Kid May Already Be Watching AI-Generated Videos on YouTube | Hacker News Reader

Kid May Already Be Watching AI-Generated Videos on YouTube

wired.com

19 pointsBrajeshwar36 comments

It's well known that Youtube already has a sea of bizarre garbage content targeted at children. It's hard for me to imagine that AI could even make it worse than it already is.

Idk just making many more things than humans could possibly filter is worse to me? A volume of content which is impossible for humans to sort through, which occasionally have something incredibly disturbing in them, generated by something fundamentally incapable of understanding and empathy. Things can be much worse than 'kind of boring'. It's sad that AI will probably be used more often for kids content at first because lots of parents don't really pay close attention.

We already got to that point like 10 years ago.

You don't need LLM's or diffusion models to overwhelm "moderation at scale" because "moderation at scale" was a fantasy in the first place, and its cracks had already been exposed and widening.

Maybe you were hanging onto some optimism and needed the jostle of this new technology to see how doomed it all was already?

I saw this when visiting relatives. At best it was auto-playing mario party game or something... At worst something entirely bizarre and pointless.

The later might as well been computer generated...

And it's stupid to get moral panicked about.

People are 99% nature, 1% nurture. We evolved to survive worse than witnessing pregnant Spider Man.

When you consider that children already watch trash on youtube, then yes I agree that there is no reason to get worried about anything. But the entire concept of letting your children watch trash on youtube is worrying.

Not having children myself it's easy for me to say that everything on the internet should be treated as if it was 18+ and therefore children don't belong on the internet. Period. Especially not on their own. There are exceptions of course like websites that are highly geared towards the education of children but youtube (kids) in it's majority is not that.

People are absolutely NOT 99% nature.

People are born with some predisposition to certain behaviors, but that is far from being set in stone.

Good, attentive parenting is the difference maker.

I think you'll find studies disagree with you.

There's not much you can do apart from providing a decent baseline and getting lucky. Which, on the one hand, is profoundly relaxing; on the other, it's pretty depressing and scary.

Social science “studies” are almost all complete garbage. Non reproducible and disproven constantly.

Social scientists are LARPing as actual scientists. There is no rigor. Most people in the hard sciences have little regard or respect for their work.

Arguing that parents play little/no role in the development of children is peak academic nonsense.

Studies like what you’re citing only give comfort to the parents who would rather live life looking at their phones than parent their children.

This isn’t just merely witnessing something, a better take would be ‘have we evolved to be be raised by videos of pregnant spooderman.’ I think that’s a negative, ghost rider.

Source? Didnt realize the nature/nurture debate was solved

Spoken with such confidence...

Ai generated is an upgrade compared to the disturbing elsaXspiderman material humans made. It is the curation thats the problem, not who and how the children content is made.

Was that stuff human made? I assumed it was ML output even at the time.

Humans responding to the rewards of the algorithm, as it turns out...

https://en.m.wikipedia.org/wiki/Elsagate

I do t see anything in that article identifying the source of the videos.

"On November 23, French-Canadian outlet Tabloïd released a video investigation about Toy Monster, a channel linked to Webs and Tiaras. They confronted the videos' creators – based out of the south shore of Quebec City – who refused to be interviewed. One of the actors featured in the videos anonymously stated that he was contractually obligated to refrain from commenting. The investigation revealed that identical content was being posted on numerous channels apparently operated by the same people."

This seems to be the video: https://www.youtube.com/watch?v=oSqqIlKHgpo

Yeah that doesn’t say how the video was being generated. I meant how, not who. It could have been ML-based.

I'm not clear how you think the ML of 2017 would have been useful for generating these videos? This points to specific actors and creators, who I expect were using whatever was cheap and easy to produce a pile of content with in 2017.

Pre-2017 tech would have been up to the task. You'd create a bunch of generic animations and characters (all rigged up to the same skeletal system so all the animations work on all the characters), and design composable elements for a scene. Then use ML categorization techniques to label both popular YouTube videos and your pieces. From there it would be mostly automated: the ML agent strings together scenes to create a bunch of videos, then use reinforcement learning to adjust and mutate content based on some metric of view counts.

My five year old watches some YouTube under supervision, and the algorithm has been recommending a new variety of "quiz" videos that are partially or entirely AI generated. It's usually around a theme, such as Super Mario Brothers, and it asks open-ended questions like "Would you marry him?" without revealing the "him" until after a countdown. Then it shows an obviously AI-generated image of a character such as Bowser.

Another variation asks the viewer things like "which car would you pick?", with options to select the "black", "white", "rainbow", or "gold" gift box, then after a countdown shows a grid of AI generated images corresponding to the colors, where one color seems to be the intended "correct" answer, but it is left to the viewer to decide.

These videos are dumb and lazy, and I'll bet they'll become a lot more common soon.

If you’re a parent with a young toddler who watches Cocomelon, you should learn how they decide to make their reedits.

The cocomelon staff creates a testing environment with a child watching two tvs:

Tv 1 with Cocomelon on Tv 2 with mundane video like gardening on

When the child shifts their attention away from tv 1, the cocomelon staff reedits the scene to capture more of the child’s attention.

It makes me value high creator effort shows like Peppa Pig and Bluey even more.

Sometimes I think the old cartoons that were churned out by Korean(?) artist-sweatshops were no different from AI-generated. People with an imperfect understanding of Western culture and values, drawing vague representations of 'American stuff'.

Everything is synthesis in the end. Americans churn out vague representations based on imperfect understandings too.

The koreans only drew the inbetweens for most cartoons, the storyboards and scripts were written by americans.

Which ones do you mean? This made me really curious.

I’m pretty sure this article was at least partially ai generated

I imagine most of the new dialog on TV & theaters is heavily AI assisted. Still human edited and curated though.

Compared to all the effort that goes into a series or film writing the dialogue seems like a relatively small amount of effort.

Except for stuff that is clearly low effort in the first place.

To what end?

Studios could have replaced the expensive professional writing talent they compete to hire with cheaper and worse writers already, even while staying on good terms with the WGA. Yet they were making a choice to pay lots of money to people who were charging more. There's a reason for that!

Writing is not a cost center for modern production and trimming that cost gets you little except a much higher risk of generating garbage that doesn't work for your audience. Studios pay talent for reliability, consistency, and team compatability -- and that's worth quite a lot to them.

Might generative AI have a place in film/TV writing eventually? Maybe. But it's going to take a lot more than what current LLM's deliver. You're a decade ahead of yourself.

There are mediums that likely are moving towards content generatirs, but film/TV just isn't going to be one of them for a while. They just don't really need what it currently offers.

I guess I'm of the opinion that most of the writing in Hollywood recently has been total crap.

brain nursery egg tv is incredible

TikTok filters are AI generated video, right?

Think of the children! The (probably paid for) fear mongering campaign is getting rediculous.

“If you tell a big enough lie and tell it frequently enough, it will be believed.”

― Walter Langer

What's the lie? That AI is a systemic danger and should be heavily regulated and only large corporations should control it for the safety of civilization.

What's the truth? AI is disruptive and a threat to those very large corporations that are pushing the big lie.

Won't someone think of the small corporations??

Monopolies are bad, mm kay.

Reply on news.ycombinator.com