Vector generation does poor job of understanding prompts
Vector generation is horrible at parsing prompts, frequently ignoring entire concepts contained in a multi-themed prompt. For example, "cute robot detective" completely ignores the "detective" portion of the prompt, instead churning out endless variations of standard looking robots. It appears that creativity based on combining concepts is not within the AI's abilities, but is exactly what artists regularly require.
Example images, as requested. Changing word order did not noticeably affect results, either.
Jay, can you please share some examples though?
Not that it works better for others, but just to have images in a report about images :)