Comment by cornholio | Hacker News Reader

Comment by cornholio | Hacker News Reader

Parent Full thread

I believe there is a technical solution to this that does not involve sanitizing the whole Internet to the level of broadcast TV (good luck with that).

A "Child mode" flag would be set system wide when the device is configured by the parent to be used by a child. This should enforce a top down policy that only allows installing child safe apps or apps that propagate and enforce a similar policy. For example, Chrome would be installed but would only allow access to sites that respect the child safe flag. Youtube would load but only run videos and suggestions form a carefully curated set of publishers that respect the age. Etcetera.

Compared with previous approaches that failed due to lack of adoption (RASCi, ICRA) this system would provide a strong incentive for app and site owners to properly support the age flag, else they are simply invisible in the children market. A strong initial effort would be required from Google/Microsoft to bootstrap the initial whitelists.

This approach is contrary to youtube’s Business model. They’d rather just block kids. Which makes kids (and parents) just pretend to be 18.

Goog just needs to give up and partner with pbs or something to make kids videos a charity.

There already is a YouTube Kids that has a similar set of features.

https://www.youtube.com/yt/kids/

This tool claims to have those features, but does not. The article covers this.

It’s actually kind of worse, because it lulls parents into thinking it’s safe. And it’s even harder to police since adults never use it and the UI is different than the YouTube parents expect.

Yes. We found our two year old had found some pretty...strange YouTube Kids videos on her iPad, so we nuked YT Kids and only gave them PBS/Sesame Street.

And the "weird" videos from the article seem to still exist there too, the way I understand the article (I haven't checked myself, it was disturbing experience the last time).

I've tried it, and found it to be worthless.

The problem isn't technical at all. In fact, the problem stems from trying to use technology to replace humans. The problem comes because the thing that your entire proposal rests on, curation, isn't being done.

That's my point: don't try to do content curation, pass it down to the next level. A limited number of manually vetted publishers should have access to the kids sections, and that access should be a very valuable asset which they lose when publishing bad content; as opposed to a burner Youtube account anyone can setup in minutes.

Fine, lets say Google does that.

Then they're blamed for allowing a media oligopoly form, whilst ignoring small content producers. We're seeing this play out with the demonitazation of popular vid creators.

Long story short, with the public, damned if you, and damned if you don't.

Nobody would complain if Youtube offered a curated channel with guaranteed kid safe videos. If I type "Peppa Pig", I want the official genuine Peppa Pig, not some stupid knockoff.

The problem is that Youtube wants to show you an endless stream of content, and they never ever want to show you an empty search result.

So they show you sketchy videos uploaded by bots with no quality control, and their stupid algorithms somehow manage to always put the fake stuff next to the real stuff.

And to make it clear: Nobody is requesting that Google should block questionable videos. They just shouldn't shove them in your face like they do now.

To make it clear: People are very much requesting that Google block questionable videos on their Kids app.

There is a difference between not including something in a curated channel (like the youtube kids app) and blocking it.

Exactly. You see the same thing happen in every other spot on the internet where users contribute content, and only an algorithm controls what dies, what lives, and what gets rewarded. See also: the huge counterfeit goods problem on Amazon, review system gaming on Yelp, spam in almost any unmoderated online forum, etc.

This isn’t going to be solved with better algorithms or machine learning. The bad actors also have computing resources. I think it can only be solved by human curation.

I have to disagree, not because you're wrong, but because I am jaded. I think it will be solved by better algorithms and machine learning after it gets worse for a very long time. If there is a temporary replacement of algorithms with human curation the scale of the economic incentive to replace those workers is vast. An institution like Google or Amazon cannot look at a huge, low-skill workforce and not actively wonder how to replace them. I lived this personally as I was briefly an adwords editor. I was one of a huge cadre of BA's employed to enforce complex and messy rules about text ads. Simultaneously, embedded engineers watched us so they could better craft algorithmic replacements. Eventually they did.

That experience was a non-zero factor in my transition into a technical career. Big companies will automate anything they can and being on the automation side rather than the automated seems like the better idea.

As automation gets better, do you think the bad actors are just going to shrug, say “well that was good while it lasted” and give up? Their automation will keep improving too. I’d argue that the spammers’ algorithms will always be better than their targets’ because their effort directly impacts their bottom line, whereas their targets’ anti spam efforts have to only catch enough to avoid turning their legit audience away.

The same problems apply to email spam. It was bad for a long, long time but now it's a largely solved problem if you use (say) Gmail.

Isn't that because most of the spam centers simply get blocked?

It's instructive to view Hacker News' own "new" page with "showdead" set. The torrent of spam is astounding. There seems to be absolutely nothing online that people aren't intent on ruining.

This times a thousand. The problems we're seeing with social media sites and trolling, spam, harassment, low quality content, fake reviews, etc is because they're operating at a scale where human moderators are too expensive to use and stuck trying to use algorithms to solve human problems.

But that doesn't work. Until we have true human level AI (and honestly maybe not even then), decent curation has to be done by humans rather than machines. That's why forums tend to be less 'toxic' than Twitter or Facebook, because the moderators can view content and use their personal disgression while figuring out whether something breaks the rules and then act accordingly. It's why (somewhat surprisingly), the likes of SMW Central and MFGG have better quality control standards than Steam, the App store or other such services; because the moderators physically play through every game submitted and determine whether they're good enough to be accepted. It's why Amazon has an issue with counterfeit products nowadays, and why Google search has spam and blackhat SEO.

And it's not just humans that are needed for good moderation or curation either. It's humans with knowledge/experience in the field in question, and with enough interest to go all in on their work to make the community better. Those 'moderation sweatshops' some social media services use (the ones that check for illegal content and what not by getting a bunch of low wage workers to scan through content en mass) are as flawed an idea as the algorithms; they don't know what's acceptable beyond the very minimum.

But I guess that's the issue with many businesses and services and startups now. Everyone's trying to solve human problems with only code, and realising it (at least for the time being) just often doesn't work that well.

> Youtube would load but only run videos and suggestions form a carefully curated set of publishers that respect the age.

Isn't Youtubes whole schtick that they don't want to do any curation, but instead want to just turn the algorithms loose on the firehose of data that's being uploaded and recommend whatever the algorithm spits out? Pivoting to human curation from there sounds like a really difficult task for them.

They do SOME curation. I just need a flag that lets me filter out the 99.9% that isn't curated.

It's either that or I won't use Youtube (or rather, my kids won't). So it's definitely in their interest to curate.

They could whitelist a large number of content providers, and just rely on their judgement for this curated platform.

This could work; whitelisting is a far better and easier approach to trying to weed out "bad" stuff.

By far the best idea I've ever heard was the creation of a .kids TLD that would be essentially what you described. I believe there was a proposal for that that went nowhere.

The problem is who moderates it?

Social mores are dramatically different in different places. What if all content was required to meet Saudi Arabia's regressive social moores? Or the disconnect between the US when it comes to nudity and violence?

A moderated community can work, but it has to be much, much more granular, and the effort to achieve that kind of granularity is basically not feasable from an economic standpoint.

There are various voluntary ratings associations in existence, focused on their own version of "acceptable."

Commonsensemedia is a good example, they have a US-centric Christian bent that doesn't agree with my own, but for the most part their information about movies and TV for kids is useful to me when deciding what to show my kids.

So there seems to be a treasure trove of voluntarily produced data on content by various groups on one side, and a desire to automate applying those ratings to my kids in a default whitelist then allow-by-exception model.

If it were technologically possible, I'd even be thrilled to have a default whitelist created by the parents who are my Facebook contacts. We do that manually today. "Hey, have you seen The Incredibles? Is it too scary for my three year old?"

Anything that allowed parents to do that, and that worked in a widespread way, would be adopted immediately and would probably be worth a subscription fee.

Google Family Link does exactly this. App installs require parental approval and browsing is filtered.

except, YouTube Kids already exists, and the whole point of the NYT article was that it's not working.

Because it wasn't made the way the parent suggests, but relied on algorithms trying to identify appropriate content and after-the-fact moderation in case of complaints.

As a parent of a young child, what I want is a very limited list of videos; not the vast wastes of content that is YouTube.

Marble run videos ok. Crazy Russian science guy, mostly ok, as long as it's not too pyromaniac or alcohol. Thoughtful Minecraft videos ok, annoying kids playing terrible games and freaking out about nothing, no thanks.

YouTube kids was totally worthless. Either you turn off search and only see featured videos which nobody wants to watch, or you enable search and basically everything is there, because it only blocks a few categories.

Reply on news.ycombinator.com