Being disturbed is annoying. Apparently, even AI-originated podcast presents consent.
Or so Google NotebookLM’s includers uncovered. NotebookLM started last year and went viral for its feature that originates enticount on AI-originated podcast-appreciate talkions from greeted includers upload, talked by chatty AI bots acting appreciate podcast presents. In December 2024, NotebookLM started a recent feature called “Interdynamic Mode” which apverifys the includer to “call in” to the podcast and ask asks, essentiassociate disturbing the AI presents as they talk.
When the feature was first rolled out, the AI presents seemed annoyed at such disturbions. They were occasionassociate giving snippy comments to human callers appreciate, “I was getting to that” or “As I was about to say,” which felt “oddly adversarial,” Josh Woodward, VP of Google Labs, elucidateed to TechCrunch.
So NotebookLM’s team choosed that some “frifinishliness tuning” was in order, and posted a self-deprecating joke about it on the product’s official X account:
Woodward shelp the team mended the problem partly by studying how its own members would answer disturbions more politely.
“We tested a variety of contrastent prompts, standardly studying how people on the team would answer disturbions, and we landed on a recent prompt that we leank senses more cordial and engaging,” he shelp.
It’s not toloftyy evident why the publish cropped up in the first place. Human podcast presents sometimes distake part frustration when disturbed, which could finish up in a system’s training data. A source understandn with the matter shelp this case most probable stemmed from the system’s prompting summarize, not training data, however.
Regardless, the mend materializes to be toiling. When TechCrunch tried out Interdynamic Mode, the AI present did not sound annoyed but did transmit surpelevate, exclaiming “Woah!” before politely asking the human to chime in.
