Boston Computation Club

Alex is a friend of mine and a mathematician. He did his PhD in mathematics with other Alex (Martsinkovsky), who has presented at the seminar group previously. We're huge fans of both Alexes at the Boston Computation Club and we hope you enjoy this talk as much as we did!

Today Daniel Yu gave a pure math talk on an extension he performed independently of work he had conducted during his math semester abroad in Budapest.

Today Woohyeuk (Kevin) Lee joined us to discuss local AI (and the question of just how local it really is). This is exciting work at the intersection of computer science and social science! Enjoy.

Casey Martin is a dirtbag intellectual, local savant, and recent PhD in the Boulder metro area. He's a party favorite and today he joined us to discuss cellular automata and their generalizations. Enjoy!

Zach Maas is an independent AI safety & mechanistic interpretability researcher in Boulder, Colorado, funded by Coefficient Giving. Today Zach joined us to talk about some of his recent work tracing introspection across model depth. This is, I think, the first mech interp talk we've hosted other than ChessGPT, and it was a good one! We hope you enjoy it as much as we did!

Today Peter McGrath An Extremely Short Proof of the Hairy Ball Theorem. Peter is a professor of mathematics at NC State, where he researches geometric analysis, minimal surfaces, PDEs. Today's talk was an elegant presentation of the classical result -- Peter makes it accessible, and explains each concept in a clear and transparent manner. This was a really fun talk and really back to our roots as a group, doing pure math on an iPad and loving it! We hope you enjoy the talk as much as we did!

Pascal Kesseli is a software engineer and technical lead at Microsoft AI, with a PhD in Computer Science from Oxford (or a DPhil? Or whatever they call PhDs on that side of the pond). Today Pascal joined us to discuss work he completed while at META FAIR, focused on the conjoining of large language models with symbolic reasoning systems (ultimately, dispatch to SAT) as well as future research directions building on said work.

Daniel Melcer is a PhD student at Northeastern University, where he researches formal methods, reinforcement learning, and large language models, among other things. Daniel also has the most colorful hair in the business (bright red for this talk, other colors for other occasions). Today he joined us to talk about some really exciting work he completed at Amazon, and to expand on his general vision of where constrained inference problems are heading in the future.

Max von Hippel is ... me, the organizer and founder of the Boston Computation Club. Today I hosted an extremely informal event to chat about some of the early experiments in FMxAI I was involved with at Galois, two years ago.

Chenfeng Wei is a PhD student at the University of Manchester, where he researches formal guarantees for large language models. Today he joined us to talk about his latest work exploring bugs in smart-contracts. This is a really interesting project at the intersection of explainable AI, smart contract debugging/security, and cybersecurity/symbolic analysis, and I hope you enjoy it as much as we did!

Today Xiao Mao joined us to discuss his groundbreaking work, Breaking the Sorting Barrier for Directed Single-Source Shortest Paths. It's not every day you beat Djikstra at something. This was a good one. Thank you for talking to us, Xiao!

David A. Noever and Forrest McKee are researchers at PeopleTec, where they work on problems at the intersection of security, defense, and AI/ML. Today David joined us to present their joint work Infecting Generative AI with Viruses. This was a really great presentation that took a rigorous approach to defining the security boundaries and limitations of AI tools, and it fostered one of the better discussions we've hosted in a while. We hope you enjoy it as much as we did!

Mingwei Zheng a Ph.D. candidate in the Department of Computer Science at Purdue University, advised by Prof. Xiangyu Zhang since 2021. Before that, she received her Bachelor’s degree in Computer Science and Technology from Huazhong University of Science and Technology in 2020. Mingwei's research lies at the intersection of Large Language Models (LLMs) and Software Engineering. She builds LLM agents that combine program analysis with LLMs for deep codebase understanding that improve software correctness, robustness, and trustworthiness. Today Mingwei joined us to discuss some ongoing work which she previously presented at IEEE Security and Privacy LangSec, applying language models to bug detection in protocols with IETF RFCs. This was a great talk with deep, technical content and a good conversation. We hope you enjoy!

Today Kaiyu Yang from Meta joined us to discuss formal reasoning using LLMs, particularly in the context of interactive theorem provers. This is a really fast-moving and exciting field in which reinforcement learning and theorem proving combine to provide a new frontier for fully automated reasoning, and Kaiyu is at the bleeding edge of it. We were really lucky to get an hour of Kaiyu's time and we hope you enjoy the talk as much as we did!

Michael Sammler n assistant professor leading the Programming Languages and Verification Group at the Institute of Science and Technology Austria (ISTA). Today he joined us to talk about his three primary projects: RefinedC, which uses a refinement and ownership type system to verify C code, Islaris, which shows how to scale verification of assembly code to realistic models of real-world architectures, and DimSum, which provides a decentralized approach for reasoning about multi-language programs (with a particular focus on RefinedC). We had a small but really dedicated crowd which facilitated an excellent discussion. This was a really fun one and we hope you enjoy it as much as we did!

Dimitri Mitropoulos is a Michigan-based typescript dev, linguist, and classicist who joined us to talk about his completely unhinged, odyssean, and frankly just unwise project to get DOOM running completely within Typescript's type system. Someone give the dude a PhD, please.

Jenna DiVincenzo is an Assistant Professor in the Elmore Family School of Electrical and Computer Engineering @ Purdue University. She is broadly interested in research spanning software verification, programming languages, and software engineering, especially research aimed at making verification techniques and programming languages more usable and scalable. Today Jenna joined us to talk about her broad research program in gradual verification. This was a really interesting talk with great Q&A and we hope you enjoy it as much as we did!

Prashant Anantharaman is a long-time BCC group member and has presented as both a solo researcher and a panelist to prior events. Today he joined us to present some of his work with NARF, to appear at IEEE S&P, on generating grammars for fuzzing using LLMs. This is a super exciting new frontier for LLMs and LangSec generally and his talk was wonderful.

Ramit Das is a formal verification engineer at Intel and an avid Boston Computation Club group member. Ramit and I have been speaking for ages about formal methods, exchanging papers, etc. and today he finally agreed to come give a talk to the group about his area of expertise -- descriptive complexity. This was a really fun talk and an excellent introduction for anyone looking to get their feet wet with complexity theory, some language theory, and even a smidgeon of model theory and underpinnings of abstract interpretation. It was really fun and we can't wait to host another talk by Ramit sometime in the future!

Today Marc Denecker joined us to present How and Why to extend First Order Logic for Knowledge-Based Systems. This presentation provided the setup for a follow-on that Marc's student Simon Vandevelde is set to give on IDP-Z3, a formal reasoning machine that Marc and Simon have built. This was a really interesting talk touching on a variety of forms for formal logic, decision procedures, and industrial use-cases thereof, potentially with profound implications for the future and realizability of so-called AGI.

Today Daniel Melcer joined us to present Constrained Decoding for Code Language Models via Efficient Left and Right Quotienting of Context-Sensitive Grammars (https://arxiv.org/pdf/2402.17988). This is work he completed while at Amazon, and it's a really interesting project around how to constrain, guide, and check language models such that they generate valid code within a given context. We really appreciate that Daniel took the time to talk to us and hope you like the talk as much as we did!

Michael H. Borkowski is an Assistant Professor of Practice in the Department of Computer Science at Purdue University. Before joining Purdue, he earned his Ph.D. from the Department of Computer Science and Engineering at UC San Diego, where he was affiliated with the ProgSys Group. Today Michael joined us to discuss LiquidHaskell, a very cool project that incorporates a kind of refinement types, with SMT-based proofs, into Haskell. This was a really compelling talk and we hope you enjoy it as much as we did!

Today Brook Santangelo and John Sterrett

Gaspard Baye is a Cyber AI Ph.D. Candidate at the University of Massachusetts Dartmouth, where he researches AI-driven offensive and defensive security applications. Today Gaspard joined us to present "Hacking GenAI with LLM Red Teaming and Beyond" based on his recent DefCon talk. This was a really fun event with a great Q&A. Thanks to Jacob from the Trust Lab for hosting!

Navid Hashemi recently defended his PhD at USC and is about to begin a post-doc at Vanderbilt. His research focuses on the intersection of Artificial Intelligence and Temporal Logics, with applications in Formal Verification of Learning Enabled Systems and Neurosymbolic Reinforcement Learning. Today Navid joined us for a really exciting presentation about his work on metrizable logics for reinforcement learning, and a technique for verification thereof based on the over-approximation of reachable sets using ReLU.

Chengpeng Wang working with Prof. Xiangyu Zhang. His research focuses on program analysis, especially software analysis, and in particular how existing analysis techniques intersect with emerging approaches from AI such as Large Language Models. Today Chengpeng joined us to talk about his recent NeurIPS paper proposing a novel static analysis technique based on LLMs. The technique is very interesting and highly informed by prior works in the static analysis space, but leverages LLMs as a kind of "oracle" to solve problems which, when handled statically, quickly become untenable. This was a really interesting talk and we're very greatful that Chengpeng took time out of his Sunday afternoon to talk to us!

Harry Eldridge is a Cryptography PhD student at Johns Hopkins, advised by Abhishek Jain and Matthew Green. His research (so far) touches on security and privacy implications of commodity hardware, which is a fascinating topic deserving of the mathematically disciplined, cryptographically informed approach his lab takes to such problems. Today Harry joined us to talk about his research into the problem of AirTag stalking, and how it can be ameliorated, while retaining acceptable performance, through cryptographic protocols. This was a very interesting talk with serious, real-world implications, and we hope you enjoy it as much as we did!

Ian Bicking is an engineer at Brilliant, which is also what he is. (Sorry, dad joke). Ian joined us today to talk about his super charming (and extremely interesting) weekend of experiments hacking various LLMs to solve puzzles using z3. The presentation was roughly the first 2/3 of the event and the remaining third presented a fantastic conversation about the future of AI, tool use, chain and tree of thought, o1, and more. Thanks again for joining us Ian!

Dhekra Mahmoud at LIMOS in Clermont-Ferrand, France, where she researches the formal analysis of cryptographic protocols under the supervision of Pascal Lafoucade and Jannik Dreier. Today Dhekra joined us to present her recent USENIX paper Shaken, not Stirred -- Automated Discovery of Subtle Attacks on Protocols using Mix-Nets. This was a really interesting presentation with a good conversation afterword touching on some subtler points around the Dolev-Yao threat model, the limitations of ProVerif, and proof optimization.

Matej Panciak holds a PhD in mathematics and is a software engineer at the Argument Computer Corporation, where among other things, he works on Lurk. Lurk is a LISP for defining computations that can prove (in the ZKP sense) that they ran, which is probably useful for all sorts of cool things we haven't thought of yet, but right now, is pretty important for doing stuff on-chain. (I can easily imagine this being applicable to building something like a dweb version of AWS ... in some theoretical future where FHE is so good that you can just trust randos to run code for you). Anyway, Matej presented a super rad intro to Lurk, gave us a code demo (it worked!) and then enjoyed our usual über-nerd conversational segment at the end.

Derek Egolf is a PhD student (since 2021) at Northeastern University, advised by Stavros Tripakis. His primary research focus is the automatic generation of correct-by-construction systems from high-level specifications (synthesis). Today Derek talked about his recent paper in this vein, Efficient Synthesis of Symbolic Distributed Protocols by Sketching, to appear in FMCAD. This was a very interesting talk with a technical conversation afterword. We hope you enjoy it as much as we did!

Joshua Ramette (https://x.com/RametteJoshua) recently completed a PhD in physics at Mass Tech, and today he joined us to talk about his new project, Undermind. Josh and his friend Tom Hartke (https://www.tomhartke.com/) founded Undermind (YC S24) to radically improve academic literature search using a mixture of AI techniques. Their system is slow, deliberate, and very high quality. You can check out Undermind at www.undermind.ai , or peruse the query I did during the Q&A section here: https://www.undermind.ai/query_app/display_one_search/c743b66ee4378b12ae8bad1fe58975ba95da71e7f7a1d5c2f0c6973c677648cc/

Evan Pu ( https://evanthebouncy.github.io/ , @evanthebouncy on X ) is a senior research scientist at Autodesk AI Lab, working on code-generation for human-machine collaboration in CAD, and industry scale instruction-following dataset annotation. Today Evan joined us from a toilet (with the lid closed) so as not to wake up his wife due to a rather large time-zone delta, which was hilarious and a first for the Boston Computation Club. Anyway, this was a really fun talk with excellent Q&A and we hope you enjoy it as much as we did!

Arthur O’Dwyer is a C++ programmer and blogger who today joined us to talk about his musings on the algebraic structure of the popular web-game Infinite Craft. Infinite Craft is a clever little experiment in sandboxed exploration, and it turns out to give rise to a rather complex mathematical structure with some interesting background in theoretical CS. Arthur covered all this and more in his presentation, which was super interesting and a lot of fun to watch. Check out Arthur's original blog post here: https://quuxplusone.github.io/blog/2024/03/03/infinite-craft-theory/ Check out Arthur's slides here: https://bstn.cc/artifacts/arthurODwyer/infiniteCraft.pdf

Evan Boehs is a HIGH SCHOOL STUDENT who broke the freaking internet. What more do I need to say? Hire this kid. Maybe I will. It's a race. Evan made an npm package called everything which installs everything. Then he got stuck in a dependency loop when someone tried to delete something. It turns out this is a nearly impossible problem to solve and he totally broke npm. Then a bunch of adults got made at him, when really, they should have been mad at themselves for building a bad system. You can read Evan's full story here: https://boehs.org/node/npm-everything

June Marcuse

Adam Karvonen was my coworker at Galois and is a bright guy doing really interesting stuff in the ML interpretability space. Today he joined us to present his work on Chess-GPT, you guessed it, a GPT model that can play chess. The punchline isn't so much how good the model is as it is how the model "thinks" -- Adam provides compelling evidence that the model internally reasons about an actual board state, and learns to make legal moves. The discussion on this one was great and we really appreciate that Adam took the time to talk to us! Also -- you should hire him! He's doing MATS but will be on the job market at the end of the Summer.

Max Ammann is a cybersecurity researcher at Trail of Bits, where he's recently been working on extending his Master's thesis work on fuzzing cryptographic protocols into an industrial-grade fuzzing tool. That work resulted in an S&P publication which is what he joined us to present today. This was a really good talk but also a great discussion, in large part because of the highly engaged audience (with representation from Galois, TwoSix, and academia!).

For this event, Holmes Wilson of Fight for the Future moderated a panel retrospective on the Pegasus malware. Our panelists were: - Jonathan Rugman: Foreign Affairs Correspondent at Channel 4 News, BAFTA Award-winning journalist, visiting lecturer at University of London, and Senior Associate Fellow at RUSI. - Raya Sharbain: education and communities coordinator at the Tor Project, and digital rights activist at the Jordan Open Source Association as well as the Digital Arabia Network. - Elina Castillo Jimenéz: feminist human rights lawyer and digital activist at the Amnesty International Tech Lab. - Prashant Anantharaman: former speaker at the club who completed his PhD at Dartmouth under Sergey Bratus and now works at Narf Industires. And … - Hinako Sugiyama: international human rights lawyer and law professor at UC Irvine. This was one of our best events ever and well worth the listen.

Mathias Preiner is a Research Scientist at Stanford University in the Centaur lab. He is one of the main developers of the SMT solver Boolector (since June 2012) and Bitwuzla -- which is what he joined us to discuss today. This was a good talk, but an excellent Q&A, and we really enjoyed it. Thanks Mathias for joining us today, and to the awesome audience for showing up with such deep and technical questions!

Joe Shiraef is a professional card counter and indie game dev. Today he joined us for a very fun, free-form conversation on advantage play, indie game development, avoid arrest, and pursuing your passions. https://www.inktalestudios.com/

Today puzzle-maker Roger Barkan joined us to talk about the creation and solution of cave puzzles, a category of puzzle for which he's quite famous as a puzzle author. Jacob lead the conversation, using an interactive puzzle that he implemented with the help of ChatGPT (:0), and it was a ton of fun. We're super grateful to Roger for joining us today and we look forward to doing a follow-up event sometime in the future! Jacob's interactive: https://bstn.cc/artifacts/jacobDenbeaux/cave.html Buy Roger's book (you know you want to!): https://www.barnesandnoble.com/w/colossal-cave-collection-roger-barkan/1125542215

Jan Hennecke is an engineer and roboticist in Boston, MA. Jan has been a buddy of mine for ages, ever since we met at the Bernardo Faria Jiu Jitsu Academy where he told me a hilarious story about placing top-3 in his first half ironman while munching down on snickers. Today Jan joined us to talk about his work at RBTX, a marketplace and platform for low-cost automation. This was a really fun talk with a lot of audience engagement and I think many of you will find it interesting!

Today Christian Williams joined us to talk about his dissertation project, Logic in Color. This is a really exciting project which he is now working on post-graduation, which aims to re-frame the way we think about logic, and logics, using a largely visual medium. The key insight is that certain mathematical observations are made completely obvious simply by adding color to the areas enclosed by arrows in monoidal string diagrams. But from this key observation comes the more foundational view that really, all of mathematics and logic not only can be expressed visually, but in some sense, perhaps _is_ visual; that the medium is exposing something fundamental about the nature of thought itself. This sounds a little pretentious but it's actually just the opposite: it's a fairly radical effort to _simplify_ logic and category theory using a visual medium. And it's enormously exciting. We were really happy Christian gave us this ground-floor view on his project and we're super excited to see where it develops.

Todd Schmid an Assistant Professor in the Computer Science Department of St. Mary's College of California. They are generally intereted in the algebraic, coalgebraic, and logical foundations of program semantics, and recently completed a PhD as a part of the PPLV group in the Computer Science Department of University College London. Today Todd joined us to talk about coequations, a fascinating (categorical) subject relating to the how we add algebraic structure to a space, how we think about relationships between spaces, and more. It turns out that coequations show up all over the place -- in DFAs, Markov chains, various PL concepts, etc. -- and so this is a place where the more abstract categorical stuff turns out to be really useful and illuminating for fairly concrete computer science ideas. Plus, coequations are just plain neat! We were really lucky to steal a little over an hour of Todd's time on this beautiful Saturday and we hope you enjoy the talk as much as we did.

Avijit Ghosh is a Research Data Scientist at AdeptID and a Lecturer in the Khoury College of Computer Sciences at Northeastern University. He's a good friend of mine and was an element of my PhD cohort at Northeastern. He's also a well-respected researcher at the intersection of machine learning, ethics, and policy. You can read about some of his innovative and cross-disciplinary work, for example, in the New York Times. David Widder is a Postdoctoral Fellow at the Digital Life Initiative at Cornell Tech, and earned his PhD from the School of Computer Science at Carnegie Mellon University. If you know me personally, you might remember David because he and I were simultaneously involved in parallel antics to fight non-consensual workplace sensors at CMU and NEU, respectively. Another funny coincidence is that David and I attended the same international boarding school program, called United World College. But most importantly -- David is a first-class researcher in the space of AI ethics. Fabio Tollon is a…

Eli Sennesh is a recent graduate of the PhD program in computer science at Northeastern, in which I (Max) and many other BCC group members are currently enrolled. Eli's research is highly interdisciplinary, taking into consideration various topics in mathematics (statistics, measure theory, probability theory, optimization), programming language theory, and neuroscience, with the unifying goal of building useful probabilistic programming languages. Today Eli joined us to discuss that research, with a particular emphasis on important open problems -- problems which he intends to study as a post-doc! This was a fun one and an excellent introduction to the world of probabilistic programming, and we really appreciate that Eli took time out of his weekend to come talk to us. Eli's website: https://esennesh.github.io/ Eli's advisor's book on probabilistic programming: https://arxiv.org/abs/1809.10756

Bill Dalessandro is a philosopher of science and mathematics at Oxford University. Today Bill joined us to discuss proofs -- specifically, what it means for a proof to be fixable, what it means for a proof to be transferable, and the apparent tension between these notions. This work built on prior work by Northeastern's Don Fallis, who attended the talk and participated in the lively and fascinating conversation that ensued. We also discussed what it's like to work in an interactive theorem prover. In such an environment, you don't really make mistakes -- because the prover doesn't let you -- but you might prove the wrong thing, and/or, you might not learn much despite having proven something. This was a great talk with a great with a really strong discussion section and we hope you enjoy it as much as we did! - Bill's website - The paper in question

Today Nathaniel Mitchell and Dan Scott joined us from Intel to discuss the ChipSec project, an open-source platform security assessment framework, available at https://github.com/chipsec/chipsec . Specifically, ChipSec "is a framework for analyzing the security of PC platforms including hardware, system firmware (BIOS/UEFI), and platform components" -- for both Windows and Linux (although as we discuss, getting it to work on Windows requires some leg-work). This was a really interesting talk and it included a very impressive demo! We learned a lot and we're very thankful that not just one but two busy engineers from Intel took the time to talk to us today about their fascinating software tool.

Bernardo Anibal Subercaseaux Roa Marijn Heule. He has a background in engineering and is passionate about mathematics and computer science. Bernardo's research attacks the following question from a variety of angles: what can and cannot be done (efficiently?) by a computer? Today, Bernardo joined us to talk about Packing Chromatic, a fascinating research area at the intersection of pure mathematics and SAT solving. Bernardo and his advisor recently solved an open problem in the space, finding the packing chromatic number for the infinite 2D grid. We discussed the proof and a bunch of related problems, including some intriguing (and open) questions about periodic vs aperiodic tilings. By the way, this work was also covered (briefly) in the NYT! (Not a lot of 2nd year PhD students get covered in the grey lady!!)