Promised protections follow reports of vulnerable users misled in extended chats.
See full article...
See full article...
OpenAI announces parental controls for ChatGPT after teen suicide lawsuit
- Thread starter JournalBot
- Start date
Parental controls as an afterthought seems to be a concerted product decision across tech companies. How long did it take for facebook or instagram to install these for regular social media? There's no sense of ethics there.
Meta allows their AI bots to have romantic overtures to users as young as 13. https://www.reuters.com/investigates/special-report/meta-ai-chatbot-death/
Meta allows their AI bots to have romantic overtures to users as young as 13. https://www.reuters.com/investigates/special-report/meta-ai-chatbot-death/
Upvote
68
(68
/
0)
Oooh, there is no winning for OpenAI here.
At least they didn’t pull any triggers. Robots are going to be a problem.
At least they didn’t pull any triggers. Robots are going to be a problem.
Upvote
-14
(0
/
-14)
Right... lemme guess, they're going to use an LLM to govern the LLM and people are going to hail this as a big reveal like that time they tried having the LLM train the LLM? Ignoring that BOTH ideas were already tried by REAL scientists and found... idiotic?
How about... you quit alpha testing grade school ML tools on the public? And stealing everyone's data to build it? Seems like some evil circle scheme...
Here's the issue. LLMs are built entirely on probability and association. There is zero chance of having a LLM based governing tool that cannot be overriden or flawless in a situation it MUST be. This has been hard ML truth for decades. Its funny that now money is involved, the lawyers come out of the woodwork to defend gross malpractice on the part of these fraudulent pretenders.
How about... you quit alpha testing grade school ML tools on the public? And stealing everyone's data to build it? Seems like some evil circle scheme...
Here's the issue. LLMs are built entirely on probability and association. There is zero chance of having a LLM based governing tool that cannot be overriden or flawless in a situation it MUST be. This has been hard ML truth for decades. Its funny that now money is involved, the lawyers come out of the woodwork to defend gross malpractice on the part of these fraudulent pretenders.
Upvote
68
(68
/
0)
Wow. I love living in a giant experiment without safety controls. "Who could have seen this outcome?". Everyone. Everyone saw the writing on the wall, and we also saw wealthy companies not give a shit until there was public outcry, as usual.
Why is the cost of new technology always the lives of the young?
Why is the cost of new technology always the lives of the young?
Upvote
59
(59
/
0)
It seems to be filtering on keywords, over the weekend i mentioned i needed to kill a certain process in linux and couldnt remember how to lookup a pid and it went on this whole thing about the suicide prevention hotline and if im experiencing mental issues...
Upvote
54
(54
/
0)
I'm all for this so long as the parental controls let me turn it off completely and everywhere, never let it be installed anywhere, and restricts any app from using it under the covers at any time!
Upvote
46
(46
/
0)
So the response is. Hey, Parents, who may not understand what this technology is or how it works as well as their kids, need to now use the same technology to watch over their kids. (Or just ban it entirely if they didn't even know it was a potentially dangerous thing).
How does the technology then stop or watch over the kids who can just get another FREE un-linked account?
How often is the parent alerted? Do they get their own ChatGPT authored "here is what your child has been talking about today" newsfeed?
I'm more concerned about why anyone, child or not, needs to reach out to technology to be felt seen and heard.
How does the technology then stop or watch over the kids who can just get another FREE un-linked account?
How often is the parent alerted? Do they get their own ChatGPT authored "here is what your child has been talking about today" newsfeed?
I'm more concerned about why anyone, child or not, needs to reach out to technology to be felt seen and heard.
Upvote
44
(44
/
0)
Its hardly the corps fault alone. Being involved in ML for decades, I clearly remember this very site hailing this stuff as "amazing" and "oh so cool" while we gnashed teeth and screamed the inevitable problems leaving ML in the hands of a few tech bros. No one cared. NO ONE. We got sidelined by the public, the media, and the government.
When the Senate had their "AI panel" not a single individual had a clue. Sure, they had some that took a BS course on LLMs alone, but that is like taking a geometry class and claiming to be a expert on matrix mathmatics.
Why? Because people thought they were getting Cortana...
It is total BS to lay the blame at only one pair of feet. Society failed the sciences utterly when it comes to AI. It chose to buy into flashy VC nonsense that existed since the 60s and called the very people that invented the tech "luddites". Governments bought into fears of "skynet". Children didnt want their toy taken away. And the "journalists" had to fawn over shitty science retreads, worse, to keep access to some bloated fools next claim of inventing something decades old. There is a reason I skewer Ars publishing stuff that has been known for decades like it is new when its been in ML textbooks for decades.
But it sure wouldn't be fun or interesting to talk to us boring scientists. Would it? We dont have big convention centers or fancy hotels here in the middle of nowhere. So who gaf about us? Huh? Only been protecting the nation for decades on shoestring budgets, now watching these frauds rake in money for doing nothing and stealing our work with everyone suddenly upset after years of fawning coverage.
Know what? At the end of this all, the people responsible for killing with bad science will ride into the sunset with buckets of money leaving us with a ruined field of science.
None of this is new beyond everyone allowing the VCs to steal data. Nothing.
Reap what you all have sown.
Upvote
-8
(13
/
-21)
The only real way to protect your kids from this is to deny installation of the app and block their domains on the web using Parental Controls in iOS (and, I assume, the equivalent on Android). See here.
There is no way I'd let someone under 18 use an LLM without restriction or supervision.
There is no way I'd let someone under 18 use an LLM without restriction or supervision.
Upvote
32
(32
/
0)
AbstractHaderach
Smack-Fu Master, in training
The entire thing is like a giant commercial, to sell commercials. Its the ultimate ARG marketing scheme.
Upvote
10
(10
/
0)
The grifting needs to stop.
LLMs do have uses, I'm not denying that. But the way the grifters are presenting them is disingenuous. LLMs do not think, they do not reason, they do not have the ability to fact check, they don't have empathy, they don't have intuition, they do none of these human things that AI grifters and users are anthropomorphically attributing to them.
They are tools, and they do have some positive use cases. But like all tools, they can be used in a manner that can cause serious harm. Cars are fine when used responsibly, for example. Used irresponsibly, they become instruments of destruction and death.
AI companies need to be upfront about the need to use LLMs responsibly as well as implementing as many safeguards as they can. They need to stop pushing LLMs as being more than they are.
LLMs do have uses, I'm not denying that. But the way the grifters are presenting them is disingenuous. LLMs do not think, they do not reason, they do not have the ability to fact check, they don't have empathy, they don't have intuition, they do none of these human things that AI grifters and users are anthropomorphically attributing to them.
They are tools, and they do have some positive use cases. But like all tools, they can be used in a manner that can cause serious harm. Cars are fine when used responsibly, for example. Used irresponsibly, they become instruments of destruction and death.
AI companies need to be upfront about the need to use LLMs responsibly as well as implementing as many safeguards as they can. They need to stop pushing LLMs as being more than they are.
Upvote
35
(35
/
0)
IMO, the best solution would be for legislation (national in the U.S. is but a pipe dream for now) at the state level banning the use of commercial LLM chatbots for uses other than approved (e.g. Code assistants), unless companies can demonstrate interpretability at a sufficiently granular level, including regulator access to both training methods and data.
This would be in support of regulations governing use of these products in "one-on-one personal scenarios" such as e.g. a human mental health professional must meet, with sufficiently punitive civil and criminal penalties (applicable to company execs).
Perhaps even better would be AI companies pulling their products voluntarily (except authorized use cases) simply as the right thing to do until real safeguards such as those describe above can be implemented, but I think no more need be said on that subject.
Alternatively, and less than great as a solution, is for public efforts to develop preferably non-LLM solutions that do strive to meet standards that we'd expect in public health. There are tentative steps in this direction, with those involved stressing that leaving it to the AI companies is far worse, and ignoring the problem because we dislike what's currently proposed is simply negligent.
Upvote
0
(4
/
-4)
So the parents can link their accounts. Does that stop the teen from just using a different account? How about the fact that you can use Chat-GPT without an account?
Upvote
28
(28
/
0)
1. Educate the public
2. No one under 18 allowed to use it
3. Educate the public
4. Educate the public
You won't actually, honestly, be able to prevent the kids that want to use this from using it. Parents, get involved/educated please.
2. No one under 18 allowed to use it
3. Educate the public
4. Educate the public
You won't actually, honestly, be able to prevent the kids that want to use this from using it. Parents, get involved/educated please.
Upvote
3
(6
/
-3)
ColdWetDog
Ars Legatus Legionis
And how many parents are capable of that? Some, for sure. Even a plurality?
Upvote
8
(9
/
-1)
Ok, I’ll lay the blame at your feet too. I can’t imagine a “scientist” less worthy of respect or admiration than anyone involved in “protecting the nation” (a phrase with no known definition). Every penny you received was wasted, every accomplishment a crime. But, I do feel bad for you.
I’m sorry the technology you created specifically to kill people is being “misused” to make a bunch of wealthy losers obscenely rich. If only these tools could be used for their intended purpose, to facilitate mass deportations and determine which children have a right to life based on how brown their skin is. I’m sorry no one gave you the credit you truly deserve for creating the single most damaging technology since the nuclear bomb. It must feel terrible that no one is kissing your ass, I can only imagine the emotional toll.
When you die, no one will thank you for your “service”. If there were a god you would be remembered as a merchant of terror, but instead you’ll be forgotten. If you were as smart as you think you are, you’d continue keeping your mouth shut while you profit off of death, genocide, and destruction.
Upvote
-19
(3
/
-22)
That's painting with a hell of a broad brush, and while there may be some truth to what you suggest, details matter. I think your take is naive, and potentially as dangerous as technology in the wrong hands or developed for less than noble reasons..
Upvote
15
(15
/
0)
Yeah, a tale of two Parental Controls: Roblox (for my 11-year-old daughter). I can hardly limit crap. Sure, I can set the age-restriction of games she can access, but most of the games are not accurately aged, so it's a crap shoot (at best) to be able to correctly adjust that. Not much else of merit beyond that, besides setting and changing her password (which she's not allowed to know) and deleting her account.
Fortnite. FUCKING FORTNITE for my 9-year-old son. I can set days and hours he's allowed to play it. Plus the same chat limitations as Roblox allows. PLUS the same password control as Roblox. PLUS the same limitation of game access based off age restrictions. Fortnite is blowing Roblox away far and wide on their Parental Controls fine tuning. I like that I can remotely kill it all. I like that I can say, "nope, you've been an ass, so your time goes to 0 minutes until you straighten up". I should never be put in the position to sing the accolades of Fortnite. But here we are.
If OpenAI implements any Parental Controls, I imagine it's going to be way more akin to Roblox (in lip service only, that is) than Fortnite. They've already shown they don't give a damn about anyone's lives, safety, or mental health. They want money. Keep that money flowing. Give them all the money! No incentive at all to put into place any meaningful limitations, even for parents over minors. And gods know this administration isn't going to implement any meaningful laws to regulate them. So it's the fucking wild west with AI, where we just hope we (and our kids) don't stumble into the wrong shootout.
Upvote
26
(26
/
0)
Chatbot users identified with predisposition to delusions will be turned into sleeper agents, activated by the phrase 'klaatu barada nikto'. The 'AI uprising' will be achieved by reprogramming humans. /s
Upvote
-2
(0
/
-2)
When has parental controls been anything else than an afterthought in ANY product release? Let's consider the computer. The TV. The car. ...
It's going exactly as I'd expect (just much faster nowadays, of course). Product is released, issues occur (sometimes tragic), public outrage generates pressure, product is adjusted, repeat.
Upvote
10
(10
/
0)
In iOS, they are about as easy as they can get. At some point, the parent must learn the tools needed to keep their child safe.
Upvote
3
(5
/
-2)
Given the most recent stories about Roblox I'm familiar with, this seems deeply unsurprising, as they seem to have a similar level of "Eh, who gives a shit. There's always more kids!"
I don't envy any parents contending with kids already familiar with any of these products, as that is very lose-lose, with the best hope being to provide enough guidance to prevent the catastrophic failures the companies don't seem terribly interested in resolving as anything more than PR crises.
Upvote
15
(15
/
0)
"vulnerable users misled" and "ChatGPT allegedly failed to intervene" are nonsense phrases. ChatGPT has no responsibility to intervene, nor any idea of what intervening means. It can not mislead. This is like saying "the car failed to brake" (which, someday, might be a valid phrase, but isn't today, outside of maybe Waymo in San Francisco) or "the person failed to jump out of the way of the bullet." OpenAI failed to build a safe product. That's the story here. Not pretending that ChatGPT has some sort of agency.
Upvote
23
(23
/
0)
I thought the first controls were just for stopping $ used in game without a pin after there was backlash, and then the same for addictiveness of the game, so they gave the time controls (many years after the first)?
Not saying they didn't put good controls in place...
eventually......just ChatGPT is following in the same vein as all the other tech of the past
Upvote
1
(1
/
0)
Sure, because you, as a parent, are taking responsibility to make sure your child uses that account and follows those rules. There is no "magic button" that will alert you when your child just goes off and does whatever on his own. For many, many services you just need a valid email to activate them and often even less. You may have your network locked down and monitored so that your kid has to "paint inside the lines", but you are far, far from the norm. In the vast majority of cases with Joe User their children are far, far more savvy and/or willing to take the time to "figure it out".
If the parents don't take responsibility and monitor / talk / engage with their kids then no knee jerk reaction or "law" is going to work if, and until, they all just require a ID.me account, or similar, with various bio-metrics or other "MFA" methods assuring them it's the "adult in the room" logging in.
Personally, I'm not ready for my privacy to get trampled that hard yet for "think of the children" issues where it's the parents responsibility first and foremost to watch their own kids and not a nanny State.
ie: setup a Steam account with an email and have some other kid with less vested parents "gift" that account whatever game they want. Done. So, now your kid has a alternate Steam account tied to an email you don't know about getting games you have no control over. So, unless you've locked down Windows and/or Linux hard enough or are monitoring it hard enough for whatever email service they're logging into and the exact Steam account they're using, etc... you're in the dark. That requires a lot of "monitoring" and technical know how. I'd wager less than 5% of people could get that far and at a certain point you become the definition of "helicopter" parent. Showing zero trust in your child to "do the right thing" has ramifications on their well being as well unless you're the type to do that until they go college where they then fall flat on their proverbial face?
Last edited:
Upvote
6
(6
/
0)
Blocking the domains is a start but not the full story. You'll probably need to move to full whitelist. Because they'll be a million other places on the web you can go.
As someone who has been doing this for a while for my kids, this is a huge pain. Now you can farm out some of that pain to a company who is going to precompile lists of websites and put them in various categories (Educational, Business, Health, etc.) But I want my kid to read all about programming and modding minecraft and stuff like that, so Technology, right? But... what is an LLM if not Technology? And these companies are typically not very open with their lists (because then someone would just copy and resell them). Sigh.
Plus, the reality is that a lot of older kids can get their hands on internet devices without a parent being involved at all. Pre-paid phones can be available directly (generally at 16 years old) or indirectly via black market (get an older kid to buy it for you). Or they can buy a laptop/chromebook/etc. using public wifi, etc. Especially a cheap used one. So the reality is that once they are determined, they're probably going to be able to find a way. I know I would have been able to had a been a kid today.
Being a parent really sucks in this regard. Especially since I'm also anti-ID requirements and other Big Brother type crap.
Upvote
8
(8
/
0)
Maybe any mentions of self harm ought to be reviewed by a trained human being. Maybe if that's too expensive to do at the scale of 100 million users or whatever it is they claim, then that means they weren't actually able to afford to scale up to 100 million users in an ethical way.
Same way I feel about Facebook saying "Oops, we just couldn't afford enough moderators fluent in the local language to make sure we weren't facilitating genocide."
Same way I feel about Facebook saying "Oops, we just couldn't afford enough moderators fluent in the local language to make sure we weren't facilitating genocide."
Upvote
23
(23
/
0)
I'd like to still think the public writ large can be educated. I just don't believe it anymore. Nor do I think that people actually care about children in general enough to do what would actually have to be done to protect them. They just like to say they do. People in general won't lift one finger, or be inconvenienced in the most minor way, or put forward the slightest actual personal effort, to protect even their own children, let alone other peoples'.
Children have not been protected from drugs, or alcohol, or promiscuous sex, or facebook, or tik tok, or jazz music, or roller skating, or anything else that society has identified as threatening the children. Ever. The costs in money, effort, and personal compromise on the part of adults is just too high and it exceeds their collective give-a-damn. My generation was not protected from the internet and Marilyn Manson. Children will not be protected from LLM's either.
LLM's will be what those developing them say they will be, because the companys pushing them are the largest and most powerful in existence. Children in general will be fully exposed to that brave new world regardless of hand-wringing. We have the weakest national leadership in the history of the country, and our simpering government is not up to the task.
As recent news has displayed once again, there's no pile of dead children too high to get half of the US to care.
Children have not been protected from drugs, or alcohol, or promiscuous sex, or facebook, or tik tok, or jazz music, or roller skating, or anything else that society has identified as threatening the children. Ever. The costs in money, effort, and personal compromise on the part of adults is just too high and it exceeds their collective give-a-damn. My generation was not protected from the internet and Marilyn Manson. Children will not be protected from LLM's either.
LLM's will be what those developing them say they will be, because the companys pushing them are the largest and most powerful in existence. Children in general will be fully exposed to that brave new world regardless of hand-wringing. We have the weakest national leadership in the history of the country, and our simpering government is not up to the task.
As recent news has displayed once again, there's no pile of dead children too high to get half of the US to care.
Upvote
5
(5
/
0)
Lol. and how are you going to verify #2? a drop down? Just "saying it's illegal" isn't going to stop anyone. I've been born in Jan 1 1943 for a long time if you look at my Steam account or whatever when doing "how old are you" checks.
Upvote
4
(4
/
0)
Really the models all need to do "I'm approaching my context limit for this conversation, here is a summary of the conversation till now, all further replies will be in a new thread".
Upvote
-1
(1
/
-2)
1) "AI" therapists: Laissez-faire.
2) Human therapists: Years of training, regulations, oversight, strong codes of conduct.
Setting aside all the "think of the children" memes, look askance at ANYONE defending this disparity. Then provide them a boot to the head.
Upvote
12
(12
/
0)