want to have a conversation with a neural network for an hour?

    2. RB

      Oh, God, no. (laughs)

    3. LF

      (laughs)

    4. RB

      No. There are not that many people that I would want to talk for an hour.

    5. LF

      (laughs)

    6. RB

      Uh, but-

    7. LF

      But there are some people in this world, alive or not, that you would like to talk to for an hour. Could a neural network of achieve that outcome?

    8. RB

      Uh, so I think it would be really hard to create a, uh, successful training set which will (laughs) enable it to have a conversation for an h- a- to conduct actual conversation for an hour. Uh-

    9. LF

      So you think it's a problem of data, perhaps?

    10. RB

      I think in some ways it's not a problem with data. It's a problem both of data and the problem of, um, the way we're training our systems, their ability to truly to generalize, to be very compositional. In some ways it limited, you know, in- in the current capacity. At least, you know, we can translate well, we can, um, you know, find information well, we can extract information. So th- there are many capacities in which it's doing very well, and you can ask me, "Would you trust the machine to translate for you and use it as a source?" I would say, "Absolutely." Especially if we're talking about newspaper data or other data which is in the realm of its own training set, I would say yes. Uh, but, uh, you know, having conversations (laughs) with a machine is not something (laughs) that I would, uh, choose to do. But you know, I will tell you something. Uh, talking about Turing test and about all this kind of ELIZA conversations, uh, I remember visiting Tencent in China, and they have this chat board, and they claim there is like really humongous amount of the local population which like for hours talks to the chat board. To me it was, "I cannot believe it," but apparently it's like documented. There are some people who, uh, enjoy this conversation. And do you know, it brought to me the, another MIT story about ELIZA and Weizenbaum. I don't know if you're familiar with his story. So Weizenbaum was a professor at MIT, and when he developed this ELIZA which was just doing string matching, very trivial, uh, like restating of what you said, with very few rules, no syntax, apparently there were secretaries at MIT that would sit for hours and converse with this trivial thing. And at the time, there was no beautiful interfaces, so you actually need to go through the pain of communicating. And Weizenbaum himself was so horrified by this phenomena that people can believe enough to the machine that you just need to give them the hint that machine understands you and you can complete the rest, that he kind of stopped this research and went into kind of trying to understand what this artificial intelligence can do to our brains. M- uh, so my point is, you know, how much... I- i- i- it's not how good is the technology, it's how ready we are to believe (laughs) that it delivers-

    11. LF

      Uh-huh.

    12. RB

      ... the goods that we are trying to get.

    13. LF

      Uh, that's a really beautiful way to put it. I- I, by the way, am not horrified by that possibility but inspired by it because...I mean, um, human connection, whether it's through language or through love-

    14. RB

      (laughs)

    15. LF

      ... it, uh, it seems like it's, uh, very amenable to machine learning and the rest is just, um, challenges of psychology. Like you said, the secretaries who enjoy spending hours... I would say, I would describe most of our lives as enjoying spending hours with those we love for very silly reasons. All we're doing is keyword matching as well. So I- I'm not sure how much intelligence we exhibit to each other in- w- with the people we love that are- we're close with. So it's a very interesting point of what it means to pass the Turing test with language. I think you're right, in terms of conversation. I think machine translation is a- has a very clear performance and improvement, right? What it means to have a fulfilling conversation is very, very person dependent and context dependent and- and so on. That's, uh, yeah, that's very well put. So but i- in your view, what's a benchmark in natural language, a test that's just out of reach right now, but we might be able to, that's exciting? Is it in machi- is it perfecting machine translation or is there other ... is it summarization? What's- what's out there just out of reach?

    16. RB

      I think it goes across specific application. It's more about the ability to learn from few examples for real, what we call few-shot learning and all these cases because, you know, the way we publish these papers today, we say if we have like naively we get, uh, 55, but now we had a few example and we can move to 65. None of these methods actually realistically doing anything useful. You cannot use them today.

    17. LF

      Mm-hmm.

    18. RB

      And, uh, their ability to be able to generalize and to move, uh, or to be a- autonomous in finding the data that you need to learn, uh, to be able to perfect new task or new language. Uh, this is an area where I think we really need to- to move forward to, and we are not yet there.

    19. LF

      Are you at all excited, curious, by the possibility of creating human level intelligence? Uh, is this- 'cause you've been very, in your discussion, so if we look at oncology, you're trying to, uh, use machine learning to help the world in terms of alleviating suffering. If you look na- nat- natural English processing, you focus on the outcomes of improving practical things like machine translation. But, you know, human level intelligence is a thing that uh, c- our civilization has dreamed about creating, uh, um, superhuman level intelligence. Do you think about this? Do you think it's at all within our reach?

    20. RB

      As you said yourself earlier, uh, talking about, you know, how do you perceive, uh, you know, our communications with each other, that you know, we're matching keywords and certain behaviors and so on. So and then whenever, uh, one assesses, let's say, relations with another person, you have k- separate kind of measurements and outcomes inside your head-

    21. LF

      (laughs)

    22. RB

      ... that determine, you know-

    23. LF

      Yeah.

    24. RB

      ... what is the status of the relation. And so one way, this is a classical dilemma, what is the intelligence? Is it the fact that now we are gonna do the same way as human is doing when we don't even understand what the human is doing? Or we now have an ability to deliver these outcomes, but not in one area, not in NLP, and not just to translate or just to answer questions, but across many, many areas that we can achieve the functionalities that humans can achieve with the ability to learn and do other things. I think this is ... a- and this we can actually measure how far we are, and, uh, that's what makes me excited that we, you know, in my lifetime at least so far what we've seen is like tremendous progress across these different functionalities. And, uh, I think it will be really exciting to see where we will be. And again, one way to think about it is there are machines which are improving their functionality. Another one is to think about us with our brains, which are imperfect, how they can be accelerated by this technology w- as it becomes stronger and stronger. Uh, coming back to another book that I love, Flowers for Algernon.

    25. LF

      Mm-hmm.

    26. RB

      Ha- have you read this book?

    27. LF

      Yes.

    28. RB

      Uh, so there is this point that it- then- the patient gets this miracle cure which changes his brain and all of a sudden, they see life in a different way and can do certain things better, but certain things much worse. So you can imagine this, uh, kind of computer-augmented cognition, where it can bring you is that now the same way as, you know, the cars enable us to get to places where we've never been before, can we think differently? Can we think faster? So ... a- and we already see a lot of it happening, uh, in how it impacts us. But, uh, I think we have a long way to go there.

  14. 1:05:42 – 1:08:48

    Augmented cognition and behavior feedback: from Neuralink to everyday nudges

    1. LF

      So that's sort of, uh, artificial intelligence and technology affecting our- augmenting our intelligence as humans. Yesterday, uh, a company called Neuralink, uh, announced ... they did this whole demonstration, I don't know if you saw it, it's, uh, they demonstrated brain-computer, brain-machine interface, where there's like a- a sewing machine for the brain. Do you, uh ... you know, a lot of that is, uh, quite out there, in terms of things that some people th- would say are impossible, but they're dreamers and want to engineer systems like that. Do you see, based on what you just said, a hope for that more direct interaction with the brain?

    2. RB

      Uh, I think there are different ways. One is a direct interaction with the brain, and again, there are lots of companies that work in this space and I think there will be a lot of developments.... when I'm just thinking that many times we are not aware of our feelings of motivation, what drives us.

    3. LF

      Hm.

    4. RB

      Like let me give you a trivial example, our attention. The- there are a lot of studies that demonstrate that it takes a while to a person to understand that they are not attentive anymore.

    5. LF

      Mm-hmm.

    6. RB

      And we know that there are people who really have strong capacity to hold attention, they're on other end of the spectrum, people with ADD and other issues and they have problem to regulate their attention. Imagine to yourself that you have like a cognitive aid that just alerts you based on your gaze-

    7. LF

      Mm-hmm.

    8. RB

      ... that your attention is now not on what you are doing, and instead of writing a paper, you are now dreaming of what you're going to do in the evening. So, even this kind of simple measurement things, how they can change us. And I see it even in simple ways with myself. I have my Zone app from, that I got at MIT gym. It kind of records, you know, how much did you run and you have some points and you can get, uh, some, uh, status, whatever. (laughs) Like I, (laughs) I said, "What is this ridiculous thing? Who would ever care about some status in some app?" Guess what? So to c- to ma- con- maintain the status, you have to do certain number of points every month. And (laughs) not only is it that I do it every single month for the last, uh, 18 months, it went to the point that I was running, uh, that I was injured, and when I could run again, I, (laughs) in two days, I did like some humongous amount of running-

    9. LF

      You tried to catch up?

    10. RB

      ... just to complete the points.

    11. LF

      Oh, no.

    12. RB

      It was like really not safe.

    13. LF

      Yeah.

    14. RB

      It was like, "I'm not gonna lose my status-

    15. LF

      Yeah. (laughs)

    16. RB

      ... because I want to get there." So, you can already see that this direct measurement and the feedback is, you know, we're looking at video games and see why, you know, the addiction aspect of it, but you can imagine that the same idea can be expanded to many other areas of our life, we, when we really can get feedback. And imagine in your case in relations, um, uh, when we are doing keyword matching, imagine that the person who is generating, uh, the keywords (laughs) that person gets direct feedback before the whole thing explodes. That maybe-

    17. LF

      (laughs)

    18. RB

      ... at this happy point, we are going in the wrong direction.

    19. LF

      That is so true. (laughs)

    20. RB

      Maybe it will be really, uh, behavior-modifying moment.

  15. 1:08:48 – 1:13:40

    Teaching machine learning: student struggles, prerequisites, and finding a mission

    1. LF

      So yeah, it's, uh, uh, relationship management too. So yeah, that's, that's a fascinating whole area of psychology actually as well, of seeing how our behavior has changed with basically all human relations now have other non-human entities, uh, helping us out. So you've, uh, you teach a, a large, a huge machine learning course here at MIT. I can ask you a million questions, but you've seen a lot of students. What ideas do students struggle with the most as they first enter this world of machine learning?

    2. RB

      Actually, this year was the first time I started teaching a small machine learning class-

    3. LF

      (laughs)

    4. RB

      ... and it came as a result of what I saw in my big machine learning class that Tommi Akala and I built maybe f- six years ago. Um, what we've seen that as this area become more and more popular, more and more people at MIT want to take this class. And while we designed it for computer science majors, there were a lot of people who really are interested to learn it, but unfortunately their background was not enabling them to do well in the class. And many of them associated machine learning with the word struggle and failure-

    5. LF

      (laughs)

    6. RB

      ... um, uh, uh, primarily for non-majors.

    7. LF

      Yeah.

    8. RB

      And that's why we actually started a new class which we call Machine Learning From Algorithms to Modeling, which emphasizes more the modeling aspects of it and focuses on, um, i- it has majors and non-majors. So we kind of try to extract the relevant parts and make it more accessible, because the fact that we're teaching 20 classifiers in standard machine learning class is really a big question do we really need it. But it was interesting to see this from first generation of students, you know, when they came back from their internships and from their, um, you know, jobs, w- what different and exciting things they can do that I would never think that you can even apply machine learning to. Some of them are like matching, you know, the relations and other things-

    9. LF

      (laughs)

    10. RB

      ... (laughs) like variety of different applications.

    11. LF

      Everything, everything is amenable to machine learning. You know, that actually brings up an interesting, uh, point of computer science in general. It almost seems, maybe I'm crazy, but, uh, it almost seems like everybody needs to learn how to program these days. If you're 20 years old or if you're starting school, even if you're an English major, it seems, it seems like programming unlocks so much possibility in this world. So in, uh, when you interacted with those non-majors, is there skills that they were simply lacking at the time that you wish they had in, that they learned in high school and so on? Like, how will they, how should education change in this computer- computerized world that we live in?

    12. RB

      See, because they knew that there is a Python component in the class, they, you know, their Python skills were okay, and the class isn't really heavy on programming. They primarily kind of add parts to the programs. I think it was more of the mathematical barriers.

    13. LF

      Mm-hmm.

    14. RB

      And the class, again, was a design on the majors, was using the notation like big O for complexity and others, uh, p- people who come from different backgrounds just don't have it in their lexicons, not necessarily very challenging notion, but, uh, they were just not aware. Uh, so I think that, you know, kind of linear algebra and probability, the basics, the calculus, multivariate calculus, are things that can help.

    15. LF

      What advice would you give to students interested in machine learning, interested, uh, y- you've talked about d- detecting, curing cancer, drug design. If they want to get into that field, what, what, what should they do, uh, g- get into it and succeed as researchers and, um...... entrepreneurs.

    16. RB

      Uh, the first good piece of news that right now there are lots of resources, uh, that, you know, are created at different levels and you can find online on, or with your school classes which are more mathematical, more applied and so on. So you can find and kind of, uh, preach it, which preaches your own language where you can enter the field and you can make many different types of contribution depending of, uh, you know, what is your strengths. Uh, and the second point, I think it's really important to find some area which you- which you really care about and it can motivate your learning and it can be for somebody curing cancer or doing self-driving cars or whatever, but to find an area where, you know, there is data, where you believe there are strong patterns and we should be doing it and we're still not doing it or you can do it better and just start there a- and see where it can bring you.

  16. 1:13:40 – 1:17:13

    Meaning, mission, and vanity: being true to yourself in research and life

    1. LF

      So you've, um, you've been very successful in many directions in life, but you also mentioned Flowers of Organon. And I think I read or listened to you mention somewhere that researchers often get lost in the details of their work, this is per our original discussion with cancer and so on, and don't look at the bigger picture, bigger questions of meaning and so on. So let me ask you the impossible question (laughs) of, uh, what's the meaning of this thing, of, uh, life, of, uh, of your life, of research? Why do you think we, descendant of great apes, are here on this spinning ball?

    2. RB

      You know, I don't think that I have really a global answer. You know, maybe that's why I didn't go into humanities (laughs) .

    3. LF

      (laughs)

    4. RB

      I didn't take humanities classes in my undergrad. But the way I'm thinking about it, that e- each one of us inside of them have their own set of, you know, things that we believe are important and it just happens that we are busy with achieving various goal or busy listening to others and to kind of try to conform and to be part of the crowd, uh, that we don't listen to that part. And you know, you- we all should find some time to understand what is our own individual missions and we may have very different missions and to make sure that while we are running 10,000 things, we are not, um, you know, missing out and we're putting all the resources to- to satisfy our own mission. And if I look over my time, uh, when I was younger, most of these missions, uh, you know, I- I was primarily driven by the external stimulus, you know-

    5. LF

      Mm-hmm.

    6. RB

      ... to- to achieve this or to be that and now a lot of what I do is driven by really thinking what is important for me to- to achieve independently of the external recognition. Uh, and you know, I- I- I don't mind to be viewed in certain ways. Uh, eh, t- the most important thing for me is to be true to myself, to what I think is right.

    7. LF

      How long did it take? How hard was it to find the you that you have to be true to?

    8. RB

      So it takes time and even now sometimes, you know, the vanity and the triviality can take-

    9. LF

      Of course.

    10. RB

      ... uh, you know-

    11. LF

      At MIT. (laughs) No, that's-

    12. RB

      Um, yeah, it- it can everywhere, you know, it's just the vanity at MIT is different, the vanity in different places-

    13. LF

      Yeah.

    14. RB

      ... but we all have our piece of vanity. But I think actually for me the- m- many times the place to- to get back to it, is, um, you know, when I- when I'm alone and also when I read. And I think by selecting the right books, you can get the right questions and learn from what you read, so. B- bu- but again, it's, uh, not perfect like-

    15. LF

      (laughing) Nothing is.

    16. RB

      ... uh, vanity sometimes dominates.

    17. LF

      Well, that's a beautiful way to end. Thank you so much for talking today.

    18. RB

      Thank you.

    19. LF

      It was fun. That was fun.

    20. RB

      Oh, it was fun.

Episode duration: 1:17:28

Install uListen for AI-powered chat & search across the full episode — Get Full Transcript

Transcript of episode x0-zGdlpTeg

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

iOSAndroidClaudeChrome