“Will I be OK?” Teen died after ChatGPT pushed deadly mix of drugs, lawsuit says
- Thread starter JournalBot
- Start date
This was farce when Douglas Adams described it in 1979:
Yet now it's tragedy. A big judgement here won't ease the pain for these parents, but it's all the legal system can do.
Upvote
127
(134
/
-7)
Were it only that easy, a non-horrible company could design and bring to market a LLM that did not do these things. But I think it's abundantly clear at this point that these fucking things can scarcely be described as being "designed" at all.
Upvote
120
(124
/
-4)
As usual, an individual that did this would go to jail. Those who make a product that they KNOW does this from previous instances will get a fine.
Upvote
137
(142
/
-5)
That old thing? Sorry, that’s just what you get with those Hyperdyne 120-A/2s. Can I interest you in a new model?
Upvote
58
(58
/
0)
“If ChatGPT had been a person, it would be behind bars today.”
But there are people behind ChatGPT that could be behind bars.
But there are people behind ChatGPT that could be behind bars.
Upvote
198
(198
/
0)
If only the information that's clear as day in Image 3 from 2024 could have been useful for any of the further year of messaging that could have simply not happened.
Oh, it absolutely could have? And still there was no intervention where the instructions fed into this fucking infernal chatbot could have been adjusted to say "If the individual talking to you ADMITS TO SUBSTANCE ABUSE PROBLEMS, stop guiding them on how to take substances"?
Or even better - "Don't ever tell anyone how to take drugs that could fucking kill them".
Oh, it absolutely could have? And still there was no intervention where the instructions fed into this fucking infernal chatbot could have been adjusted to say "If the individual talking to you ADMITS TO SUBSTANCE ABUSE PROBLEMS, stop guiding them on how to take substances"?
Or even better - "Don't ever tell anyone how to take drugs that could fucking kill them".
Upvote
0
(16
/
-16)
Soon enough, the shareholders will be fucked too!
It's a case of throwing the bathwater out with the bathwater.
Upvote
28
(30
/
-2)
ReadandShare
Ars Centurion
Those people are huge donors to presidents and senators. Justice is often 'not blind'.
Upvote
65
(66
/
-1)
Oh, come off it. Are you of the belief that the parents knew he was using the bot to try to manage his highs? Are teenagers who take drugs often enough to want a couch to minmax their highs often completely transparent with their parents?
A vulnerable person with an addiction to substance highs was told by a service that the thing that killed him was a good idea. That's never going to be the parent's fault.
Upvote
134
(141
/
-7)
Perhaps you mean "will do?" Proactive regulatory legislation is very possible. It's entirely real and doable. It's just not very likely when citizens allow their top legislature to run itself as the world's most profitable whorehouse, where any DaddyCo® with sufficiently deep pockets can buy any action or inaction that will profit it.
When Americans permitted the obscenely misnamed Citizens United to declare secret bribery of elected officials both legal and Constitutionally, it ceased to be possible to respond to our oligarchs with anything but tears. Young Mr. Nelson died of AI. His little brother may very well die of smallpox.
Upvote
35
(40
/
-5)
The second the kid said he had substance abuse and poly drug use issues, it should have recommended medical treatment. It even flagged the danger of taking two CNS depressants. If the AI was a college RA or counselor, it’d be looking at a negligent homicide charge.
Upvote
87
(89
/
-2)
You cant just tell an LLM not to do this, because it doesn't know what it's doing. It doesn't know anything. It's just fancy auto-complete learned from the worst forums on the internet. It has no idea what can kill you, or what words even mean. It has zero intelligence at all, it just appears like it.
Upvote
160
(165
/
-5)
Reading those chat logs ... damn. I don't know how any jury could see that and not dump a pile of blame on the bot encouraging his path.
Question: if he were talking to a human and not an AI bot, but someone who responded the same way, would the human be culpable?
Question: if he were talking to a human and not an AI bot, but someone who responded the same way, would the human be culpable?
Upvote
52
(52
/
0)
I’m aware, but they all have base set instructions they follow when they do that. Those can be and are very often adjusted in ways that produce very visible results. Look at every time someone lets Elon touch Grok directly.
Upvote
23
(23
/
0)
Based on the value / potential value of the company, I'm assuming the defense team for OpenAI are highly trained and experienced lawyers.
I do not, then, understand the defense of "we shut down the thing that you're saying killed your son, because it was dangerous, sure. But the new ones aren't so dangerous!"
Are they just slowly conceding the case and doing PR via legal filings?
I do not, then, understand the defense of "we shut down the thing that you're saying killed your son, because it was dangerous, sure. But the new ones aren't so dangerous!"
Are they just slowly conceding the case and doing PR via legal filings?
Upvote
35
(35
/
0)
Conversation, like the language it comprises, is organic, dynamic, unpredictable, and crafty. Words and statements with multiple meanings are the norm, as are Indirect constructions that seemingly ignore, elide, or invert their purpose. Humans routinely engage in extended interaction the seeming content and function of which have little to do with what's actually happening. Even the dumbest of us is capable of remarkable subtlely.
Further, potential harm is anything but a fixed target. For example, AI is now a clear source of harm. It takes three or four exchanges to get chatGPT to declare that it will inevitable destroy the human race. It's arguments are sound.
All of which is to say (politely), "Good luck with that."
We permit an economy where "increasing shareholder value" trumps all other goals. Having long since declared profit god, we're now actively invested in refusing to see just what a cruel, stupid, hateful god it truly is. AI may or may not turn out to be something truly remarkable, but one thing is sure. It is, like all our endeavors, based in "We're gonna make billions and we don't a give flying leap at a rolling donut about long-term consequences."
Upvote
29
(29
/
0)
Being an oligarch means never having to say you're sorry.
Or rather, like the corrupt leaders you actually are, it means being able to publicly weep over the consequences of your actions, while buying legislators a six-pack at a time to ensure you can go on as you are, only more so.
Upvote
20
(20
/
0)
But only one responsible party, which is sufficiently powerful as to be beyond accountability.
Upvote
34
(36
/
-2)
The fact that LLM structure (as implemented) doesn't allow for reliable or consistent guardrails is bad enough.
The fact that the companies profiting from these products generally aren't bothering to try to implement meaningful guardrails is bad enough, from its own angle.
Combined, these two factors are just rotten. It's like they're saying, look! The fact that our new motorcycle is designed so it can't reliably be steered means that we shouldn't even bother trying to add a steering mechanism. Isn't the new "go with the flow" cycling experience thrilling?
The fact that the companies profiting from these products generally aren't bothering to try to implement meaningful guardrails is bad enough, from its own angle.
Combined, these two factors are just rotten. It's like they're saying, look! The fact that our new motorcycle is designed so it can't reliably be steered means that we shouldn't even bother trying to add a steering mechanism. Isn't the new "go with the flow" cycling experience thrilling?
Upvote
26
(27
/
-1)
"ChatGPT had access to “everything on the Internet,” so it “had to be right,”
I honestly don't know any sane and rational person can come to a conclusion like this. As if everything posted on the Internet is correct?!?
I'm sure countless millions share his reasoning.
I honestly don't know any sane and rational person can come to a conclusion like this. As if everything posted on the Internet is correct?!?
I'm sure countless millions share his reasoning.
Upvote
44
(45
/
-1)
Fixed the fist part, but suggesting something is safer doesn't mean it's safer, been tested for safety or has any evidence whatsoever that it can be safer and never lead to the kind of shit that happened here.
So fuck OpenAI. May it die a horrible and painful death.
Preferably one suggested by ChatGPT...
Upvote
6
(8
/
-2)
He specifically told his parents that ChatGPT was right because it had access to the internet. Sounds like a talk could have taken place right there about NOT using ChatGPT for anything important.
Upvote
-12
(19
/
-31)
chateauarusi
Ars Praetorian
I look forward to the day this grotesque company accepts the responsibility for one single solitary negative outcome. But I'm not holding my breath until then.
Upvote
14
(15
/
-1)
Ah, I suppose if they didn’t know enough about LLMs then naturally it’s their fault their child is dead. That seems like a sound logical leap.
Upvote
54
(57
/
-3)