Hacker News Reader

Comment by 13years | Hacker News Reader

The internet is going to become bedlam and madness drowning in the noise of auto generated content.

This is just the beginning of the unfolding of an enormous set of unexpected consequences of AI.

FYI,I have written extensively on other disruptions that are likely coming

"We had hoped technology would replace monotonous tasks while leaving us more time for creative exploration. Instead we are getting exactly the opposite. It is replacing what humans love to do. Previously it had been thought that creativity would be the area untouchable by machines. Ironically, it seems to currently be one of AI’s most successful areas."

from https://www.mindprison.cc/p/ai-and-the-end-to-all-things

View on HN

>The internet is going to become bedlam and madness drowning in the noise of auto generated content.

I'm less pessimistic, simply because this content is trash. It's not just that it's inaccurate, it's also that it's uninteresting and repetitive. Who will consume it? If no one will consume it, who will host it, and for what purpose?

I agree.

I also keep thinking: If ChatGPT (etc.) can write such content, and I'm interested in it, then why won't I just ask ChatGPT for the information directly rather than going to someone's bloated website?

That is an evolution I perceive coming at some point. It will be essentially the end of all middleware.

Leveraging AI to make your business more productive is a short term temporary evolution. Longer term most businesses will have no purpose.

Why is ChatGPT allowed to do it and not Google?

ChatGPT writes the averagest content possible. Only creatively dead websites will use it

A lot are hosting it because they found a way to generate revenue off of it.

"Per the report, over 141 "blue chip" — or internationally recognized — brands are unknowingly supporting hundreds of low-effort AI content farms."

https://futurism.com/content-farms-ai

Children raised on AI-generated storybooks, who think that’s just what “content” is.

(That’s unrealistically cynical… I hope.)

It will be worse. People already let cellphones and social media become toys/babysitters for children.

Next, AI will become your child's closest companion and they will bond with it greater than to their own parents.

Try finding anything good on the internet when 99% of search results are dominated by AI-generated garbage alongside the SEO-generated garbage we already have.

If I were substack I'd start figuring out how to ban AI generated content on their platform. Trust in content on certain platforms being non-AI seems like a significant selling-point going forward.

It is impossible. Reliable AI content detection does not exist. You could only implement some heuristics, like some unreasonable rate of content publications.

Authenticity is going to the make or break for this problem. However, I just haven't seen any promising solutions. Theoretically, there won't be solutions as the very intent of AI is to be human like in every way if achievable.

I recently tried a AI art detector while writing my latest - https://www.mindprison.cc/p/ai-art-challenges-meaning-in-a-w... - and found it to be abysmal. The false positives made it completely unusable.

Does reliable detection matter?

I think platform holders saying "AI content is unwelcome here", and loudly kicking off violators is more than enough to send the required signal.

The goal is not to have 0% generated content, it's to brand yourself as _the_ source for thoughtful, human writing. Having an occasional article slip under the radar is fine.

When you put a lot of effort into your own creation and then it is rejected because it is flagged as AI, then yes it will matter significantly. 50% of my own art I created is flagged as AI.

There may be an appeal process, but the point is AI will make verification of authenticity painful.

You're misunderstanding what I'm saying.

My argument isn't "imperfect detection is fine".

I'm saying that _you don't need detection at all_.

Just make it very clear that AI content is unwelcome on your platform and loudly boot everyone who tries to "experiment" with your limits.

Yes, that is really the only thing that you can do. Set expectations and reasonable heuristics.

What do you mean "going to become"?

I mean, not to just be a jerk or trot out a tired cliche for the heck of it, the "internet" - particularly the web - has, IMO / IME, massively degraded in the past 5 - 7 years. And, unsurprisingly (given the nature of these sorts of processes, looked at either from a more "naive" statistical / distributional POV, or more mechanistically), the rate of degradation is swiftly / noticeably accelerating (i.e., d^2/dx^2 >> 0).

The flood of "AI" BS is just about the most apropos way to continue what's already in motion. A way that could hardly be more perfect - LLMs and the like can work 24/7/365 recursing infinitely until there's no information content left at all, just a "gray goo".

The dark hilarity of our "intelligence" is endlessly amusing to me (yet another stupid human).

I don't disagree, there isn't exactly a test for abstract terms of bedlam and madness. So to whatever degree it already exists, AI will accelerate that an order of magnitude. It is the perfect tool to continue out the self destruction from our attention addiction.

The decay of internet and digital communication has been undeniably accelerating but this will push it to a level previously thought physically impossible.

The hill we're rolling down continually steepens and indeed is now very steep, but we're about to ignite a 1,000,000 ft-lb rocket booster with the engine nozzle pointed directly at the sky.

I disagree.

1) Garbage replaced with garbage is still garbage.

2) spam replaced with spam is still spam.

3) artisanal writing will always be higher quality and demand higher premiums, viewership, margins, and stickiness.

4) generative AI can be used by humans to craft reasonable boilerplate and edited by humans to add value and depth. Similarly calculators can produce mathematical results but they’re only interesting the context of a humans agency and synthesis. The use of generative AI for writing is more like using a code completer or calculator and at no stage removes a human - the AI must be prompted and guided, and as is shown in the article must be edited and post processed to be appealing. The machine covers only a small part of the exercise. Unless you’re making garbage or spam, in which case see (1),(2)

Which will make aggregators, filters, sifters all the more important.

Unfortunately the high quality solution is to throw humans in the mix and have their brains absorb the shrapnel, and they can approve what others should look at. Like techmeme.com and drudgereport.com

This why I'm skeptical of superintelligent AI.

The output of AI's is essentially lossy, so when you feed that back in as an autonomous loop you get the equivalent of a 100x re-saved jpeg, aka complete garbage.

AI's wont have any more useful data to gobble up.

I don't think we need any more useful data for the AIs to gobble up — we just need to make the architectures we train more data-efficient. In my mind, a few petabytes of text, images, audio, and videos from the pre-AI era should be more than enough to train an AGI, we just haven't figured out how to do it yet.

There's a book in that.

AI becomes ubiquitous, and provides media, information, and general assistance and is trained exclusively on content from before AI was on the web. This freezes pop culture in what, the late 90's through the 2020's? What are the implications for that moving forward? Do we stagnate? Do we become so reliant on it that innovation is dying?

I find that idea interesting.

It is a similar concept that I touch on a bit in the article I linked above.

In the section "AI dominance, stagnation and human skill debt" which mentions the idea that we could break the cycle of learning new information and skill where there is nothing to fill that void.

Makes you wonder whether the internet does something similar to our human minds. Too much information feedback and signal loss actually harming our ability to think and understand things deeply.

Reply on news.ycombinator.com