A lot of the "humans will still be needed" arguments implicitly rely on the assumption that the trend of AI progress stops here.
The original DALL-E was released just a year ago. In one year, AI image generation went from generating images that looked slightly normal if you squinted really hard, to the beautiful, detailed and creative results we see today.
The same speed of progress can be observed in almost all other domains, from code generation to question answering and music synthesis.
For many of these problems, it almost feels like the next step up in quality is just two hours of planning and a couple months of training away. Like they're already solved and just waiting for someone to implement the solution.
So yeah, humans are still needed right now. But I wouldn't base my hopes for the future on that, because the world of AI capabilities will probably look very different in just one year.