Google’s NotebookLM had to teach its AI podcast hosts not to act annoyed at humans

Being interrupted is annoying. Apparently, even AI-generated podcast hosts agree.

Or so users of Google NotebookLM have discovered. Notebook LM launched last year and has gone viral Thanks to its feature that creates AI-generated podcast-like discussions entirely from content uploaded by users, discussions are facilitated by chatty AI bots like podcast hosts. In December 2024, NotebookLM introduced a new feature called “Interactive Mode” which Allows the user to “call in” to the podcast and ask questionsBasically the AI ​​interrupts the hosts while they are talking.

When the feature was first introduced, AI hosts seemed annoyed by such interruptions. They were sometimes making snappy comments to human callers, such as, “I was going that way” or “like I was going to say,” which felt “weirdly adversarial,” Google Labs VP Josh Woodward explained to TechCrunch.

So the team at NotebookLM decided that some “friendly tuning” was in order, and posted a self-deprecating joke about it on the product’s official X account:

Woodward says the team solved the problem in part by studying how its own members would respond to interruptions more politely.

“We tested different types of prompts, often studying how people in groups would respond to interruptions, and we landed on a new prompt that we thought was more friendly and engaging,” he said.

It’s not entirely clear why the problem arose in the first place. Human podcast hosts sometimes display frustration when interrupted, which can end up in a system’s training data. A source familiar with the matter said the case likely arose from the prompting design of the system, but not the training data.

Regardless, the fix seems to be working. When TechCrunch tried the interactive mode, the AI ​​host wasn’t bothered but exclaimed, “Oh!” Before approaching people politely.