4 ms·

I agree that it's questionable. I commented the same in the thread I linked: https://news.ycombinator.com/item?id=25513161 https://news.ycombinator.com/item?id=

I agree that it's questionable. I commented the same in the thread I linked: https://news.ycombinator.com/item?id=25513161 https://news.ycombinator.com/item?id=25513161 However, at least for Pi-Hole users, more is usually better, so I added the list to my Pi-Hole.

phire 6y ago
> > We were testing an AI that could show some basic emotions about internet content, and turns out it was very precise at getting “annoyed” by ads and “unsolicited” third party connections… Holy shit that's such bullshit. They are basically claiming they invented a artificial general intelligence, with feelings, that happens to feel the same way about ads as us. It's basically sentient instead of publishing research papers, they turned it into an ad blocker.
- srtjstjsj 6y ago
  It's just colorful language for the fact that ads and spyware score high on their model for bad websites.
  - phire 6y ago
    First: Marketing bullshit is still bullshit. Even if it's not morally wrong, it makes you look like an idiot who doesn't understand the technology you are selling. In the worst case it might even be used as evidence that your work is a fraud. There is no benefit; To the lay person, It would sound just as impressive to say "We trained a machine learning model to detect ads and spyware" and that wouldn't immediately set off alarm bells with people familiar with the current state of machine learning. Second: Talking about fraud, the evidence linked above is pretty strong. Their alleged AI is somehow detecting test domains that authors of other lists as "ads or spyware". Test domains that aren't linked anywhere on the internet. In one "smoking gun" example, the test domain doesn't even have a DNS entry. The alleged AI can't even load the domain to scan it.
llacb47 6y ago
No, more is not usually better. Especially with a garbage ""AI-generated"" (not) list with untrustworthy maintainers like this one. It's better to add a low number of lists with trusted maintainers, who actively curate their lists and respond to false positives. That means no "mega-list" abominations like oisd.nl. I suggest: https://www.github.developerdan.com/hosts/ https://www.github.developerdan.com/hosts/ https://gitlab.com/curben/urlhaus-filter/raw/master/urlhaus-filter-hosts.txt https://gitlab.com/curben/urlhaus-filter/raw/master/urlhaus-... https://raw.githubusercontent.com/notracking/hosts-blocklists/master/hostnames.txt https://raw.githubusercontent.com/notracking/hosts-blocklist... https://raw.githubusercontent.com/anudeepND/blacklist/master/adservers.txt https://raw.githubusercontent.com/anudeepND/blacklist/master...
- Merman_Mike 6y ago
  Can you explain why more is not usually better? I added the 4 lists you recommended to my Pi-Hole, which added a net new 73,253 domains to my Pi-Hole. My total is now close to 2M.
  - PixyMisa 6y ago
    You could just blacklist *.com and be done with it.
    - freebuju 6y ago
      You joke but I would be most happy if all my web needs could be served on .onion addresses

phire 6y ago

> > We were testing an AI that could show some basic emotions about internet content, and turns out it was very precise at getting “annoyed” by ads and “unsolicited” third party connections… Holy shit that's such bullshit. They are basically claiming they invented a artificial general intelligence, with feelings, that happens to feel the same way about ads as us. It's basically sentient instead of publishing research papers, they turned it into an ad blocker.

srtjstjsj 6y ago
It's just colorful language for the fact that ads and spyware score high on their model for bad websites.
- phire 6y ago
  First: Marketing bullshit is still bullshit. Even if it's not morally wrong, it makes you look like an idiot who doesn't understand the technology you are selling. In the worst case it might even be used as evidence that your work is a fraud. There is no benefit; To the lay person, It would sound just as impressive to say "We trained a machine learning model to detect ads and spyware" and that wouldn't immediately set off alarm bells with people familiar with the current state of machine learning. Second: Talking about fraud, the evidence linked above is pretty strong. Their alleged AI is somehow detecting test domains that authors of other lists as "ads or spyware". Test domains that aren't linked anywhere on the internet. In one "smoking gun" example, the test domain doesn't even have a DNS entry. The alleged AI can't even load the domain to scan it.

llacb47 6y ago

No, more is not usually better. Especially with a garbage ""AI-generated"" (not) list with untrustworthy maintainers like this one. It's better to add a low number of lists with trusted maintainers, who actively curate their lists and respond to false positives. That means no "mega-list" abominations like oisd.nl. I suggest: https://www.github.developerdan.com/hosts/ https://www.github.developerdan.com/hosts/ https://gitlab.com/curben/urlhaus-filter/raw/master/urlhaus-filter-hosts.txt https://gitlab.com/curben/urlhaus-filter/raw/master/urlhaus-... https://raw.githubusercontent.com/notracking/hosts-blocklists/master/hostnames.txt https://raw.githubusercontent.com/notracking/hosts-blocklist... https://raw.githubusercontent.com/anudeepND/blacklist/master/adservers.txt https://raw.githubusercontent.com/anudeepND/blacklist/master...

Merman_Mike 6y ago
Can you explain why more is not usually better? I added the 4 lists you recommended to my Pi-Hole, which added a net new 73,253 domains to my Pi-Hole. My total is now close to 2M.
- PixyMisa 6y ago
  You could just blacklist *.com and be done with it.
  - freebuju 6y ago
    You joke but I would be most happy if all my web needs could be served on .onion addresses