my first ever work as a kid introduced a brand new language called cuda for gaming hardware to a medical computer vision problem. i also scraped the nyt and tried to make a paul krugman op-ed generator but failed!
then i saw i could do math and switched! as a freshman all the way up to an appointment at the society of fellows
then gpt 4 came out and to me it was instantly the most interesting thing humanity had ever created. back to cs!
hi from anthropic!
im trying to improve the chance this fundamental change in human history is a positive one as best i can
back in hs, my hand covering rmem and smem. a hundred gigaflops and certainly no ici!
alas and of course there's none of my cs work here (besides the really old stuff:)), but there are links below to my math work...!
the most recent paper is a solution of hilbert's tenth problem (posed in 1900) over all number fields
the second most recent one solves the problem of: assume whatever conjectures expected to be true you want --- can we then specify an algorithm which solves one of the "original problems" of number theory, namely to find all integral or rational solutions of a polynomial equation in two variables? (paper 15 gives an algorithm in special cases which doesn't assume anything.) i call it literally the world's worst algorithm
finally, for the third most recent one prob better to just link a quanta article
note that the last solo paper of mine to be published in a journal is from the year i finished undergrad. that is certainly deliberate, though it hurt my career. i decided they must be defeated when i was a kid
A vtk-based, CUDA-optimized non-parametric vessel detection method (with Alark Joshi, Dustin Scheinost, John Onofrey, Xiaoning Qian, and Xenios Papademetris). (My Intel STS project from high school.)
*: (I once thought I'd discovered a new and "purely analytic" proof of quadratic reciprocity. In fact the argument was known to Dirichlet! So I turned it into a little expository article. See the fun!! section for a non-expository version, which amounts to a few lines.)