Personalised ads are beside the point. The issue is how they are personalised, namely by building a rich profile of user behaviour based on non-consensual tracking.
It isnt even clear that there's a meaningful sense of 'consent' to what modern ad companies (ie., google, facebook, amazon, increasingly microsoft, etc.) do. There is both an individual harm, but a massive collective arm, to the infrastructure of behavioural tracking that has been built by these companies.
This infrastructure should be, largely, illegal. The technology to end any form of privacy is presently deployed only for ads, but should not be deployed anywhere at all.
If I wanted to, right now, I could build a deep profile for every single user of HN, simply by downloading the public pages, and cross-referencing comments, upvoted/favorited stories, etc with usernames. I could then create a weighted index that tells me how likely a user is to be a libertarian, gay, wealthy, etc. Then I could e-mail those users and offer to sell them privacy-focused freedom-loving lgbtq+ products.
I can pretty much do whatever I want with this database, partly because you don't even know I have it, but also because it's all public information you've posted to the web voluntarily. Maybe the ToS will say I can't, but they have to catch me/stop me. I could just hire some Russians to do it for me and collect the data later.
I'm not saying this should be allowed, but it's probably going to be impossible to stop, and the implications (esp. for political concerns) are enough of a motivator that just making it illegal probably won't end the practice. We have to consider alternatives so that we aren't stuck in some information arms race that makes the problem worse.
For example, we could say that private data should remain private, and public should remain public. Data which everyone has a reasonable expectation to be private - like the private photos you upload to Google Drive - should never become public, and thus should never be aggregated into some product (trained for an AI, etc), used to sell you something, etc. But data which does have a reasonable expectation to become public - like comments on a public forum, likes on public posts on Facebook - should remain public, and thus be used the way any other public thing can be. We already have legal limitations on uses of some public things, but we can expand that if need be.
Then we can legally define what constitutes private and public, and construct tech so that it's very clear to people what's public and what's private, and then they can decide what they will post where, or what sites they will/won't use in what ways, etc. It's already clear what's private and public out in the real world. We just need to make that same distinction clearer for other cases, like when and how companies collect data and what they can use it for. It's going to require case-by-case analysis, but we can totally get there without having to ban everything or allow everything.