Gwynne Michele
@thecurioushermit
Curious Dialogues: A Personal Learning Project
2 replies
2 recasts
4 reactions
Kristin Piljay🇨🇦
@kristinpiljay
This sounds like a really interesting project and I'm very curious about the notebook podcast development. 100 $DEGEN
2 replies
0 recast
1 reaction
Gwynne Michele
@thecurioushermit
The NotebookLM podcasts are FASCINATING. They glitch a bit on occasion, but not all that much. Sometimes, the pacing gets a bit weird. But overall, close enough to human it would fool a LOT of people - one of the reasons I'm super upfront about what's AI and what's human in my work. I don't think AI is bad, I love it, but I do think transparency is necessary.
1 reply
0 recast
1 reaction
Kristin Piljay🇨🇦
@kristinpiljay
Yeah, I listened to one of those once that someone posted a while back (can't remember who). Only thing was that one person was the "support" (the woman) and was just agreeing to what the first person said, rather than a more balanced conversation. That was the only thing that would "give it away." It was very good. Of course, there are real conversations like that too. LOL
1 reply
0 recast
1 reaction
Gwynne Michele
@thecurioushermit
Yeah, it learned that pattern from the way actual podcasts go - Google basically had every podcast ever to train off of, and those biases came through strong in the training. It's an issue across models.
1 reply
0 recast
1 reaction
Kristin Piljay🇨🇦
@kristinpiljay
Well, there is usually the host and then who they are interviewing or who te guest is (the main focus), but the host does have more to do then just support the other person. But I think that is subtle and so the AI's aren't sensitive enough to understand that (yet).
0 reply
0 recast
1 reaction
