There have been a tonne of discussions on it, and people have fallen into different camps ..
But I think it’s three issues
I think it’s usability , and accessibility..
The rather lovely Matt bliss wrote a thing , and has done some small talks on image refinements I can’t find the article but hav e his talk
https://www.youtube.com/watch?v=gEA8lGEJXBs▾
And I am sure there is a article somewhere by a Mike Masey or others on context alt text