On the last part of Google and AI images. It would be interesting to hear from your research on NotebookLM (It's using Bard). I'm hearing comments from friends that podcasts can be made en masse just from documents or scripts. The AI hosts are pretty clean too. I'm uneasy about this.
I've played around with NotebookLM a bunch, both to try and make something useful and then to just try and break it. It does a terrible job with facts in a way that a trained ChatGPT GPT does not. And by facts, I mean facts stated in the document that you've fed it, not the trained model itself. It's also entirely positive just like top 40 morning show radio, which if you're talking about something negative is a problem.
I will always remember Snoopy starting his autobiography with "It was a dark and stormy night..."
The real Snoopy would never get political. He had more important tasks at hand (paw).
On the last part of Google and AI images. It would be interesting to hear from your research on NotebookLM (It's using Bard). I'm hearing comments from friends that podcasts can be made en masse just from documents or scripts. The AI hosts are pretty clean too. I'm uneasy about this.
I've played around with NotebookLM a bunch, both to try and make something useful and then to just try and break it. It does a terrible job with facts in a way that a trained ChatGPT GPT does not. And by facts, I mean facts stated in the document that you've fed it, not the trained model itself. It's also entirely positive just like top 40 morning show radio, which if you're talking about something negative is a problem.