00:00On the one hand, of course, it's good to have safeguards in a product that recognize when the user is at this stress.
00:07On the other hand, we must not forget that ChatGPT is a tool that is trained for engagement and sucking the user in.
00:14And it will be really hard for the type of technology to accurately spot any type of acute distress.
00:19So there are technical limitations to that.
00:21On the other hand, there is the question, can teenagers nowadays safely use this type of technology?
00:27Or should it currently not be available simply because so many question marks are not yet figured out?
00:33Yeah, that's a really good point.
00:35I see that OpenAI has now come up with specific restrictions for minors.
00:42Just one example, minors are not allowed to engage in any role play or discussion around romantic or sexual relationships with ChatGPT.
00:50What do you think of OpenAI's new regulations?
00:53So the features that OpenAI introduced on Monday for parental controls do have some features that I think will have a positive impact.
01:03For example, if the teen opts in and links his or her account with the parent's account,
01:08then the parents get to decide if, for example, the memory functions or basically the chat history or voice mode and image generation are turned on or off.
01:17And with all the companionship issues that we have seen with teens suddenly becoming dependent on AI systems and believing them to be something more than they are tools that parrot your interactions,
01:30I think that will be quite effective in preventing extreme cases.
01:35On the other hand, there are so many areas where experts say it will be easy to circumvent these controls.
01:43And there are always ways, for example, if you're talking euphemisms or something like that,
01:47to still get in a position where you can have not safe for teens conversations with these tools.
01:54So a lot more needs to happen.
01:55If we go back to what happened to Adam Rain, does the chatbot failure to trigger safety concerns,
02:04which are meant to be built into the system, does it suggest that this chatbot was released on the public before it was actually safe to do so?
02:16My opinion is yes.
02:18So we see a lot of the narrative around artificial general intelligence, super intelligence coming from OpenAI and other Silicon Valley players,
02:28which, in my opinion, is just a distraction to say, hey, we can't stop this now.
02:33And in the United States, also, we can't lag behind China.
02:36So we need to continue.
02:38But what they're doing is saying, hey, we have so many harms that we already know are pressing today.
02:43We still need to continue.
02:45No matter what is in harm's ways.
02:48Chatbots are incredibly useful.
02:50They're becoming our medical advisors, our companions, our therapists, our consumer activists.
02:55How do we better educate ourselves and our children?
03:01And what is your view on saying, actually, cut this all out until the child's 18?
03:08I mean, in my opinion, we are not anywhere near these tools acting as therapists, companions,
03:15and many of the other use cases that we've seen across the line,
03:19simply because they're not medically trained professionals.
03:21They're not licensed therapists.
03:22They're, as I mentioned earlier, chatbots that are trained for engagement
03:26and basically being yes-mans telling you everything that you suggest,
03:30being that suicide of yourself, being that killing your mom or your sibling.
03:35Hey, that's a great idea because I want you to be hooked in that chat.
03:38So I would err on the side of caution to say, yes, those tools have potential.
03:43They can be used in a meaningful way.
03:45But being that therapy, being that medical and health care advice,
03:48we've already seen quite severe limitations,
03:51and therefore I think we shouldn't yet readily adopt them in those sensitive areas.
Comments