Why Does Parag Agrawal Treat Human Click Data as a Bug?
Parag Agrawal says Parallel treats human click data as a bug. An agent that searches in order to get work done should learn from agent feedback, not from what humans click.
Human ratings and click logs used to be the scarce signal for whether one result was better than another, and they were hard to collect at scale. Expert ratings can now be created much more cheaply, and the models that compress information can be applied to indexing and ranking. Agents matter as a customer, not only as a technology. That is the argument he made when starting the company: agents, as that customer, made a new web index tractable.
Parallel’s Parag Agrawal: Building a New Web for AI Agents ·