“Yes — I can imagine it. A simple sphere, maybe sitting in a soft pool of light.” “I’m picturing it as a bright red ball, glossy and catching a bit of light on

“Yes — I can imagine it. A simple sphere, maybe sitting in a soft pool of light.” “I’m picturing it as a bright red ball, glossy and catching a bit of light on one side.” Great, huh? Except that’s what ChatGPT said when I asked it those two questions. It certainly isn’t picturing anything. If a robot which only ‘thinks’ in terms of chain-of-thought of abstract tokens can act as if it truly sees things, what makes you think this test has any validity at all?