It is the closed model companies that manufactured AI apocalypse hysteria....to cripple open models and truth seeking competitors. The hysteria is leading to compute constraints which favor open model companies and SPCX, NVDA and PLTR. There are always karmic consequences. The…
SpaceX is currently the only company that can solve that massive compute demand that is about to come up. AI companies are limited by compute and compute is limited by energy. Currently two companies: SpaceX and Nvidia sit exactly right where the greatest bottleneck to…
@saranormous Good thing we’ve got zero LLMs in our critical path. Deterministic AI is not dependent on compute and runs without GPUs! One of the biggest benefits of it, besides the ability to fully resolve up to 1 million tokens in 200ms or less.
@saranormous We want to help! Heavy Metal Network is building the inference plane that maximizes the utility of your existing compute. DM for more.
@saranormous AI that always need more compute is mediocre. Like an engine always burning more gas. Stupid.
@saranormous have B300s for anyone interested
@mhdempsey @timzaman it is now true for inference if you grow to any scale, and inference is about to get hard for companies because ppl don’t realize it’s constrained and not a liquid market. much much easier to get capacity in 3 months than in 3 weeks, but end users don’t care
@timzaman I don’t mean teams can’t afford compute. I mean they have enormous money to spend and there isn’t available compute
many great ai companies about to be desperate because their growth is limited by compute
@mhdempsey @timzaman better make friends with your inference vendors
@timzaman Hasn’t been true for CO’s outside of the big labs until recently :)