When you are in an echo chamber created by social media, the dehumanization required to make you a killer is mich more subtle, more easy and more covert.

All the methods you have mentioned, are, to some extent in the public eye. Imagine the uproar if anti-semitic articles occured in any of the daily news papers. It is a matter of public record, even if some of the methods come out later. Now imagine how many such articles are published and propagandized in online media, away from the gaze of critics, away from serious action by concerned authorities. The power is in the hands of big data companies and they are driven by greed and stock market performance.

Once the papers die, people will start dying too.

There having been equivalent evils before the internet does nothing to disprove the thesis of it causing harm. Even if you could show that the internet, is net positive (pun intended), I don't see how that would be an argument against possible further improvements.

Analogy: People being killed in accidents with horses is not an argument for disregarding traffic deaths. Nor can that fact be used to argue against seat belts or airbags. Only if someone were proposing to go back to horse & carriage would it make sense to reference total effects.

If "the internet" wants to avoid the fate of more regulation, it is most sensible to take its critics' worries serious, and to try to find ways to be a positive force in the world without inviting the sort heavy-handed regulations that will otherwise follow.

As but one example, see spam, or child pornography: in those areas, tech companies, users, and governments found ways to effectively work together mostly without resorting to draconian measures.

For an example of such a process failing, look at the EU's recent "link tax": Because the tech community has been so adamantly opposed to taking journalism's existential fears into account, this rather unwieldy regulation will become law. This could have been avoided if, for example, ad blockers had settled for blocking only dangerous and/or the most annoying ads, an approach that could have easily reduced annoyances by 80%+ while costing publishers far less. Or micropayments that actually work, and users to use them. Etc.

The ads that aren’t “annoying” still violate your privacy. I don’t want to load any of them.