Back to all posts Next post — Data Privacy, Palm Oil, and Petty Consumerism

Notes from Heidelberg: Talking with Laureates

Updated on Oct 16, 2019   •   Anmol Kabra   •   experience  •  cs  •  climate-change  •  conference   •  

I can’t believe that two weeks have passed since I returned from the 7th Heidelberg Laureate Forum (HLF). After wading through a tricky two-week stream of assignments and prelims, I was finally at leisure to introspect and clear my head, thanks to Fall Break (classes were off for the last two days). While I have solely utilized the previous Fall Breaks to read ahead, I thought I should also use this Fall Break to reflect on the recent past. As I was reading through my notes, I found several pages of incoherent bullets scribbled when I was at the HLF. Among those notes were also recounts of in-person interactions I had had with the laureates. In this blog post, I want to formalize and share those experiences.

Me @ HLF
Me @ HLF, credits to Ham

The HLF was a unique setting, where 200 young researchers (undergraduate, graduate, postdoctoral, and junior researchers) from all over the world could interact with laureates in Computer Science and Math (awardees of Abel Prize, ACM Turing Award, ACM Prize in Computing, Fields Medal, and Nevanlinna Prize). In the 1-week Forum, we attended lectures from laureates, workshops, and panel discussions — AI, Machine Learning, and Climate Change were the most widely topics. On Wednesday, the organizers had planned visits to different CS and Math institutes for the young researchers and a boat trip on the Neckar river to encourage informal interactions, followed by a boat party. We could also informally interact with the laureates during meals. The dinners were particularly extraordinary, in my opinion, as the good food was supplemented with short programs and opportunities for 1-on-1 chats with the laureates. Moreover, the every dinner was held at a different exotic location — we ate at Halle02, Speyer Museum of Technology, Weldegarten (theme: Bavarian Evening), and the Heidelberg Castle.

At the end of every day, I would recollect my conversations with the laureates on paper, in an effort to engrave the discussions in stone. Yesterday when I revisited my notes and relived the HLF experience, I realized that the daily efforts had reaped benefits. While I will recount my discussions with the laureates in this post, I cannot emphasize enough the importance of meeting like-minded students. Getting to learn from the young researchers has had a remarkable impact in my outlook. Alas, I’m incapable of describing it in a public blog. Keeping this in mind, I’ll get going. I’ll use the structure of my notes and discuss the conversation I had with laureate A in a section named “A”. The sections are arranged in order of the day I took the corresponding notes.

Edvard Moser

Edvard Moser
Edvard Moser

The Nobel Prize in Physiology or Medicine 2014, “for their discoveries of cells that constitute a positioning system in the brain”.

After Yoshua Bengio gave the Turing Lecture on Deep Learning and AI, Moser gave the Lindau Lecture (lecture by a Nobel Prize awardee at the Forum) on his group’s work on the brain’s positioning system. He talked about how the mammalian brain had evolved for orienting animals in the real world by creating intricate structured grids of neurons in the Hippocampus and Entorhinal Cortex regions of the brain. For example, when a rat was placed inside a room, a distinct set of neurons would fire at coordinate (0, 2) in the room than the set of neurons firing at coordinate (2, 3). This illustrated how those neurons helped the animal identify the position and orientation in the room (there were specialized neurons for directional orientation as well).

After the talk, I asked him whether these grids of neurons and their activation patterns persisted in imagination and dreams, when the animals likely didn’t have a real-world frame of reference substantiated by vision. He strongly believed that the grids and firing patterns persisted in imaginations, as the brains could construct a virtual reality very similar to actual reality that gave animals the ability to associate with their surroundings in ‘thin air’. He remarked, however, that experiments to test this hypothesis are tricky and he wasn’t aware of such experiments.

When asked about how these grids were created in the Hippocampus, he said that most of the creation took place by the time of birth and ‘learning’ wasn’t prominent during the life. He pointed to the common grid structures across very different species, indicating that evolution had had a key role in building the brains’ positioning systems. I associated this comment to the ongoing research in robot positioning systems, which many researchers argue must be self-learned from scratch. Moser’s answers indicated how evolution had ‘pre-trained’ or pre-established several features of the brain, while allowing for learning other features during animals’ lives. In the context of the Forum’s focus on AI and Machine Learning, these lessons from a Neuroscientist were very interesting!

Vinton Cerf

The ACM Turing Award, with Robert E. Kahn, for pioneering work on internetworking, including the design and implementation of the Internet’s basic communications protocols, TCP/IP, and for inspired leadership in networking.

Finding Cerf outside the lecture hall was a great opportunity to talk about his recent critiques of internet governance policies, cybercrime, and malicious actors. I asked him what he thought would be essential to reduce misinformation and cyberattacks, and he described both technological and social modifications in detail. I learned about the importance of Border Gateway Protocol, which streamlines and standardizes the routing of identities and IPs among internet nodes, and fingerprinting mechanisms, which are needed to verify internet actors’ identities. He also pointed out the superiority of multi-factor authentication techniques in verifying users’ service accesses, while complaining about the annoying picture-captcha that are seldom easy on the first try. Very relatable! He also briefly mentioned how SSL certificates (and so HTTPS) had been instrumental in establishing trust (in the technological meaning) between servers and clients.

On the social side of solutions, he ideally hoped that humans would cooperate and develop a common set of moral values when using the internet, but remarked at the sheer complexity of the task. He was, however, hopeful of comprehensize government policies and better moderation on social networking websites and news media, which he thought played key roles in disturbing the existing democratic institions across the world. Even after safeguarding the internet infrastructure with BGP, authentication mechanisms, HTTPS etc., he still claimed that users needed to be better educated in using the internet, and that services were as much responsible as everyone else in maintaining the digital world.

Richard Stearns

The ACM Turing Award 1993, with Juris Hartmanis, in recognition of their seminal paper which established the foundations for the field of computational complexity theory.

During lunch on the first day, I sat beside Stearns and listened to his stories about finding research problems and working on the complexity class hierarchy. He recounted how Hartmanis (co-awardee) and he worked together at General Electric in the 1960s on problems that seemed interesting at that time. When I asked if he thought the work on complexity class hierarchy would become so influential, he amusingly said that he didn’t even consider that work his favorite or the most challenging. The paper was merely a result of proving many small results here and there.

He also celebrated his friendship with Hartmanis, and commented that they still remain good friends and visit each other. Stearns said that he recently visited Ithaca and didn’t mind getting snowed in as he was used to such weather, having lived in Albany.

Ivan Sutherland

The ACM Turing Award 1988, for his pioneering and visionary contributions to computer graphics, starting with Sketchpad, and continuing after.

I got the rare chance of sitting beside Sutherland during dinner on the first day. I was amused that he was flaunting a light-blue UNC-Chapel Hill hoodie while everyone else was in business casual attire; when someone at the table asked if he was a UNC alum, he clarified that he was wearing it because his friend had gifted it.

Since I was not well-versed with Sutherland’s work, I asked many questions about his graphics work, his recent interests in computer architecture, and his time at various companies between 1970 and 2000. He said that he liked computers from his early days, and always wanted to play with them. He was interested in building something tangible, which led to Sketchpad and the head-mounted display, the first VR headset. His primary reason for continuing to work on graphics and displays was because it was “cool and exciting”. Cool.

I asked him why he switched to computer architecture, and he exclaimed, “Why not?”, and waited for me to answer. I was taken aback and quickly replied that I thought graphics and architecture to be very distinct subfields. He smirked, asked me what the speed of light and the size of a computer chip was, and started to identify challenges in packing logic gates in chips. I was utterly baffled as no professor had used such seemingly unrelated stats to motivate a problem statement before; I briefly thought he wasn’t a good teacher.

We briefly discussed his time at RAND corporation and Sun, and he shared a story about Eric Schmidt of Google. It was funny, but I can’t remember it now.

During the rest of the dinner, he talked about arbitrary physical phenomena to emphasize the importance of curiosity in science: why plywood has an odd number of layers; why balloons stay afloat above candle-light. In my opinion, he was by far the most eclectic person to talk to at the Forum.

Raj Reddy

Raj Reddy
Raj Reddy

The ACM Turing Award 1994, for pioneering the design and construction of large scale artificial intelligence systems, demonstrating the practical importance and potential commercial impact of artificial intelligence technology.

On the second day, I met Reddy at the dinner table in the Speyer Museum. He talked about his recent big project of convening scientists and engineers in India to build a mobile application that could immediately translate one spoken Indian language to another. He was very excited about the impact this technology would have in India, which has tens of languages and hundreds of dialects. The app would ease communication across linguistic and state borders, which will be highly influential in boosting research, business, and the economy. He also talked about the technological challenges involved, such as collecting speech data from different speakers in different regions and improving the current state-of-the-art to work with speech translation on non-Latin derivatives.

Some of the other young researchers also asked him for advice for graduate students, and he emphasized networking, communication, and presentation skills on top of academic proficiency. He pointed out that academic proficiency is straightforward to obtain from textbooks, courses, and a willingness to learn concepts, whereas communication skills can only be learned through real-world interactions that demand practice and time.

Jeff Dean

Jeff Dean
Jeff Dean

The ACM Prize in Computing 2012, with Sanjay Ghemawat, for their leadership in the science and engineering of Internet-scale distributed systems.

I met Dean on the boat, after the young researchers had visited local institutions and the laureates had visited local schools. He described his trip to the schools, where he attempted to teach neural networks to students without using calculus. I asked him how he explained backpropagation to them, and he said that he gave them a high-level understanding of how each layer gets updated using dependencies in the computation graph. He said he wasn’t so sure if the students understood it very well, so he moved to answering their questions. Many school students, it seemed, were concerned about their career in an automation-centered future, and Dean argued that many of the jobs that will get automated will be replaced by new opportunities. He also mentioned how careers that involve creativity and imagination on the part of individuals will remain in demand.

Following this discussion, we went to the lower deck of the boat to get drinks and we started discussing misinformation and Google’s actions in this sphere. He said that Google had many automated solutions to counter misinformation on websites, one being the reduced page rank Google Search assigned to persistently malicious websites. Although he couldn’t describe more strategies in detail, he acknowledged the difficulty of countering misinformation when spread through visual media like videos rather than text, the former being harder to automate monitoring systems for.

I also brought up the New Yorker article about him and Ghemawat’s love for pair programming, and he became visibly excited. He joked about how their coding and thinking styles were almost orthogonal but complimentary. They were always acceptive of criticisms from each other, which made the coding sessions fun and productive. He highly recommended pair programming to any pair who could work together. When I asked how to find someone willing to work together for a long-term when students have to change schools regularly, he contended that the pair need not last several years; even pair programming on small projects can help build understanding and produce efficient code. I appreciated how willing Dean was to share his personal-professional relationship with Ghemawat openly.

Vinton Cerf, again

Vinton Cerf (I took this pic after this chat)

During the boat trip, I met Cerf again at a table with many other young researchers. Someone asked him to share his motivations of creating the internet in the 60s, to which he replied that the internet and its design infrastructure were a product of strict requirements and resources set by the existing technology and DARPA. Most of the design decisions were a result of decomposing the network connectivity problem into smaller tasks and realizing the resource and research bottlenecks to solve those tasks. At the onset, he personally didn’t have a big vision for the project and wasn’t anticipating the internet to become mainstream in the future. He said that he didn’t know the idea of the internet would stick around; he claimed that it is impossible to know if anything would stick! He advised us to work on problems that we found interesting and not worry too much about the hypothetical future impact.

When asked about his recent interests, he said he had become interested in microbiology, especially how RNA controls the messaging systems for transporting nutrients and information inside cells. He wonders if there are analogies between the intra-cell communication networks and the internet. He also wants to better understand neural networks’ inner-workings and hopes that the research community will work on this.

Stephen Cook

Stephen Cook
Stephen Cook

The ACM Turing Award 1982, for his advancement of our understanding of the complexity of computation in a significant and profound way. His seminal paper, “The Complexity of Theorem Proving Procedures,” presented at the 1971 ACM SIGACT Symposium on the Theory of Computing, laid the foundations for the theory of NP-Completeness. The ensuing exploration of the boundaries and nature of NP-complete class of problems has been one of the most active and important research activities in computer science for the last decade.

I briefly chatted with Cook during a coffee break on Thursday, and asked him about his current research and his motivations to work on SAT and NP-completeness. He said that he had retired now, but still does some research in theoretical CS. He said that many researchers were working on SAT and computational complexity, in general, during the 60s and 70s, and so he worked on similar problems. He had liked problems in theoretical CS throughout his life, and he thought that coming up with the SAT reduction would be a good example of a framework that could help other researchers. As expected, he didn’t anticipate his work to take the center-stage of CS.

Robert Tarjan

Robert Tarjan
Robert Tarjan

The ACM Turing Award 1986, with John E Hopcroft, for fundamental achievements in the design and analysis of algorithms and data structures.

I met Tarjan on Friday during a coffee break, and had a short discussion about finding research problems and looking for solutions. When asked about how to confidently pick good problems from the large basket, he said that practice and experience played an important role in honing that skill; there was no other real way he knew of. Another young researcher seeked his views on handling multiple research projects. He endorsed the idea of concurrently working on problems as a way to stash temporarily unassailable problems, thus helping researchers in two ways:

  1. Working on another project when stuck on one allows the researcher to make progress.
  2. When coming back to the stashed problem, one might be able to think of another solution, thus making progress.

Stephen Smale

Stephen Smale
Stephen Smale

Fields Medal 1966, worked in differential topology where he proved the generalized Poincaré conjecture in dimension n>=5: Every closed, n-dimensional manifold homotopy-equivalent to the n-dimensional sphere is homeomorphic to it. Introduced the method of handle-bodies to solve this and related problems.

Smale gave a lecture on mathematically modeling the heartbeat on Thursday, and I got to meet him just before Friday’s dinner at the Castle. He was clearly interested in Biology, and when asked why, he said that he found it interesting and “cool”. I asked him if people from Math and Biology appreciated this work, and he said that a few from both sides did. I had always wondered how researchers doing interdisciplinary work fought ‘traditional’ scientists constantly trying to categorize the work into buckets, and he, interested in Math and Biology, recalled that he would often ignore the traditionalists but draw on research from all fields to take his work forward.

I hadn’t internalized how old he was (89!) until we talked about his advisees, two of which, Jim Renegar and John Guckenheimer are now very senior professors at Cornell. Since these advisees did more Applied Math research, I inquired about his experiences of advising Applied Math students when he himself had done early work in Pure Math (he won the Fields Medal for a work in topology). He said that he didn’t think either he himself or his advisees minded this disconnect as they would only meet a few times in a week, which wasn’t enough to tangibly alter interests. Moreover, he didn’t believe in these divisions of Pure and Applied Math! It was fun talking to him.

By the end of the Forum, I was both energized and disappointed. Energized from meeting so many amazing young researchers and interacting with the brilliant laureates, and disappointed because I couldn’t meet more. I also couldn’t take pictures with everyone. The Forum refreshed my love for CS research, and I left a very motivated student. All laureates emphasized the freedom in and love for research one ought to have to succeed. No laureate I talked to anticipated the impact their work would have on the world, and I gathered that none cared either. At the end, they were looking to solve problems that they found ‘cool’ and interesting.

If you’re interested in CS/Math research and are in a graduate program/postdoctoral position/senior undergraduate standing, I will highly recommend applying to attend the Forum. It is an opportunity of a lifetime.

I want to thank Cornell’s CS Department for funding my travel and HLF for funding everything else — I’m grateful for their support.