It should be called an "automatic annoyance generator".

The comments generated are not just useless, they are harmful. They made it harder to recognize and act on actually meaningful comments.

That was my intial reaction as well but then I thought, if the AI can't make sense of it then a human probably wouldn't either. So, I guess a tool like this can serve to check if code is actually readable. I think variables names play a big role in this (i.e. communicate intent).

Agreed! A useful comment would only be present to describe an edge case, a related business rule, a future change that might affect this code, virtually anything another engineer might need to be aware of, that is not the code they are looking at.

It might be very interesting to write the "no value" comments automatically. It can help us focus on and emphasize the part that we think is missing from the autogenerated ones.