“Maybe it’s more important to influence the agent than even the human, because with a human you influence one person,” Mobian cofounder and CEO Jonah Goodhart told Digiday. “When you influence ChatGPT, you’re influencing potentially all of ChatGPT. If ChatGPT changes what it says about [a brand], it’s massive and it’s more than any one campaign could ever do.”
If I was this dumb I surely would be smart enough to keep my mouth shut.
I forget the article, but some author wrote about how he wanted AI companies to ingest his stories so that there was always a chance his words and stories could influence their output. Not just “how they write”, but what they might recommend to someone who’s asking for suggestions as to what they should read next.
To me, this is super interesting! Because right now the publishing industry has a bigger problem than AI (it started just a few years before ChatGPT): Discovery
As in, you want to find a book that covers the things/keywords/details you want, but just searching book sites makes that almost impossible.
(Hint for the publishing industry: software developers solved this problem long ago… It’s called “user-driven tags”. Let the users add tags to your stuff! Stop being Nazis about controlling every little thing!)
Anyway, LLMs are actually pretty good at discovery if you give them a super detailed prompt. That’s actually the problem they were invented to solve!
They can still hallucinate and make stupid mistakes, but using an LLM to find information is probably the best way to use them. Note: This doesn’t mean, “use the LLM to give you the answer.” Noooo! That’s how you end up with rocks on pizza, haha. Instead, you use the LLM like a search engine to find the link or location of what you’re looking for inside of something else (e.g. find the website, find the spot inside a document).
AI is super unpopular on Lemmy, but out in the real world it’s actually quite useful for it’s actual purpose (and the accidental side purpose: Coding).