I'm currently in a process of getting access to a multi-GPU cloud instance for zk-verified ML experiments (well, I'm just waiting for the Google guys to process a request and thinking about another blog post and a little infrastructure-related project but mostly trying to clear my head to see strategy). Any interesting results will inevitably be published in the company blog but I just wanted to express how I love this opportunity to play with shiny toys helping meaningful cutting-edge research.
I only recently joined Inference Labs, had to catch up on a lot of things and don't have many cool results to show yet but I already feel right at home there with all the support and resources being available to me whenever I need something while not restricting my creative freedom in any way. Well, it helps that I have a long history with some of the guys and also that absolutely everyone is just brilliant, if I had to trust the future of the world to just one team I'd think no further (luckily we're a part of a much bigger community and with that much support survival of the human race is almost certain). I guess I'm in the honeymoon stage with this job and let's hope it will develop into a healthy long-term thing, just like before everything.
Anyway, practical limits of Zero-Knowledge Proof systems for Machine Learning are about to be thoroughly tested and massaged. One experiment at a time. Gotta sleep and hopefully my request will be processed tomorrow. I wonder if some folks at some data center are physically moving some GPUs while I wait, could be. Please hurry guys, I have a thing to test.