I see GP is talking more about Netflix and the like, but user-generated video is horrendously expensive too. I'm pretty sure that, at least before the gen AI boom, ffmpeg was by far the biggest consumer of Google's total computational capacity, like 10-20%.
The ecology argument just seems self-defeating for tech nerds. We aren't exactly planting trees out here.
Excuse the throwaway. It's not even just the employees, but it doesn't even seem like the technical leadership seriously cares about internal AI use. Before I left all they pushed was code generation, but my work was 80% understanding 5-20 year old code and 20% actual development. If they put any noticeable effort into an LLM that could answer "show me all users of Proto.field that would be affected by X", my life would've been changed for the better, but I don't think the technical leadership understands this, or they don't want to spare the TPUs.
When I started at my post-Google job, I felt so vindicated when my new TL recommended that I use an LLM to catch up if no one was available to answer my questions.
The ecology argument just seems self-defeating for tech nerds. We aren't exactly planting trees out here.