Sam Altman: AI Should Stop Reading Human Screens
The OpenAI CEO believes AI is trapped in interfaces designed for humans, and that new hardware is the answer. But others are skeptical: an API could solve it—why build a new device?
Sam Altman recently said something worth pondering.
He said: “AI is constrained by something designed for humans.” That “something” is the computer screen.
AI has to read the UI, read buttons, read pixels that were originally rendered for human eyes. Altman thinks this is pretty silly: “The computer spends all its energy rendering an interface for a person to look at, and then AI has to read it, instead of just taking the data directly.”
He went further: AI needs a new form of hardware. “I believe this technology wants a new carrier.”
This isn't a sci-fi fantasy. Altman's reasoning is very practical: most of his life doesn't happen in front of a computer. If AI really wants to reduce our cognitive burden and take initiative, it has to be embedded where we are, not just stay inside a screen.
Interestingly, someone in the comments replied: “Funny, the solution to reading pixels is usually just a good API—no new device needed.”
That makes sense. When AI reads screens, it is essentially working around bad design. If all systems exposed open interfaces, AI could consume structured data directly, which would be far more efficient. But the reality is that the world won't be rewritten for AI. A huge amount of information still exists only in interfaces designed for humans.
The new hardware Altman talks about may not mean another screen. Maybe it is a carrier that lets AI directly perceive its environment and access data directly. The question is whether such a device can be accepted by the market. Will it become another pair of Ray-Ban glasses?
Right now, no one knows. But at least Altman has put the question on the table: AI's evolution may be stuck on hardware form factor. And that bottleneck is harder to solve than model parameters.
发布时间: 2026-10-27 00:03