Given how wrong/ridiculously oversimplified those AI summaries usually are, it scares me that so many people would stop there like, "Ehh, good enough".

I think Alk is referencing, the concept of perverse incentives. Without explicitly saying it. It's a concept, or a way of refering to an incentive structure that gives un-desirable results, in economics.

Example: When clicks give you ad revenue. And hurt kittens nurtured back to health gives the most clicks. People start hurting kittens, so they have more to nurture back to health for clicks.

Edit: my example is unfortunately a very real thing. Multiple channels on YT have been found to do it. Someone else will have to find the articles about it. I don't want to ruin my day reading about it again.

A lot of my queries only call for oversimplified summaries. Either I'm simple like that or I google stupid shoot no one else would bother. A recent example:

Are there butterflies or moths that don't have mouths? (No but some have vestigial mouths connected to non-functioning digestive systems.) Good enough!

That said, I'm very skeptical about answers if it's anything I care about or need to act on.

Are you Google's official bootlicker?

Less depressing example. A country (I think it was India but my memory is hazy and I'm too lazy to go google it right now) had a problem with a certain venomous snake. They decided to offer a bounty for every snake corpse brought to them. The goal was to incentivise people to hunt snakes. What actually happened was people started breeding the snakes to turn in for the bounties. They realized the program wasn't working and cancelled it at which point the breeders dumped their snakes into the wild making the whole situation even worse.

So I had this joke idea of "they'll just start showing the ads to the AIs", but the more I thought about it the more it started to sound less like a joke. Imagine if someone figured out how to cram ads into the AI training models and it skewed the outputs. Why astroturf when you can train the AIs to astroturf for you. This is some black mirror shit and now I've made myself a bit depressed.

There is an exploit used by scammers to change the official support number from summaries to theirs.
I dont know exactly how it works but its bad when you call the indian it support instead of your airline supportdesk

Youre right, it was in British India, where the British government came up with this scheme of rewarding the locals with money for captured / killed snakes in hopes of controlling the local snake population.

The AI answer mostly just parrots whatever the site that has won the referencement war is spewing. If it's easy enough, it can luck out and find an easy ready answer on wikipedia or something. Beyond that, most of those high referenced sites are the shitty aggregators that already pollute the search results.

I often search for the correct way to do do something. For example, there's a lot of baseless bullshit in gardening. If there wasn't an AI answer, I would not trust the first result and stop there, I would look for a few, check what sources they have. I would not even take the wikipedia answer at face value without at least confirming where they got their info.

We know AI doesn't do that. We have examples of it not even recognizing obvious parody, it can't be trusted with recognizing unsourced shit.

AI literally produces better answers than 99% of ad supported, SEOptimized websites.

That's saying not a lot about AI though. It tells you how utterly awful searching the web is thanks to those sites.

I'd say AI search summaries are somewhat useful for me 30% of the time. And I click through to the sources to confirm its summaries anyway, because they're often oversimplified.

Often though, they're goddamn useless.

OroborAIos

We do not need to worry. The amount of scraping and traffic those ais are doing are already killing every website. At least they all have full backups of the whole internet by now... Right? Righttt?

Yes, but that's besides the point.

Yes you said it better than I could have. Not only the perverse incentive, but also just the way ads have annihilated the usability of the internet for the average user. I know some sites can't exist without ads, but the web now is an unusable mess of for-profit click bait SEO slop and the average non-profit oriented enthusiast with a website for something has a harder time than ever existing because of it.

I am not smart enough to know what to change, but I know something has to change. Short of a complete upheaval of the current web, the ones profiting off the current model will do everything in their power to make sure nothing changes.

This is why I'm conflicted. AI destroying ad revenue is that upheaval that could be fast and powerful enough to disrupt the status quo, but at what cost?

I fear the world must be destroyed first, for humanity to rise beyond Capitalism essentially. The Capitalists will never give up their power and control willingly. To get to -> Advertisements are illegal and banned worldwide, we first need to become the Phoenix. Kinda doubtful personally that it'll ever happen, I feel like climate change and WW3 will ensure our collective extinction.

Thanks you for letting me know that my interpretation wasn't completely off base.

Well I have an idea and its a bit archaic but it just might work. Local homegrown news bulletins. Like how punk rock bands and other subcultures back in the day spread around. Registers and news, and compilations of cool sites you and your group of friends or "club" have found. The old internet had loads of sites or BBS's that were link lists.

It doesn't have to be janky paper magazine's. But communities need to engage more in genuine material and sites. Remember happy tree friends? No algorythm spread that. Kids did! Same thing with meatspin and all those crazy sites and content. Word of mouth is crazy powerful. Like take peertube for example, finding content you like there ain't as easy as on YouTube. But if you in a group / forum honestly recomend something or someone you found, chances are someone like minded that didn't know of it, now finds it. But we can't, on the other hand, go around and spam everything we find.

So monthly bulletins of content, sites etc in a forum would be my 2 cents.

Honestly that's kind of what lemmy is, in a roundabout way. I think you are right, but actually getting people to engage with that would be difficult. Today, word of mouth with younger people mostly revolves around individual things inside centralized platforms like a TikTok meme or something. I think in addition to independent sources of content, there needs to be a cultural change in how everyone accesses content. That's the hard part.

Indeed! But let's be the change we want!

Do you have any good mastodon, pixelfed or peertube or other fediverse recommendations to follow ? I'm still new here so my feeds are pretty monocultural (to borrow an agricultural term).

The top result is always some AI-gen, 2000-word essay response to a simple yes/no question like "Can a dog eat onions?"

I swear they do it to train us to just use the shitty AI summary of the shitty AI essay.