That’s all ok, but you can only speak for your own preference in this regard. Others should be afforded the opportunity to opt-out in a way that’s not just “don’t ever use the Internet”, scraping the internet is an action that assumes by default that everyone is ok with their personal work being used in this way.
Me posting this response here with the expectation that it be read by others as part of this discussion cannot reasonably be said to also be explicit consent to have my writing scraped and used to produce things that emulate how I write or to use my style as a means of unique identification (something which already occurs across many social media sites) or any other unforeseeable use of the data in the future.
It may be legal and may happen whether I want it to or not, but if asked explicitly if I would like my writing used in this way I personally would definitely say no, so it’s hard for me to see the “consent by default” argument as truly ethicsl. That’s the issue people have with the way these developers are training their models. It’s not about the LLMs themselves or the current quality of their output, it’s about people basically being unwilling participants in an experiment and having their data used to profit others.