## Intro [(00:00:01)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1s) - [[NotebookLM | Notebook LM]] is a tool provided by [[Google]] Research that allows users to index and make their handwritten notes fully searchable, and also interact with the content in an interesting way, getting the most out of their research and notes together with the content provided by the large language model [(00:00:21)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=21s). - The main page of Notebook LM has a section with example notebooks that introduce users to the particular usage of the tool, and another section for users' own notebooks, where they can create and manage their own notebooks, such as the "History of AI" notebook [(00:00:47)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=47s). - Notebook LM can work with handwritten notes from any device, including the reMarkable paper Pro, iPads, Surface tablets, or even photographs of handwritten notes in a conventional paper notebook, making it a versatile tool for capturing and working with handwritten content [(00:01:21)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=81s). - The tool allows users to create new notebooks and add particular data to them, with the first screen prompting users to add data when creating a new notebook [(00:01:57)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=117s). - Notebook LM's functionality is not limited to a specific device or platform, as it can work with handwritten content from various sources, such as OneNote or other note-taking apps, making it a useful tool for users who capture handwritten text in different ways [(00:01:46)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=106s). ## Import files and documents [(00:02:01)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=121s) - The underlying large language model is powered by [[Google]] Gemini, a trained model on a variety of internet sources, allowing users to add their own documents to enhance their experience and interact with their knowledge base in the notebook [(00:02:01)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=121s). - Users have the option to work with Google Docs and slides, website links, [[YouTube]] videos, and can paste any arbitrary text from other sources, such as websites with hard-to-digest content, directly into the document [(00:02:34)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=154s). - There is a source limit of 50 documents or entities, which is notable given the typical context size limits of large language models, and the tool is currently free and in an experimental state, created by Google Labs [(00:03:08)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=188s). - The tool's future is uncertain, as it may change frequently and could potentially be incorporated into Google Keep, another separate product, so users should be aware of the risks of relying on it as their only note-taking tool [(00:03:27)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=207s). - Users can import files, such as PDFs, into the notebook, and immediately receive a notebook guide with suggested questions, allowing them to interact with their document by asking questions, similar to interacting with a chatbot [(00:04:06)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=246s). ## Get an instant summary [(00:04:32)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=272s) - The document is made visible on the left side, allowing scrolling through it to view a summary of the content, which includes social media and platforms planned for future business use, along with comments on conversion rates and engagement [(00:04:32)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=272s). - The summary highlights [[YouTube]] as having the highest conversion rate, but notes that it does not allow direct message interaction with users, while [[Instagram]] has a longer engagement period compared to other platforms like Tik Tok and Snapchat [(00:04:40)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=280s). - Other social media platforms mentioned include [[Twitter]], which is suitable for posting small updates, Facebook, [[LinkedIn]], Pinterest, and Amazon Kindle Direct Publishing, with LinkedIn being the main business platform and Pinterest used for visual content [(00:05:27)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=327s). - The document is correctly captured, allowing interaction and questioning, with features such as finding specific keywords, similar to Apple Notes, but with more advanced options due to the integration of the [[Google]] Gemini large language model [(00:06:08)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=368s). - The platform provides a full-fledged chat interface, enabling users to search for particular keywords and interact with their notes in a more advanced way, thanks to the background support of the Google Gemini large language model [(00:06:36)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=396s). ## Search your handwritten content [(00:06:52)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=412s) - The search function allows users to find specific keywords, such as "[[Instagram]]", in their handwritten content, and it returns relevant results quickly, including references to the source documents, but does not highlight the exact position in the document [(00:06:52)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=412s). - When searching for a keyword like "Amazon", the system can identify the exact page where the keyword is mentioned, but it does not provide highlighting of the keyword, which can be a disadvantage, although it still provides useful results [(00:07:53)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=473s). - The search function has limitations, but the integration of AI technology offers more advanced options, enabling users to ask questions and interact with their documents in a more sophisticated way, going beyond basic searching [(00:08:30)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=510s). ## Summarize the document [(00:08:40)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=520s) - A request is made to summarize all mentioned social media platforms and provide bullet points to consider, with the instruction being acknowledged as vague [(00:08:40)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=520s) - The request involves creating a bullet-pointed list in response to the instruction [(00:08:48)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=528s) - The instruction to summarize social media platforms and provide bullet points is noted, with the expectation of a list being generated [(00:09:00)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=540s) ## You get an "enriched" search result (with knowledge from the LLM) [(00:09:09)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=549s) - The search results are "enriched" with knowledge from the large language model, providing additional information not present in the original document, such as the mention of [[Twitter]], which was not in the original source [(00:09:25)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=565s). - The search results include links to respective pages, allowing for easy navigation, and direct messages are also possible, making the search function more interactive and useful [(00:09:47)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=587s). - The search function can also summarize notes, providing a concise overview of the content, and can scroll to specific pages where relevant information is mentioned, such as [[LinkedIn]] being mentioned on the second page of notes [(00:10:07)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=607s). - The search results may include information from various sources, such as [[YouTube]], [[Instagram]], LinkedIn, Pinterest, and Amazon, and can provide a brief summary of the notes, without transcribing them entirely [(00:10:14)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=614s). - The large language model can add new information to the search results, such as the conversion rates of different social media platforms, including YouTube having the highest conversion rate and Instagram having the second highest [(00:09:09)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=549s). ## Asking for "transcription" gives you a perfectly usable transcription of your entire notes [(00:10:31)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=631s) - The transcription process is able to understand and accurately convert handwritten notes into a usable text format, even when the handwriting is not perfect, as demonstrated by the example where the transcription correctly identifies words and phrases such as "image one YouTube highest conversion" and "LinkedIn business platform" [(00:10:31)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=631s). - The large language model is able to recognize and correct typos and minor errors in the handwritten text, resulting in a highly accurate transcription, including the recognition of specific details like the arrow symbol and the phrase "Max two per day" [(00:11:00)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=660s). - The transcription is also able to identify and extract relevant information from the handwritten notes, including key phrases like "High conversion rate" and "CTA", and even recognizes the context of the text, such as the mention of social media platforms like Facebook, [[Twitter]], and [[Instagram]] [(00:11:19)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=679s). - The transcription process is able to handle multiple pages of handwritten notes, processing each page as an image and extracting the relevant information, including text and symbols, as seen in the example where it correctly transcribes the text on page two, including the phrase "no links" [(00:11:53)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=713s). - The resulting transcription is highly accurate and usable, even when the original handwriting is not clear or well-written, as demonstrated by the example where the transcription correctly identifies phrases like "[[LinkedIn]] shorts" and "recruiting options" [(00:12:25)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=745s). - The transcription is able to capture the overall meaning and content of the handwritten notes, including flowcharts and processes, and is able to recognize and extract relevant information, such as the mention of "comment flowcharts and processes" and "simple to create" [(00:12:52)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=772s). - The final transcription result is a highly accurate and usable text format that can be used for further processing or analysis, as demonstrated by the example where the transcription correctly identifies the content and meaning of the handwritten notes [(00:13:08)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=788s). ## Save transcript as a note [(00:13:10)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=790s) - The option to save a transcript as a note is available, allowing users to refer to it at a later stage [(00:13:10)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=790s). - Once saved, the transcript can be used as a note, enabling users to access and review the information as needed [(00:13:10)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=790s). - The ability to create notes from transcripts provides a useful feature for organization and reference purposes, with the possibility of creating multiple notes [(00:13:21)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=801s). ## More generic question about the handwritten notes [(00:13:25)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=805s) - The discussion revolves around handwritten notes and how to utilize them effectively, with a focus on marketing strategies, such as using [[Instagram]], and the importance of posting strategically to avoid penalties on the platform [(00:13:25)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=805s). - To effectively market on Instagram, it is recommended to post a maximum of two times per day and to foster conversations, while selling privately through methods like private messages [(00:13:49)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=829s). - The ability to save notes and access further details, such as understanding, critiquing, suggesting related ideas, and creating an outline, is also explored, allowing for a deeper dive into the content [(00:14:04)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=844s). - When multiple documents are highlighted, the option to create a full outline, study guide, or combine them into one note becomes available, providing a fast and efficient way to work with documents and content [(00:14:20)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=860s). - The process of combining notes is demonstrated, resulting in a merged note that includes the transcript and marketing strategy, showcasing a convenient method for organizing and working with handwritten notes [(00:14:37)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=877s). ## The full power of AI at your fingertips [(00:14:55)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=895s) - The large language model is able to propose related ideas and provide suggestions based on the input, allowing users to explore different concepts and ideas [(00:14:55)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=895s). - The model can be used to compare different platforms, such as Instagram's conversion rate, and leverage X, formerly [[Twitter]], to post short messages [(00:15:05)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=905s). - These short messages can then be combined into a larger bullet-pointed list, which can later be converted into a longer story or blog post [(00:15:24)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=924s). ## AI takes handwritten notes a step further [(00:15:30)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=930s) - The process of using Apple Notes to full-text index any handwritten document from the reMarkable tablet has already been explored, and a video on this topic is available [(00:15:30)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=930s). - A note-taking approach has been developed to write scientific documents based on mathematical structures and equations, and a video is available that shows how to translate these handwritten mathematical equations into LaTeX code and flow diagrams into Mermaid code using the CLAE LM from Anthropic [(00:15:46)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=946s). - The handwritten mathematical equations and flow diagrams can be translated into LaTeX and Mermaid code, respectively, which can be used for proper typesetting and creating professional publications [(00:16:05)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=965s). - The ability to not only transcribe handwritten notes into proper text but also interact with the document has opened up a range of options, including enriching content with data from large language models trained on internet data [(00:16:23)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=983s). - Additional data sources, such as [[YouTube]] videos, images, links to web pages, and custom text, can be added to notebooks in [[NotebookLM]], allowing for further enhancement of the content [(00:16:50)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1010s). - The interaction with the document enables users to take handwritten notes and add various types of content, making it a powerful tool for creating and editing documents [(00:16:59)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1019s). ## Organization as topic-based research [(00:17:07)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1027s) - NotebookLM can be used to create topic-based research applications that allow for full interaction with specific topics and handwritten notes, combining to form a powerful tool [(00:17:07)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1027s). - Handwriting has psychological advantages that can improve personal productivity by enhancing the ability to digest, interpret, and learn information [(00:17:29)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1049s). - There is a video available that provides more information on how handwriting can improve productivity, and it is recommended for those interested in learning more about this topic [(00:17:22)](https://www.youtube.com/watch?v=w9jTSeVAdHo&t=1042s). ## Sources - [website](https://www.youtube.com/watch?v=w9jTSeVAdHo?si=zKvVTRnaTnmYv8t_)