Too much fun to be had with OpenAI’s image generator.
I have been an incredibly sharp critic of Dall-E. Yes, it was fun for silly images, but at the end of the day, nothing it produced was super usable. It couldn’t spell well, it frequently ignored prompts, and generally, it was a frustrating experience for this creative.
Recently, Open AI essentially laid Dall-E to rest and made vast improvements to its image generator. When the news broke, I was quite skeptical. And then… pleasantly surprised.
The model obeyed prompts better than previous attempts, and its spelling capabilities have improved. The quality of the images was vasty improved — they have much less of that “AI” feel to them. I also found the responses to be more conversational.
Of course, I have concerns over the strain this improvement has on the GPU usage for Open AI (hallucinations in my normal prompts have been on the rise lately), however, I also understand that the AI I use today will be the dumbest AI I ever use.
An example: To know me is to know I have a true affinity for Funko Pops. No lie, we have (I think) around 400 in our house. Pops are my family’s love language. So seeing what AI thinks I look like as a pop is TOTALLY fun for me! I also fell victim to the viral sensation of creating myself as an action figure. Well played, Open AI.
LP Innovates 