Sunny and 50
Woke up a bit later (7:30) intentionally, since practice tonight is at 10pm. Started reading one of the papers Vinod recommended. Very cool. One of the coauthors is Zvika (on our STOC paper), and another is Paul Christiano (of AI safety fame, at OpenAI). Small world! Skipped meditation and workout; leisurely breakfast with Libby. Incredible cinnamon roll. Walked up the hill with Libby and Emily. Weather was insane--balmy, perfect, sunny. At Gates, cleaned up my latest analysis. Plots looking great, feeling like a legit programmer/data scientist. Tossed with Martin, Hao, and Simon. Brown bag lunch where I learned about the Cornell CS wiki. Awesome resource but totally out of date. Had a bit of an edit party. Then went to Visit Day czar meeting. I was popping off, so full of ideas about how things could be done well this year. Overall today I enjoyed a completely insane level of mental awareness, focus, and context-switchiness. If I can keep this up... Back to work, wrote an update email to the Recursive squad, and wrote up a little email to goodnews@cis.cornell.edu about the STOC submission. An excerpt:
I have some good news! A team informally led by my advisor Noah Stephens-Davidowitz at Cornell had our work, "Lattice Problems Beyond Polynomial Time", accepted to the ACM Symposium on Theory of Computing (STOC) 2023! You can see the full list of authors and affiliations below. Here is a short summary: In the paper, we study lattice problems (central to the design and security of next-generation "post-quantum" cryptography) in the practically relevant regime where adversaries have access to an amount of computational resources that grows exponentially, but not too steeply, on the security parameter--much less than the steep exponential amount believed necessary to solve these problems. Among other things, we show that the security of private- and public-key lattice-based cryptography can be based on slightly different and potentially preferable assumptions than those currently made. At a high level, we show security assuming the worst-case hardness of harder lattice problems (problems with harder parameters than in previous assumptions); however, hardness must be assumed even for adversaries that run in "small" exponential time (that is, more powerful than the typically assumed polynomial time adversaries). Thus our new hardness assumptions are not strictly weaker/better than the hardness assumptions currently made. But, since in practice we are arguably most concerned with adversaries that have a "small exponential" amount of computing resources, it seems useful to gain something (basing security on the hardness of harder problems) from assuming such adversaries.
Enjoyed a little random CS social with food, talked about my research with Surendra, and generally goofed around. Chatted with Eva a little bit. Left work a little early (4:50ish), then did my back exercises at Keeton, then went to the cemetery in my Vibram barefoot shoes (I had them on all day) and did a little light jogging. Felt really nice on the soft ground, and should strengthen my feet. Shower, then ran to House Dinner, which featured delicious brisket. I sat with a student-athlete baseball player named Kirby and met some of his buddies. Back to Keeton, practiced a little guitar, took a 15 minute nap, wrote this log, and off to pick up my teammates for practice!