It's easy to nitpick all the details in the video, but keep in mind that 2 years ago generative AI videos consisted mostly of shape shifting mosaics that vaguely resembled the things they were supposed to be. And now we're down to "in this frame the 10x10 pixel airplane has a third wing".
That doesn't excuse the use of copyrighted material to get to this point, mind you. But to claim that this tech is going nowhere is just a contextless circlejerk.
Im amazed at the quality leaps for output for new versions of chatgpt.
I dont rely on it for project structure and the vast majority of logic but its so fucking cool that i can now give specific instructions and receive a mostly working script, give it the resulting errors, and after a couple prompts i have working code.
Not very surprising that people sticking their head in the sand have no idea what they're talking about. Our methods of interaction and completeness of input are just as important as the capability of the model.