Ben Lorica: AI's data problem moved downstream — the hard part is making information usable
bigdata · x · 2026-09-30
In Gradient Flow, Ben Lorica argues AI's data problem has shifted from raw material scarcity to usability: information can exist without being usable by models.
Robotics data is no longer scarce, but conversion is costly:
- One company trained on over 1 million hours of human video
- Another turned 1,900 hours of first-person footage into 18,000+ hours of robot-format training data
- Every such system needs substantial machinery to make human experience usable by robots
The web is becoming a market with rules:
- An academic publisher opened selected books for AI retrieval while explicitly withholding training rights
- "Use for AI" is splintering into separate permissions for training, retrieval, and inference
- Site owners can increasingly distinguish search, training, and agent crawlers; one closed-beta system supports HTTP 402 pay-per-access responses
Related event: AI's Data Problem Shifts from Scarcity to Usability, Says Ben Lorica(3 posts)→
More from Infra
- On-device inference is finally practical: NPUs shift AI workloads off the cloud — lmoroney · 2026-09-30
- Cognition runs production workloads on Vera Rubin NVL72 with 4.8x token throughput — nordicinst · 2026-09-30
- Dual R9700 local inference: how much does PCIe Gen4 actually cost vs Gen5? — IngwiePhoenix · 2026-09-30
- SwitchSD, speculative decoding that switches between neural drafting and context copying, accepted to NeurIPS — arankomatsuzaki · 2026-09-30
- Rebellions' 2048 TFLOPS NPU lands major Japanese AI datacenter deal with ai& — DavidBennett__ · 2026-09-30
- Solo Rust + Vulkan training backend now passes 14 architectures at 2e-7 parity, with full LoRA/PEFT lifecycle — PhysicsDisastrous462 · 2026-09-30