NotebookLM is an extraordinarily fun experiment to...

NotebookLM is an extraordinarily fun experiment to use. I just got access earlier today and it's really something else. You dump a bunch of text sources into a “notebook” and that trains an LLM so you can interface with it about your specific sources (with citations).

I gave it a bunch of research papers I archived from my college days. As a chatbot, it’s whatever. As a search engine, it’s incredible. This is the best way to search a large corpus of papers.

A Notebook LM chat session in a split view layout. On the left, a plain text representation of a research paper PDF. On the right, I asked it “What are some methods that platforms use to make their products more addictive?”

It responded in part:
"Methods that firms use to make their platforms more addictive:
• Automatically rolling over into the next day. Some platforms, such as TikTok, automatically roll over into the next day without asking the user if they want to continue. This can make it difficult for users to stop using the platform, as they may feel like they need to finish what they were doing before they can stop.
• Push notifications. Platforms often send push notifications to users, even if they are not actively using the platform.
This can be a way to remind users of the platform and encourage them to log in.
• Personalization. Platforms use data about users to personalize their experience, which can make the platform more engaging and addictive. For example, a platform may show users content that they are interested in, or suggest friends that they may want to follow.
• Feedback loops...”

Notebook LM with the source panel open, showing four of the six PDFs I’ve loaded in. They all are “enabled,” meaning that chat context has access to that material. You can tick them off to exclude a specific source from conversation. I asked a question: “Do casinos and digital products share any DNA in their addictive properties?”

It responded in part:
"Yes, casinos and digital products share some DNA in their addictive properties.
• Both casinos and digital products use variable rewards to keep users engaged. Variable rewards are rewards that are not given on a fixed schedule, but rather are given randomly. This makes it difficult for users to predict when they will receive a reward, which keeps them engaged and coming back for more.
• Both casinos and digital products use social reinforcement to keep users engaged. Social reinforcement occurs when users interact with other people who are also using the product. This can create a sense of community and belonging, which can make it difficult for users to quit using the product.
• Both casinos and digital products use psychological triggers to keep users engaged. Psychological triggers are things that can trigger a user's desire to use the product. These triggers can include things like the sound of a slot machine, the smell of a casino, or the sight of a social media notification.”

10 available citations

Like 24 Nov 2023 at 2:39 | Open on hachyderm.io