ML can also be used to discover new things, like here: (Mathematical discoveries from program search with large language models) https://www.nature.com/articles/s41586-023-06924-6 or combine old concepts to create new ones.
I'm not going to bother arguing with the rest of your article, but I think everyone agrees that it is better to rely on facts rather than myths.
do you like doing art or digital?
The linked article is half a year old and is more of an opinion piece than a reflection of reality.
Synthethic datasets often perform better than the raw stuff, as exampled here https://arxiv.org/pdf/2401.00368.pdf as well as here https://arxiv.org/pdf/2401.01335.pdf , as well as community driven projects in image generation.
ML can also be used to discover new things, like here: (Mathematical discoveries from program search with large language models) https://www.nature.com/articles/s41586-023-06924-6 or combine old concepts to create new ones.
I'm not going to bother arguing with the rest of your article, but I think everyone agrees that it is better to rely on facts rather than myths.