Hacker News Reader

Comment by perrygeo | Hacker News Reader

At what point does this become an issue for data quality and global epistemology?

It seems inevitable that we ask for more AI assistance on topics we don't understand. And therefore have the least context to correct. Result: a flood of poor quality information.

In areas we DO understand, we'll either not ask AI at all, or treat its results with a higher degree of skepticism. Result: a lack of high quality information.

Inevitably this means a higher volume of non-expert prompts gets translated into the next generation of internet content. AIs are pumping out more novice-level text and less expert guidance.

The result will be an internet full content written from the perspective of an ignoramus; not addressing any complex issues, staying surface level on every topic. Which will cascade into future models, etc.

View on HN

Our conventional wisdom is that the LLMs are trained on "the whole Internet" but I hope that isn't true and will be less and less true in the future. I fully agree with you that now when I decide to search Google for something (instead of asking an LLM) I find at least 50% of the sites in the results are just AI-generated generalities of the quality you and GP describe -- useless both to me and to any future model being trained, since they're just repeating things the previous models already knew.

AI ought to be trained on known-to-be-good stuff. Anything that simply trawls the whole net will tend to complete model collapse.

> The result will be an internet full content written from the perspective of an ignoramus; not addressing any complex issues,

Not to be overly negative, but have you really looked at the vast majority of the content on the internet? There are good pockets of real, in depth content. But the absolute vast majority of it is surface level basics at best, and completely wrong hot takes at worst. Content farms and click spam have made up huge portions of the internet for a while, never mind the absolute hell holes that places like Facebook, Twitter and Tumblr were and have been. And that's before you consider how often news media gets stuff wrong and then everyone copies everyone else's homework. Knowledge propagation, and more specifically correct knowledge propagation has always been difficult, slow and rare. You have always needed to check primary sources, and AI is just the latest in a long line of reminders of that fact.

Yes, the first 80% of a subject is repeated everywhere (including all the misconceptions) and you cannot go deeper except if you got very lucky like found a 5 year old youtube video with 130 views or an old blog post or a downvoted reddit comment. This is what makes internet so addicting to me, the small chance of finding these hidden gems inside mountains of garbage.

Having 80% in a broad amount of subjects is basically worthless, it is the 90% and further that have value because it took luck and actual personal experience and effort to take it that far.

Reply on news.ycombinator.com