Hacker News Reader

I don't think this follows. If they wanted, they could crypographically bias the sampling to make the output detectable without decreasing capabilities at all.

Only people using it deceptively would be affected. No idea what portion of ChatGPT's users that is, would be very interested to know.

View on HN

There’s a much more effective way: store hashes of each output paragraph (with a minimum entropy) that has ever been generated, and allow people to enter a block of text to search the database.

It wouldn’t beat determined users but it would at least catch the unaware.

Changing every 10th word defeats that strategy but doesn't defeat a cryptographic bias.

Also the cost of storing every paragraph hash might eventually add up even if at the moment it would be negligable compared to the generation cost.

They literally already store the whole conversation...

One solution is to store a hash of every n-gram for n from 2 to whatever, then report what percent of ngrams of various lengths were hits.

Did someone say Bloom Filter??

Reply on news.ycombinator.com