But progress is slowing way down, and I don't forsee it making it in 50. If things continue like the first 40 years of computing, we would get there. So are we going to tell it in english? The current Stable Diffusion is hardly able to communicate in english. The biggest issue is how do you tell it what to code? We have a method to tell computers what to do. We can't be sure this was designed at all, or just copied. That is a specific game that is easy to teach the AI with many examples. The necassary processing power has to be much further then it is.īuilding a game like Minecraft isn't a test of this. But even then, I'm not sure we'd be there. But processors that run on photons could help us, or something like that. Without something big like that, we won't ever see it. but we will see if that ever comes to fruition.
And I don't foresee quantum computing doing what we want, not at least in its current attenuation. Their is a possibility they move away from Silicon, but that's unlikely anytime soon. They have been moving towards more 3-D processor designs lately, but that too will only yield so many results. And what we need is much, much more processing.īut again, they all hit limits with silicon that we won't be getting over any time soon. Moores law is about the number of transistors, not the processing doubling. Year to a year and a half with 10% gains? That isn't enough, and it doesn't look like 10% will keep continuing. Currently, we see at most 10% increase from generation to generation of GPU's and that is slowing. They are doing that because shrinking silicoln isn't possible anymore. NVIDIA already optimizes for AI, and they are not the only ASIC processor being designed for AI processing.
GPU's if you want to call them that, will not be GPU's for much longer. Sorry, but I'm talking about limitations of Silicon and all Silicon processing.
0 kommentar(er)
