> And still, it doesn’t know we use chairs for sitting down. I imagine if I gave it the prompt "chair being used by woman" then a majority of the pics would ha

> And still, it doesn’t know we use chairs for sitting down. I imagine if I gave it the prompt "chair being used by woman" then a majority of the pics would have someone sitting in it. I realize your point is more of a "Chinese Room" philosophical objection. But it's important to note the level of sophistication these models have. I'll also add that Nvidia is currently doing research on 3D model based generation, so that the model is aware of the 'skeleton' if the subject before illustrating it.