the spam and misinformation from #LLMs is a feature not a bug to the companies that develop them
when you sell the ability to detect AI spam either by filtering results or through some "knowledge platform," being unable to access information through other means is really good for business
Diluting informational power is the point. All fake everything is what they were designed to do.
@jonny Its just having just one more massive intermediary in between everything, similar to how problematic financial transfers are without using electronic payment companies
ik most people I talk to on here already know this, but I just keep seeing ppl say things in the realm of "dang the LLMs are cool, it's a shame they also generate massive amounts of fake text" when that is in fact the point of their having been designed and their most valuable feature.
the instructGPT paper, describing how platform capture is what made the bot usable, concludes by saying maybe the best way to keep these things safe is for only a few people to have them and rent out access via the API. being able to sell safety from the weapon you manufacture is an incredible racket.
Add comment