Comment by kristopolous | Hacker News Reader

Comment by kristopolous | Hacker News Reader

Parent Full thread

Tor sites do all of this still.

I don't advocate for their content but the way they've built things is pretty great.

I've described it as "the 1998 web with better cgi scripts".

Also search engines like marginalia (https://search.marginalia.nu) and teclis (https://teclis.com/) are nice - they essentially filter out pages with advertisements and trackers which is a pretty brilliant way to defeat SEO farms and other low quality content with ulterior motives.

To see their effectiveness give it a challenge. Try queries like "how to lose weight" or heck, "lasagna recipe".

To demonstrate, I just took the time to make two videos for "lasagna recipe" with the network tab of chrome's inspector window open:

google's first result: https://www.youtube.com/watch?v=ntC5ZcgxyD4

teclis: https://www.youtube.com/watch?v=GULoL8o6NDA

Along with recipe sites, there's recently been a crop of similar looking tech help sites that basically give w3schools level answers with a bunch of rotating ads based on the hypothesis that users will look at the content, then focus on their work for the next hour or two, keeping the tab open because that's just the common pattern while the ad rotations go on the page.

As far as serving ads go, programming and recipe sites exhibit similar usage patterns. Similar to recipe sites, the answers you came for are interleaved with ads in the hope that people drag their cursor over the banner ad while selecting text from the site or at least keep that part in the viewport open for the duration of the activity. What a lovely feature!

Search has to be driven by axioms of values and purpose because different intents yield different results. There is no silver bullet.

For example, google, bing and yandex's goals are pretty antithetical to mine unless I'm actively trying to buy something. On google, I came for a lasagna recipe and someone tried to sell me car parts and a home mortgage which is fine. Those engines are just a poorly chosen tool for that job.

>Tor sites do all of this still. I like browing onion sites from time to time, typically personal sites. It is resemblant of the wild west days of the web, except most onion sites are for drugs.

It's changed since the deplatforming last year.

There's a new flood of people who don't mind mixing violence with their politics, hate groups, and a few fascinating libraries that are vast archives of incendiary frauds, quacks and malicious propagandists of the past.

Many countries free speech rules are more restrictive than the United States when it comes to hate speech and calls for violence and insurrection.

I've been meaning to scrape those archives and email academic historians with a Dropbox link. Historical hateful screeds and conspiratorial fabrications are actually hard to come by (think of say, political opportunists trying to capitalize on the 1932 Bonus March with self serving conspiracies).

It's a long list of mostly self-published pamphlets, newsletters and books. When these mountebanks pass on since they were never actually affiliated with legitimate institutions, their papers don't get cared for.

This also goes back to the search problem. GBY explicitly excludes the most virulent and harmful of this material. Fine. Agreed. Totally legitimate. But there's no way to say "I'm interested in folklore and sociology, I know it's all fiction".

People take action, sometimes murderous, based on these fraudsters. For the purpose of crafting policy, laws, and documenting manipulation techniques for media literacy, there should be a quarantined place where their lies go instead of being intentionally scrubbed after it manifests some awful tragedy. Maybe that would be a good candidate for a separate search tool.

There's some decent ML work with NLP there since (1) many things are anti-semitic dog-whistles (such as Alex Jones using the term Globalist) and would probably cluster really nicely with the quiet part using seq2seq, word2vec and might even be achievable unsupervised, and (2) many of them are just crude copy/paste amalgamations from older hate sheets, like the protocols, secret world government by spiridovich or waters flowing eastward by fry; oftentimes with merely the names and events updated to be more contemporary while usually throwing a different oppressed group under the bus such as immigrants or trans people. In the results you could display this: "See side by side comparison with this literature from the British Union of Fascists" where you clearly demonstrate some "truth about gays" is just a crude find and replace of say "truth about jews" from 1935.

It might also be helpful to assuage the victims who are falling into these libelous traps by clearly demonstrating how obvious the fakery is. Of course some seem to lack the cognitive wherewithal regardless of the clarity of the evidence but we might as well give it a go.

Reply on news.ycombinator.com