🏅 Top traders

Operationalize reliable please :)
e.g. which LLMs detecting, what text is being generated, required sensitivity/specificity/false positives/negatives etc

@CalebWithers I'd definitely bet no if it involves detecting general more generally (i.e. not constrained to certain types of prompts, tasks, or models)

Key word, "we." Someone may have it, but not, "we." There are too many LLM's, pandora's box has been opened. That being said, it is possible, just as companies have implemented software that accounts for sales tax in every single jurisdiction of every single town in the United States, you can build a mixture of expert software which runs a given set of tokens through 10,000 different LLM signatures and then filters down a response based upon the rating of the signature for each LLM silo. However 1.) That's really expensive to build. 2) Really expensive to run. 3) Would be highly custom/privatized, because who needs that? Google may have their own for SEO purposes that approximates the above, but that's going to be an internal tool and you would have to sign NDA's about it, they would not disclose the nature of that software.

@MarkIngraham didn't expect you to buy NO on this

@firstuserhere ai isn't important enough to develop solutions

@DuSusisu the market is about a software which can detect AI generated text from human text. Do you expect such a software to be reliable, by the end of the year?

@firstuserhere it won't exist

@DuSusisu it already exists. Just not good enough