Hi, my name is Ben Thompson.

I’m currently working on AI adversarial attacks and AI interpretability at Confirm Labs!

I’m excited to talk to you about all sorts of stuff. And let me know if you want to work together.

A big list of topics that excite me and I write about here: artificial intelligence, machine learning, software engineering, computational math, high performance computing, integral equations, earthquake science, earthquake prevention, adaptive experimentation, the structure of science, science policy, and many more.

Other things that excite me that I don’t write about here: growing plants, running, mountains, my dog, the time I jacked up my house by a quarter inch.

Resume  /  Google Scholar  /  GitHub  /  LinkedIn  /  Twitter  /  Email
Posts BIEBook
Recent postsall postsrss
Fluent dreaming for language models (AI interpretability method)
Extreme website and app blocking
Some thoughts on my ideal publication system.
Unexpectedly sucky things I've noticed about grad school
FAQ: Why should research scientists be at the center of our scientific workforce?
We need technical research consultants.
Open source release! Generalized linear modeling with glum and tabmat.
50 suggestions on correctness and testing for scientific software for PDEs
Gaussian quadrature is not optimal
Reflective deep work
Why it's okay to share your code
Maintaining momentum
Maybe we should stop using planar triangles for fault modeling
Cool video of Cascadia earthquake cycle simulations
130 million triangular dislocations per second with Python/CUDA
CLUDA - Write once, run anywhere with both CUDA and OpenCL.
Regularizing rate and state friction for numerical simulation
A quasidynamic spring block slider
Cloudpickle, serializing functions and monkey patching
My Python testing set up
Automated testing for scientists and data analysts
Sneaky (transparent) huge pages
Sparse n-body matrices
"Serializing" a function in C++
Setting up my website