I specified a featureless background - no floor or shadows. The light source is inconsistent. There are only a few frames where it is a fully formed chair. Only one axis of rotation... This isn't a 'scale-up' problem. This is an 'our methods are fundamentally flawed' problem.

Why not the Ministry of Peace?

Or this one... www.penny-arcade.com/comic/2001/0...

Damn you beat me by 700 minutes 😭

Oh yes... They'll "burn it down" by using it to death. 🙄 So exactly what do you permit AI to be used for? Saving lives is ok, but evoking emotion is off limits? www.axios.com/2024/11/25/g...

It's not even to feed Brazilians 🤦‍♂️ www.statista.com/chart/31839/...

estudado o verbo 'to be' dez mil vezes 😝

I get where you're coming from. AI is coming for all of our jobs (and in some cases it's already here) and that's distressing. But instead of pretending it's going away, we should be fighting for the kind of world that's going to be built with it. www.nature.com/articles/s41...

I would encourage you to learn a bit more about [embeddings](youtu.be/gQddtTdmG_8) and [transformers](youtu.be/wjZofJX0v4M). Or even just put two unrelated concepts that are unlikely to have been paired in the training data. It is quite capable of making imaginative leaps.

This is verbatim from the ChatGPT model behavior guidelines: "Our goal isn’t to train
models that take the correct viewpoint". I'm well aware that this pandering to both sides can be circumvented with prompting. But should it need to be? openai.com/index/how-sh...