TLDR Tech / Design
Applying Doherty's Threshold to Generative AI for Improved Perceived Speed
Generative AI necessitates a re-evaluation of Doherty's 400-millisecond response threshold, as requests now involve user acknowledgment and answer generation. While models may require several seconds for completion, interfaces can provide immediate feedback, display progress, and maintain user attention during generation. Prioritizing the design of this initial response can significantly enhance the perceived speed of slow AI without altering the model or inference speed.
Read the full story