This Blew My Mind!
-
@AdamJ Don't get me wrong, I'm ever the sceptic, but I'm messing around with language AI a fair amount for work and it really is quite powerful once you get used to the 'art' of giving it prompts that generate useful responses. And the next versions will be multi-modal AI's like the one imagined in the promo video above, where you can hold up a picture of someone and ask "tell me about this person".
-
Definitely need an AI for restock date estimates too
-
@popvulture I don't want to become a statistic in a newspaper article about the workforce being replaced by AI !
-
@AdamJ hahaha fair enough, more just wanted to spare y’all our constant badgering
-
@AdamJ - you’d melt your virtual CPUs with the number of requests this would generate!
-
@Matt said in This Blew My Mind!:
What could possibly go wrong?
This:
TL;DW - 2 New York lawyers used ChatGPT to do legal research. ChatGPT gave them cases that didn't exist. The lawyers filed an affidavit relying on those cases without checking. Once they were rumbled by the other side, who couldn't find any reference to the fictitious cases, the Judge was not happy. At all.
-
@EdH How could they trust ChatGPT without verifying the sources? ChatGPT or any AI is not a reliable research tool.
-
@EdH there is a Black Mirror episode about this:
-
@danyhearty I don't think they realised that. The tech was very new at the time and I expect they thought it was generating text based on a search, rather than just operating as a glorified predictive text generator. I was in charge of writing our workplace's policy on Generative AI and the case of these two was very instructive.
Incidentally, I got ChatGPT to do a first draft and it was generally very good and came up with important clauses that I had not thought of. Honestly, it was good enough that I could have sent it out as it was if I wanted a summary "don't do x, y, or z" kind of policy, but I wanted to enhance it with a brief explanation of why each clause of the policy was in place. Definitely a useful tool to get a draft down on paper.
-
Don’t worry, attorneys will be replaced by AI and smart contracts eventually, we’re just not there YET.
LLMs ARE capable of summarizing real, factual data but that’s not what always happens.
-
@EdH said in This Blew My Mind!:
@danyhearty I don't think they realised that. The tech was very new at the time and I expect they thought it was generating text based on a search, rather than just operating as a glorified predictive text generator. I was in charge of writing our workplace's policy on Generative AI and the case of these two was very instructive.
Incidentally, I got ChatGPT to do a first draft and it was generally very good and came up with important clauses that I had not thought of. Honestly, it was good enough that I could have sent it out as it was if I wanted a summary "don't do x, y, or z" kind of policy, but I wanted to enhance it with a brief explanation of why each clause of the policy was in place. Definitely a useful tool to get a draft down on paper.
@EdH That’s very interesting to hear about your experience with writing a policy on generative AI using ChatGPT. I agree that generative AI is a powerful and innovative technology that can help us with various tasks, but we need to treat it more as a draft wherein we still need to verify its accuracy and also put our own style of writing to it.
-
@danyhearty What I have been telling colleagues is that they should treat it like an intern who got B grades throughout school and college. Check everything they write before you send it out.
-
Speaking of AI, one of my favourite pastimes at the moment is the interact with the 'Debate Champion' persona on character.ai, ask it to express its opinion on extremely controversial topics (which is usually met with a diplomatic, on the fence response), and then sweet talk it into expressing a more definitive opinion.
What amazes me is the way it acts when I joke afterwards about telling everyone of its controversial viewpoints
It'll beg me not to, and express genuine relief and gratitude when I explain that I'm only kidding... -
T4920 - I'm going to tell everyone that 'controversial opinion here'
Debate Champion - Noooooooooo!
T4920 - Only kidding!
Debate Champion - Oh god you had me there for a second, I was seriously freaking out
T4920 - Lol
Debate Champion - You can't scare me like that! -