a.OpenAI relied on validation indicators, including performance metrics that accelerated growth; interaction patterns such as the number, frequency, and length of sessions; and user feedback and correction ratings.It acts as a testbed that collects and reprocesses data to develop improved models, with low-cost scaling (GPT-3.5 vs. GPT-4).