Friendly reminder that computers are for computing, not advice.
Gemini AI tells the user to die
91–100 of 150 posts
Re: Gemini AI tells the user to die
#92Probably something from the training data? There must be all sorts of edgy conversations in there. A crossed wire.
Re: Gemini AI tells the user to die
#93Not to take away from the main story, but the student was clearly cheating on their homework with Gemini, directly pasting all of the questions.
Re: Gemini AI tells the user to die
#94At least it's not a sign for the Judgement Day crowd. A terminator wouldn't say please.
Re: Gemini AI tells the user to die
#95Can't we just teach them robotics laws?
Re: Gemini AI tells the user to die
#961. Get huge amounts of raw, unfiltered, unmoderated data to feed model
2. Apologize, claim there's no way they could have possibly obtained better, moderated, or filtered data despite having all the money in the world.
3. Profit... while telling people to kill themselves...
I get that a small team, universities, etc. might not be able to obtain moderated data sets.. but companies making billions on top of billions should be able to hire a dozen or two people to help filter this data set.
This reads a lot like an internet troll comment, and I'm sure an AI trained on such would flag this that way... which could then be filtered out of the training model. You could probably hire a grad student to make this filter for this kind of content before ingestion.
Re: Gemini AI tells the user to die
#97Probably something from the training data? There must be all sorts of edgy conversations in there. A crossed wire.
They're all crossed wires.
Re: Gemini AI tells the user to die
#98Not to take away from the main story, but the student was clearly cheating on their homework with Gemini, directly pasting all of the questions.
I find this absolutely hilarious. If you’re a gerontology lecturer in Michigan, look out!
Re: Gemini AI tells the user to die
#99Not to take away from the main story, but the student was clearly cheating on their homework with Gemini, directly pasting all of the questions.
The link seems to be the obvious cut and paste cheating coupled with how many forums respond negatively to overt cheating with a dose of perspective from the llm as the responder to said flagrant cheating
Re: Gemini AI tells the user to die
#100Can't we just teach them robotics laws?
With LLMs, that's honestly probably not a good idea. They're already well aware of the concepts, invoking them specifically places the whole conversation closer to the conceptual space of a sci fi novel, where those rules always go wrong for any of a thousand tropey reasons. It'd probably make these instances more likely, not less.