NotebookLM: Another Google Experimental AI

In the 22-plus months since ChatGPT was released, I’ve shared a few observations and tips with friends and readers. One of the tips I frequently share is my morning routine of using AI for 20 to 30 minutes every day. Last week, I opened One Useful Thing, a Substack blog by Ethan Mollick, and read about NotebookLM, an AI tool new to me.

Professor Mollick wrote that he fed the text of an article to the app, which generated an audio podcast. He noted that the podcast was surprisingly accurate even though it built on some of his ideas.

Curiosity got the best of me, so I accessed the NotebookLM site and signed up to use it. Like so many of us, I skipped the tutorial and loaded the link to one of my recent blog articles, PhD Completion Rates: What Percentage of Students Finish Programs On-Time?

The app provided a summary of my article as well as a reprint of the article itself. I quickly found the Notebook Guide link and created an Audio Overview. I liked the podcast. I thought the conversation between a man and a woman was clever. I particularly liked the creativity and voice inflections.

I could only find one issue with the podcast. At the beginning, the AI chose to describe me with the pronoun “them.” My preferred pronouns are he/him/his, but I don’t publish those in my articles. Given the internet’s access to so much information, I was surprised that it couldn’t figure out which ones to use.

As I listened repeatedly to the podcast, I learned more about NotebookLM and its features.

What Is NotebookLM?

Google states that NotebookLM is a “personalized AI collaborator that helps you do your best thinking. After uploading your documents, NotebookLM becomes an instant expert in those sources so that you can read, take notes, and collaborate with them to refine and organize your ideas.”

Google emphasizes its definition of sources and how Notebook uses them. According to the website, “a source is a static copy of the source document you import or upload to the app. When using NotebookLM, the model will use the sources you upload to answer your questions or complete your requests.”

Furthermore, Google states, “NotebookLM can’t delete or edit your original files. When you upload a doc, the app makes a copy of the original file. NotebookLM doesn’t keep track of changes to the original doc automatically, including access to it, so you have to manually refresh any local file you uploaded.”

Another feature of NotebookLM is that the app creates a source overview that “summarizes the document and offers key topics and questions to ask.” I found the questions to be worthwhile considerations.

Keeping the app oriented toward the Google family, NotebookLM supports Google Docs, Google Slides, PDF, text, and markdown files, web URLs, copy-pasted text, YouTube URLs of public videos, and audio files. Each source can contain up to 500,000 words (the average word count for a non-fiction book is between 70,000 and 120,000 words).

If you want to import an article using a web URL (like I did for my blog), only the text content of the webpage will be used as a source. Images, embedded videos, and nested webpages are not imported, and paywall webpages are not supported.

If you import an audio file, the file is transcribed, and the text is saved as a new source. Google provides an extensive list of languages supported for audio file transcription—I counted approximately 70.

Features added since December 8, 2023

The following features have been added since the original release.

  • Notebooks can include up to 20 sources.
  • Sources can include up to 200,000 words.
  • Adding a note will create an independent new note instead of adding to a single notepad.
  • All written and saved notes are pinned to the Noteboard.
  • Chat responses can be pinned to the Noteboard as notes.
  • Saved chat responses will retain the original citations.
  • Clicking on a citation in a chat or saved note will take you to the quoted source.
  • You can hide the source if you want to concentrate on notetaking.
  • You can focus the AI on selected sources.
  • Source guides now appear at the top of each source in the source sidebar.
  • Notebooks can be shared in both viewer and editor modes.
  • Viewer shares your assembled sources.
  • The Editor lets you collaborate on creating notes.
  • NotebookLM automatically suggests follow-up questions for you based on your recent conversation history and the content of your sources.
  • PDFs can be uploaded as sources up to 100MB or the word limit.
  • You can copy and paste text to create a new source.
  • NotebookLM has a palette of suggested actions, including:
    • Combining all notes into a single note.
    • Providing constructive feedback on your prose or argument.
    • Creates a concise, easy-to-read overview of the content of multiple notes.
    • Convert your selected notes into an outline that is organized around topics.
    • Creates an on-the-fly study guide based on your notes, including key questions and a glossary.
    • Prompts you with related ideas from your sources based on the content of the selected notes.

Additional Podcasts Created

When I find an AI tool like this, I normally don’t walk away from it. I uploaded another recent blog article, Graduate Degrees: Risky and Unequal Paths to the Top. I created another audio podcast. You must first allow public access to the audio file to share links to these audio podcasts. Then, you can share the link.

I liked the second audio podcast as much as I liked the first. This time, there was no mistake related to pronouns. The summary of my article was accurate as well, explaining that I recommended something different than the researchers at Georgetown’s CEW, including the elimination of the Grad PLUS loan program.

For this submission, I asked NotebookLM to create a briefing document (one of the create options). I was impressed with the briefing document. I thought it accurately summarized my recommendations. I also enjoyed seeing specific quotes included to back specific points in the outline. The briefing document is appended below.

Briefing Doc: Graduate Degrees – Costs, Outcomes, and Policy Recommendations

Source: Excerpts from “Graduate Degrees: High Costs, Unequal Paths” by Dr. Boston, published October 4, 2024.

Main Themes:

  • Rising Costs and Diminishing Returns: The cost of graduate degrees has tripled since 2000, while earnings premiums haven’t significantly increased. This raises concerns about the financial burden on graduates and the value proposition of these programs.
  • GradPLUS Loan Program: The program enables increased borrowing and potentially fuels tuition inflation, disproportionately impacting lower-income students.
  • Accountability and Transparency: A lack of robust data and regulations allows low-performing programs to persist. Greater transparency in program outcomes and stricter accountability measures are crucial.
  • Systemic Inequities: Racial, ethnic, and gender disparities exist in degree attainment, field of study concentration, and earnings, highlighting the need for targeted support and interventions.

Key Ideas and Facts:

  • Graduate degrees offer the highest earnings prospects, but growing costs diminish returns. “Graduate degrees – which include master’s, professional, and doctoral degrees – offer the highest earnings prospects and best chance of finding employment of all educational credentials. However, the growing costs of attaining a graduate degree have diminished the returns that accrue to graduate degree completers.”
  • GradPLUS loans allow borrowing up to the program cost, potentially incentivizing institutions to inflate tuition. “thanks to the high level of borrowings allowed under the GradPLUS program, colleges have limited incentive to keep costs in line with expected earnings.”
  • The proposed In-field earnings premium test requires graduates to earn only 5% more than bachelor’s degree holders, which might be too low a bar. “Allowing graduates to only improve their earnings by 5% over the median earnings of bachelor’s degree graduates is a very low standard. I assume grad students have higher expectations.”
  • The proposed debt-to-earnings test would disqualify programs where median loan payments exceed 10% of discretionary income. However, data limitations hinder comprehensive evaluation. “41% of master’s programs and 67% of professional programs do not pass the debt-to-earnings test.”
  • Systemic inequities permeate graduate education. Minorities and women face barriers in access, affordability, and earnings outcomes. “Women are credited with driving the growth in graduate education, with their share of graduate degrees increasing from 35% in 1970 to 61% in 2021. … Earnings inequities are greatest among women who hold graduate degrees.”
  • Targeted grant aid for socially valuable but lower-paying professions is suggested. “federal and state governments should provide targeted grant aid to support graduate education in socially valuable professions.”
  • Expanded transparency in program outcomes is recommended, including data on completion rates, loan repayment, and post-graduation licensure requirements. “Potential students would receive information about: …The program completion rates, The program withdrawal rates, The loan repayment rates among borrowers, The share of students who have loans…”

Policy Recommendations:

  • Eliminate or Reform GradPLUS: Dr. Boston advocates for the elimination of the GradPLUS program to curb tuition inflation. Alternatively, stricter limits and accountability measures could be implemented.
  • Apply Gainful Employment Rules to Graduate Programs: Extending existing Gainful Employment regulations to all graduate programs, regardless of institution type, would ensure greater accountability and protect students from predatory programs.
  • Targeted Grant Aid: Provide grants for students pursuing socially valuable but lower-paying professions to address workforce needs and reduce financial barriers.
  • Enhanced Transparency: Mandate comprehensive data reporting on program outcomes, including earnings, debt, and employment data, to empower students to make informed decisions.
  • Address Systemic Inequities: Implement targeted interventions and support systems to address racial, ethnic, and gender disparities in graduate education access, affordability, and outcomes.

Conclusion:

The report highlights critical issues within graduate education, emphasizing the need for policy reforms to protect students, ensure program quality, and promote equitable outcomes. While Dr. Boston agrees with the need for greater transparency, he disagrees with the researchers’ hesitancy to eliminate GradPLUS and believes that applying Gainful Employment rules would be a more effective approach.

A Few More Thoughts

A friend of mine authored a nonfiction book about the Calvert family responsible for founding the colony of Maryland. He shared a PDF of his near-final draft with me for my review. The book’s word count is approximately 88,000. I uploaded the PDF, created an audio podcast, and listened to it.

NotebookLM created an audio podcast for the 88,000-word book in approximately the same time that it took to create an audio podcast for each of my two articles. I liked the podcast so much that I sent it to my friend. He liked it as well. In fact, he liked it so much that he uploaded to NotebookLM a dozen or so short non-fiction stories that he had written.

Even though I initially used NotebookLM to create an audio podcast, I realized it was much bigger than that. The audio podcast is just another tool in its quiver. In fact, if I were currently a doctoral student, I would consider using NotebookLM to curate research papers, articles, and other documents that I was using for assignments.

While NotebookLM’s capabilities are not endless, they are clearly beyond most of the tools that I use for my writing. I already have an idea for a future blog article in which I intend to maximize my use of NotebookLM’s features. But I can see why some writers would find this product helpful for audio podcasts alone.

Subjects of Interest

Artificial Intelligence/AI

EdTech

Higher Education

Independent Schools

K-12

Science

Student Persistence

Workforce