4 ms·
Yes, significant part of our effort is to build technology that detects LLM spam. We have a working model that detects LLM generated text with 90% accuracy curr
by freediver 3y ago
Yes, significant part of our effort is to build technology that detects LLM spam. We have a working model that detects LLM generated text with 90% accuracy currently. The plan is to integrate in search results and make available as an API.
- oktoberpaard 3y agoAnd just out of interest: what is the rate for false positives? An accuracy of 90% sounds very decent, as long as it doesn’t also filter out 10% of legitimate content.
- lupire 3y ago10% of infinity is infinity. That's as good as 0% accuracy. And AI text will keep changing.