Generative AI: Black Mirror is here

[BurnedOut]

So many of my friends who are not as well fluent in English have been doing something that is nearly unthinkable for me to do - talk with AI and do AI shopping to help them in their work at their office and uni both. By AI shopping I mean having several AI apps ready to ask. Another guy I talked to writes his AI prompts like he is talking to a human being. Another guy spends everyday at office chitchatting with chatgpt. They are actually having a banter at this time. Another guy told me that he 'thinks' that talking to AI like you are talking to a human provides a greater possibility of a better response. However I speculate that the way the NLP algorithms are written, it makes their job harder and not easier

 

[scorpiomover]

My brother uses AI at work all the time. He says he has learned to regularly tell the AI to "not lie to him", as he often has to catch the AI out, and tell it when it's not saying the truth.

 

[Puffy]

I don’t rely on AI for my job but I do experiment with it to see in what areas it can make me more efficient. Recently I had to do desk research and I used AI to help me identify 50 relevant case studies. I of course had to read into them and analyse them myself but that saved a lot of time and only took a couple of prompts.

 

[Hadoblado]

I was doing that too for awhile, but tbh it doesn't really work. If you catch it in a lie, most of the time it doesn't have the overall understanding to correct the lie or it wouldn't have lied in the first place.
What I've found is that I can layer my questions to prevent the lying being integrated into the corrective model. So instead of "find me a source for X claim, please don't lie", I can ask one AI for a list of sources, and use a second AI to sort those sources into valid or invalid. Because the second AI never lied in the first place, they haven't internalised the bad logic.
But yes, immensely frustrating to know that the AI is making something up, have it apologise and tell you it will correct it, then do the same damn thing again as if nothing ever happened.

 

[Cognisant]

Another guy told me that he 'thinks' that talking to AI like you are talking to a human provides a greater possibility of a better response.

Of course, it's trained on humans responding to each other and is designed to imitate how they do that, so when you ask it a question it will respond at your apparent level.

- YouTube

Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.

youtube.com