Sara Hooker says AI labs lack product surfaces and pay data firms to simulate them

AISara Hooker, who is associated with Cohere, says institutional companies with large interaction surfaces hold the advantage right now. She says there is a narrow window in which they can use their own knowledge to gain an edge.

AISara Hooker, who is affiliated with Cohere, says AI labs are paying data companies large sums to synthesize and mimic real-world surfaces. She argues that a simulated surface is never as good as the real thing, so synthetic data is a substitute rather than a replacement.

AISara Hooker, who is affiliated with Cohere, says the shift gives power to companies that hold the product surface rather than pure AI labs. She adds that frontier lab modelling teams typically do not think much about interface, leaving them at a fundamental disadvantage without a direct connection to users.

AISara Hooker argues that frontier AI labs lack access to product surface, which she calls the most important input for the next era of intelligence. She says this explains why data companies are earning large revenues. The post is a short reply to a question about data-company profits.