@redtachyon Not as daily drivers tbh since 5.5 gets mostly everything I need done. I mainly use them to generate synthetic data for fine tuning very small models, but that’s the most feasible “local” driver model imo
@max_spero_@adastroworld@rob_mcrobberson I am actually in the process of doing this at work, looking to migrate traffic from Modal into both AWS and Nebius. I have a very long list of providers and H100 prices if you want some insight, including their ability to handle “serverless”
@syshukus@HSVSphere@georgecurtiss A few months ago I learned about the subsidized market in the US that leads to corn syrup in EVERYTHING. It's a large part of why so many Americans are obese. You can buy same product in other countries with a fraction of the calories
@HSVSphere@georgecurtiss I just searched up UHT and that’s so cool, you can buy milk and keep it there for months until you need it. I need to visit Europe
@OfficialLoganK@harjtaggar I really respect and admire how receptive you are to feedback even when it's negative or mean sounding. Wishing you guys luck on the next release :D
@jparkjmc In my masters second semester at gt. At what point did you feel ready/comfortable to interview at deepmind? I did interdisciplinary ml research in undergrad, took grad ml, and work at a startup fullstack deploying models in serverless env. But it feels far off from good enough