newsfilter
← Interviews
Sequoia Capital1 min

The Problem With Testing AI Architectures at Small Scale | Jerry Tworek, Core Automation

Jerry Tworek

  • Reinforcement learning architectures generally require a baseline level of compute capacity to produce effective results and function meaningfully.
  • Validation attempts on small datasets prior to scaling may fail to accurately reveal an architecture's true capabilities if they do not meet these minimum compute thresholds.