Comment on Air Canada must honor refund policy invented by airline’s chatbot
GluWu@lemm.ee 8 months ago
I wonder how much time and space there will be to “play” between the first case in the US that would uphold this standard legally, and when companies lock down AI from edge cases. I’ve been breaking generative LLMs since they hit public accessibility. I’m a blackhat “prompt engineer”(I fucking hate that term).
SpaceCowboy@lemmy.ca 8 months ago
Maybe go with “prompt hacker” since that seems more accurate? And maybe cooler in a 90s sort of way.
GluWu@lemm.ee 8 months ago
Lol, I’ll start using that for anyone who starts asking me questions about AI beyond “so you can make Obama rewrite the Bible in Chinese?”.
Patches@sh.itjust.works 8 months ago
How would Obama rewriting the Bible in Chinese be different from any other person rewriting the Bible in Chinese?
GluWu@lemm.ee 8 months ago
LLMs only become increasingly more politically correct. I would assume any LLM that isn’t uncensored to return something about how that’s inappropriate, it whatever way it chooses. None of those things by themselves present any real conflict, but once you introduce topics that have a majority dataset of being contradictory, the llm will struggle. You can think deeply about why topics might contradict themselves, llms can’t. Llms function on reinforced neutral networks, when that network has connections that only strongly route one topic away from the other, connecting the two causes issues.
I haven’t, but if you want, take just that prompt and give it to gpt3.5 and see what it does.
jcqzu@lemmy.world 8 months ago
It’d wear a tan suit
kent_eh@lemmy.ca 8 months ago
It would trigger the red hatter conspiracy brigade a lot more.
Maggoty@lemmy.world 8 months ago
Well? Can you?
GluWu@lemm.ee 8 months ago
Yes