Extropic Founder: Running Probabilistic AI on Deterministic Hardware is a 'Demon Tax'
beffjezos · x · 2026-08-05
Extropic founder Guillaume Verdon argues that running probabilistic AI algorithms on deterministic hardware is a massive local optimum in computing. Citing Landauer's principle, he explains that classical computers pay an energy penalty for a "Maxwell's demon" every time a bit is reset to a deterministic state during a clock cycle.
He notes that core modern AI algorithms—like Transformers, diffusion models, and recommenders—are inherently probabilistic. If hardware could run probabilistically using fractional bits, it would avoid the energy loss of deterministic resets and drastically improve power efficiency.
More from Infra
- Musk: Memory is the AI Bottleneck; SpaceX to Triple Nvidia Compute by 2027 — firstadopter · 2026-08-05
- Update CUDA to 13.3 to Fix DeepSeek V4 Flash Looping Issue — Easy_Werewolf7903 · 2026-08-05
- Hands-On: Deploying 550B Nemotron 3 Ultra Locally on NVIDIA DGX Station — NVIDIA Developer · 2026-08-05
- Together AI's Monthly Token Volume Skyrockets from 30B to 400T — togethercompute · 2026-08-05
- API Key Expiry Leads to Runaway Agent, Costs $300 in Idle Compute — voooooogel · 2026-08-05
- Dev Builds Pixel-Art GPU Cluster Dashboard in 20 Mins Using GLM Agent — Porespellar · 2026-08-05