Hacker News Reader

Comment by p-e-w | Hacker News Reader

No privacy policy, no mention how search data is being used, and Twitter and Discord as the only contact options?

I like the idea, but those are some pretty major red flags.

Additionally, the promise of "No AI-Content" appears to be snake oil unless the creators have discovered a magic formula for reliably distinguishing AI output from human output. With up to 7 degrees of separation, your search network is pretty much guaranteed to include AI-generated content, which will be indexed like any other content.

View on HN

I was planning to add those pages and did not expect Grep to be on the front page of HN.

> Additionally, the promise of "No AI-Content" appears to be snake oil unless the creators have discovered a magic formula for reliably distinguishing AI output from human output.

From my testing, at least within your network, I did not notice any AI-generated content. Even if you find any low-quality sites, you can simply block the previous node that refers to the site (yet to be shipped). Grep will remove all the sites that the previous node was linking to, including any AI-generated ones.

> I was planning to add those pages and did not expect Grep to be on the front page of HN.

Fair enough, but a privacy policy isn't a "nice to have", it's a legal requirement (in many jurisdictions) before you can start collecting email addresses and other PII. It's a very bad idea to push such things to later.

> From my testing, at least within your network, I did not notice any AI-generated content.

That doesn't mean anything. Humans cannot reliably recognize AI-generated content. Why do you think all kinds of institutions are so afraid of it? If it were obvious whether something is made by an AI, there would be no problem.

> Additionally, the promise of "No AI-Content" appears to be snake oil unless the creators have discovered a magic formula for reliably distinguishing AI output from human output. With up to 7 degrees of separation, your search network is pretty much guaranteed to include AI-generated content, which will be indexed like any other content.

You can fairly effectively apply network topology heuristics to the link graph to separate the wheat from the chaff. Content farms tend to have different linking patterns from organic content. (Overall, few humans tend to link to them)

That lets you exclude most content farms, but doesn't do anything about AI generated content.

It's more or less the same thing.

I hope that's true. If it's just content farms that are affected, I'd be fine. My concern is that it will happen beyond them.

No. Many high-quality websites contain AI-generated content. Illustrations made with Stable Diffusion have become common on hacker/dev blogs. AI-generated text is posted everywhere ChatGPT is being discussed. The technique you describe doesn't filter any of those. So "No AI-Content" is at best highly misleading, and, if you're being a little less charitable, flat out false.

Reply on news.ycombinator.com