Raw LLM Responses

Coming from a tech background and having been recently enamored with AI art (Partially because it is an amazing feat of machine learning with amazing visual outputs and partially because I can't draw at all), I have a few technical bits I would like to point out. Nightcafe and Wombo are both last-generation models (maybe even two generations ago) Current models like MidJourney (Free 25 images + 30/month) and Dalle2 (free 15 + 115/15$) blow Nightcafe and Wombo out of the water. Next-generation models like Google's Imagen and Parti are said to be even better. Both Dalle2 and MidJourney are able to make some pieces of actual zero-shot (no editing) art occasionally, though we usually run it through basic filters and upscalers to fix some artifacting. Dalle2, Imagen and Parti all specialize on photorealism and a sort of stock image vibe, while MidJourney focuses on "Art" over realism. Although these models aren't yet able to replace artists in quality of output, they are getting way closer the Nightcafe and Wombo. All current generation models are also only a few months old. Dalle1 was release a year ago and was 10x worse then dalle2, although still quite impressive. Additionally, older models use some form of GAN-based generator while newer ones use Diffusion-based generators (start from white noise and work towards more detail) Midjourney can even send you a video of your generation being made in real-time. Datasets used by AI models also aren't in the thousands or even millions in count, but often in the billions. Midjourney publicly has said they use the open-source LAION 400M dataset and are moving towards using the LAION 5B dataset soon (filtered down to roughly 4bil). While I don't think AI generated art is ever going to replace human art, there are many niches that it could fill and personally, I welcome my new artistic overlords. Please reply with some insight from the artistic world!