Unverified50% confidenceOpinionExact time
Altman used a metaphor to point out that computers expend large amounts of compute rendering interfaces for human eyes to view, and then AI works hard to "read" the interface back, rather than directly obtaining the underlying data.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/21/2026
First Seen
Valid until: 1/19/2027
Sources
Related Entities
Related Claims
UnverifiedAltman believes AI needs new hardware forms, because laptops and iPhones are designed for people rather than for AI, and having AI render a UI for people to watch and then read it back is essentially wasteful.77% similarUnverifiedWhen AI "understands" interfaces through screenshots plus OCR or vision models, it is effectively downgrading already structured data back into pixels and then having the model reverse-infer the semantics, causing a dual waste of compute and latency.75% similarUnverifiedLanier believes that today's AI systems are essentially products that aggregate, recombine, and re-present vast amounts of human-created text, images, and code, and that the systems themselves do not "understand."71% similarUnverifiedThe main interaction mode of current AI coding assistants is still text input, and visual information on the screen is hard to convey efficiently in words.71% similarUnverifiedThe speaker believes that humans are essentially an asynchronous API for AI agents.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1083562API
curl https://kongchang.com/api/v1/knowledge/claims/1083562MCP
get_claim(id=1083562)