Hallucination Frustration

“I hate the hallucinations because it seems like there's no excuse for a lot of them, but it happens anyway.”

“The biggest disappointment would be like when it's I'll say confidently wrong. Like it thinks that it's right and then it starts telling you to do things or that these things are facts.”

“Right off the bat, the first time I used this like a month ago, it hallucinated a whole quote.”

“A lot of them now are making them look really cool and have vibe coding but I don't think they ever go back in and add anything just whatever they prompted and told our customers and there's parts where it says supposed to have secondary research and customer research and it's just making up pain points in there and so everything looks put together and there's a lot of words on a page but nobody's still going in for that second layer.”

“I guess I don't know if miserable failure is accurate but not far from that. Essentially figured out that what we were doing wasn't working and that it wasn't getting us where we wanted to be and the cost benefit was just not even close to being there.”

“not just relying on the transcripts alone, but introducing moderator notes to the analysis as well to help get that real world and actual findings.”

“it's generative, right? I mean, we should expect that it invents. But I think the majority of people don't think about that. I try to remind myself, it's generating stuff. So I try to reply, "Please stick to the real thing." And it even invented places to eat that don't exist,”

“even with a fake website. I went there and this restaurant doesn't exist.”

“I've also run into some situations with ChatGPT where it will just obviously hallucinate something. The chief example I always have of that is there was a time where, literally, it was last November, I needed to make a calendar for like a newsletter that would have been the month of December and I just didn't feel like making the Word table. 00:06:41 So I asked it, "Make a Word table that's a calendar for the month of December with two rows for each date," that sort of thing, and it messed the dates up. Like if November started on a Monday, it had it starting on a Tuesday where none of the dates lined up.”

“Yeah. I think at the end of the day, we've all seen some big examples, like that McKinsey report, where they charged like $100,000 and it turned out it was all AI slop. The ability for a human being to go through and discern whether something is correct or incorrect, and then be able to go back to the data set that is built on, when it comes to being able to trust your own evidence, your own sources, your own starting points, and being able to look at it and run your own analysis on it. You can tell an LLM, "Give me all the citations which you're drawing from," and it's going to give you a handful of those. It's not going to give you all of them. And even then, now I'm going to have to go through someone else's data and try to figure out whether what they actually said and whether the extrapolation was correct. That's a problem, because, now, what time is being saved here?”