Right now it takes at least a few seconds from the time you give a text prompt and the time you get an image back from one of the currently available generative A.I. programs. This is a limitation of the current technology.
But Moore’s Law keeps marching onward, in one way or another. With increased computational parallelism and newer semiconductor technologies, response time will gradually trend downward. In another decade or so the response will reliably arrive in a fraction of a second.
From the user’s perspective, this means that images and videos and simulated 3D scenes will appear even while you are describing them. But even more important, it will mean that you can edit those images and videos and simulated 3D scenes simply by continuing to talk. As you speak, the changes will happen right before your eyes.
When we get to that point, generative A.I. will truly become a fundamental new mode of human expression.
I agree that it will work great for the visuals, but interating with Grok’s almost instant LLM was really weird and belittling – as if I was playing chess with someone with ELO of 5000.
Having a conversation where the other side finished the response in milliseconds felt like conversing with an all-powerful entity to which I am just a small and slow meat-bug. “Whatever you can think of, I’ll have a response before you’ll even start speaking the question.”
I find both groq and sdxl to be fun because they are instant. I didn’t feel belittled by either; I’m interested in why it would feel like that? In both cases, fast seems fun! Neither is my workhorse AI of choice right now, but experimenting with the speed does indeed feel expressive.