3 ms·
Why is everyone with criticisms about DALLE-E, Midjourney and Stable Diffusion so focused on their day 1 capabilities? The fact that it can do any of what it do
by RationPhantoms 4y ago
Why is everyone with criticisms about DALLE-E, Midjourney and Stable Diffusion so focused on their day 1 capabilities? The fact that it can do any of what it does today, is absolutely insane. Is no one focusing on what day 36 or year 2 will look like?
I've been privy to some of the model iterations Midjourney has gone through on their Discord and in a series of a few weeks, they've made substantial gains on generating faces from their V1, V2 and V3 models.
Within a short amount of time, this will no longer be a complaint.
In relation to OP's post, I guess Reddit can't guarantee this won't also turn into DeepFake territory so they're banning it outright.
- IceMetalPunk 4y agoOh, don't get me wrong, I agree! Since the invention of transformers, progress in generative AI models has accelerated exponentially. It's an exciting time to live, and they are definitely extraordinary! (I mention transformers because basically every bleeding-edge AI in the last few years has integrated a transformer somewhere in its architecture. They really spurred an AI revolution.) That said... it would be lying to claim Stable Diffusion's current generations aren't nightmare fuel :D
- throwaway5959 4y agoMaybe lead with that then.
- naniwaduni 4y agoWhy would someone want to lead with irrelevant sermons to the choir before making the remark they came to make?
- tracerbulletx 4y agoThis, plus the day one capabilities are already useful and are actively being used. For each weird bizarre output there's also one you can barely tell wasn't made by a relatively talented human artist.