[kangalioo]

How does Pangram detect this? How do we know this is any more reliable than just asking any HN user to judge a text for AI signs, i.e. how is this more authoritative than the comments you're responding to?

[computerphage]

https://www.pangram.com/research/how-it-works

[raincole]

> How do we know this

We don't. Pangram is HN's astrology. It has 99.99% accuracy (according to Pangram.)

[computerphage]

They claim it has a very low false-positive rate, which is meaningfully more nuanced than "accuracy".

"Detect AI-generated content with 99.98% accuracy." -pangram themselves

[shantnutiwari]

you are being downvoted but yeah, all these ai detection tools are astrology. Fancy sciency looking, but still crap

[raincole]

The downvotes are understandable. People really wish there is a reliable way to detect AI-gened content. If there is such a tool thew Show HN post will break the record of the post score.

But there isn't and won't be such a tool. As a fallback people defend Pangram etc. Posting Pangram scores in comments is fundamentally no difference than posting ChatGPT responses in comments, and it's quite hilarious that the former is normalized on HN.