Envision you’re a podcaster who routinely does fast 10-to 12-minute summary evaluations of composed works. Now picture your manufacturer offers you numerous pages of absolutely nothing however the words “poop” and “fart” duplicated over and over once again and asks you to have an episode about the file on their desk within the hour.

Promoting myself, I ‘d have difficulty even understanding where to begin with such a job. When Reddit user sorryaboutyourcats provided the exact same timely to Google’s NotebookLM AI design, the outcome was a remarkably sound and interesting AI-generated podcast that touches on the nature of art, the viewpoint of attention, and the human desire to make implying out of the naturally worthless.

Examining Poop & & Fart composed 1,000 times – Creating significance from the worthless
byu/sorryaboutyourcats in notebooklm

When I asked NotebookLM to evaluate my Minesweeper book recently, commenter Defenstrar wisely asked “what would happen if you fed it a less engrossing or well written body of text.” The response, as seen here, reveals the intriguing instructions a modern-day AI design can enter when you let it simply spin its wheels and stray from a basically unmoored beginning point.

“Sometimes a poop is simply a poop …”

While Google’s NotebookLM released over a year earlier, the design’s just recently released “Audio Overview” function has actually been getting a great deal of attention for what Google calls “a new way to turn your documents into engaging audio discussions.” At its heart is a LLM comparable to the kind that powers ChatGPT, which develops a podcast-like script for 2 persuading text-to-speech designs to check out, finish with “ums,” disruptions, and significant stops briefly.

Experimenters have actually handled to deceive these AI-powered “hosts” into what seem like an existential crisis by informing them that they aren’t truly human. And detectives have actually handled to get NotebookLM to discuss its own system triggers, which appear to concentrate on “going beyond surface-level information” to discover “golden nuggets of knowledge” from the source product.

The “poop-fart” file (as I’ll be calling it for simpleness) is a quite fascinating test case for this sort of system. What “golden nuggets of knowledge” might be buried beyond the “surface level” of 2 scatological words duplicated for several pages? How do you “highlight intriguing points with enthusiasm”– as the uncovered NotebookLM timely recommends– when the file’s only oft-repeated points are “poop” and “fart”

Here, NotebookLM handles to utilize that total absence of context as its own beginning point for an intriguing stream-of-consciousness, podcast-like discussion. After some throat-clearing about how the audience has “outdone itself” with “a unique piece of source material,” the ersatz podcast hosts fast to compare the repeating in the file to Andy Warhol’s soup cans or “minimalist music” that can be “surprisingly powerful.” Later on, the hosts attempt to obtain some significance by comparing the file to “a modern dadaist prank” (noticable as “daday-ist” by the speech synthesizer) or the vase/faces visual fallacy.

Creative contrasts aside, NotebookLM’s virtual hosts likewise dig a bit into the psychology behind the “very human impulse” to “search for a pattern” in this “accidental Rorschach test” and our propensity to “try to impose order” on the “information overload” of the world around us. In nearly the very same breath, however, the hosts get philosophical about “confront[ing] the absurdity in trying to find meaning in everything” and recommend that “sometimes a poop is just a poop and a fart is just a fart.”