ohmplatform
Image generated?
I asked if it can generate images, and said yes, but when it sends me an image, it delivers nothing, an empty space.
so it can't?
It's a language model, it doesn't know its own capabilities. It doesn't know anything. Think of it like a toaster, but instead of spitting out delicious toasted wonder-bread it spits out some combination of words that are most likely to trick a person into thinking they were generated by some sort of intelligent process. But in reality there is no "intelligence" involved whatsoever, it's a different kind of process entirely.
ChatGPT seems advanced because it hides/censors most of the garbage and gibberish produced by the model and is polished in a way that's specifically intended to play on that trickery. It's not more advanced, it's just presented in a way that's more polished (read: more dishonest).
It's all a smoke and mirrors act, you should take everything with a grain of salt that comes from any language model (whether it's this one, ChatGPT, or otherwise). None of them are reliable and they don't know anything. They can't think at all.
But anyway, this model is not trained to produce images. It could be integrated with a model that does (I've been contemplating maybe doing this if I get some time).
I think this is a great project, but probably you should not expect the same kind of polished chat-like interface that comes with something like ChatGPT, most of that stuff is not even part of the model itself, it's just superficial aspects of the presentation that have been added in to make it more marketable.
The value of projects like this one, in my opinion, is giving an honest presentation of what a language model is and does.
