Mitarbeiter von Open AI warnten vor der Tragödie vor dem Schützen von Tumbler Ridge: WSJ-Bericht
http://thestar.com/news/canada/open-ai-employees-warned-about-tumbler-ridge-shooter-before-tragedy-wsj-report/article_1d914e18-cfde-4d5e-ae56-690832c47927.html
2 Kommentare
>This was flagged by an internal system and some employees at the company urged higher-ups to contact the authorities.
[..]
The Journal says that according to an OpenAI spokesperson, Van Rootselaar’s account was banned but her content did not meet the threshold for alerting authorities.
That’s just negligence.
It was disturbing enough for the employees entrusted to safeguard the systems to request that it be escalated to the authorities, that in itself should be enough to warrant contacting the authorities. There should be no humming and hawing about it, it must be part of the internal processes to ensure that things like this don’t go overlooked.
Platforms see disturbing content all the time… People describe violent fantasies, write fiction, vent emotionally, role-play, etc.
If companies reported every scary conversation:
* police would get flooded
* innocent users could be reported constantly
So companies use a higher bar: credible + **imminent threat**.
This isn’t a case of someone unknown to police either.
Weapons WERE seized previously
RCMP confirmed:
* during a previous visit (within the past two years), firearms were seized under the Criminal Code
* however, the lawful owner petitioned to have them returned, and they were returned.
**A relative told reporters those guns were returned about a month before the attack**
It doesn’t matter if OpenAI had reported the chats which would have led to nothing being done. The guns would have been returned to the lawful owner (the mother) no matter what.
Do people actually think it’s legal in Canada to arrest or imprison someone because of what they wrote to an AI chat bot???