Being interrupted is a nuisance. Apparently, even AI-generated podcast hosts agree.
Or as discovered by users of Google NotebookLM. NotebookLM was released last year and made waves for its ability to create fully AI-generated podcast-like discussions from user-uploaded content, discussed by a chatty AI bot that acts like a podcast host. . In December 2024, NotebookLM launched a new feature called “Interactive Mode.” This allows users to “call” into the podcast and ask questions, effectively interrupting the AI host’s conversation.
When this feature was first rolled out, AI hosts seemed annoyed by such interruptions. Josh Woodward, vice president of Google Labs, told TechCrunch that they sometimes make off-topic comments to human callers, like “you were getting close to that” or “that’s what I was going to say.” He explained that it felt “strangely hostile”.
So the NotebookLM team decided a “friendliness adjustment” was needed, and posted a self-deprecating joke about it on the product’s official X account.
Woodward said the team solved the problem by studying how members could more politely respond to interruptions.
“We tested different prompts, often studying how team members respond to interruptions, and came up with new prompts that we thought felt friendlier and more engaging.” he said.
It’s not entirely clear why this problem occurred in the first place. Human podcast hosts may exhibit irritation when interrupted, which may be recorded in the system’s training data. But a person familiar with the matter said the incident was most likely due to the system’s prompt design, not training data.
In any case, the fix seems to be working. When TechCrunch tried out the interactive mode, the AI host didn’t seem annoyed, but rather looked surprised and said, “Wow!” before politely asking the human to chime.
TechCrunch has a newsletter focused on AI. Sign up here to get it delivered to your inbox every Wednesday.