5 ms·
There is inherent conflict in having both an AI tool business and an AI tool detection business. If the first does a good job, the second fails. And vice versa
by kylecordes 3y ago
There is inherent conflict in having both an AI tool business and an AI tool detection business.
If the first does a good job, the second fails. And vice versa.
(On the other hand, maybe there is a lot of money to be made selling both, to different groups?)
- sebzim4500 3y agoI 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.
- zarzavat 3y agoThere’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.
- sebzim4500 3y agoChanging 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.
- doliveira 3y agoThey literally already store the whole conversation...
- explaininjs 3y agoOne 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??