Like nearly everybody, we had been impressed by the power of NotebookLM to generate podcasts: Two digital folks holding a dialogue. You may give it some hyperlinks, and it’ll generate a podcast based mostly on the hyperlinks. The podcasts had been attention-grabbing and interesting. However additionally they had some limitations.
The issue with NotebookLM is that, when you may give it a immediate, it largely does what it’s going to do. It generates a podcast with two voices—one male, one feminine—and provides you little management over the outcome. There’s an non-compulsory immediate to customise the dialog, however that single immediate doesn’t can help you do a lot. Particularly, you possibly can’t inform it which subjects to debate or in what order to debate them. You possibly can attempt, nevertheless it gained’t pay attention. It additionally isn’t conversational, which is one thing of a shock now that we’ve all gotten used to chatting with AIs. You possibly can’t inform it to iterate by saying “That was good, however please generate a brand new model altering these particulars” like you possibly can with ChatGPT or Gemini.
Study sooner. Dig deeper. See farther.
Can we do higher? Can we combine our data of books and know-how with AI’s potential to summarize? We’ve argued (and can proceed to argue) that merely studying use AI isn’t sufficient; you must discover ways to do one thing with AI that’s higher than what the AI might do by itself. You’ll want to combine synthetic intelligence with human intelligence. To see what that may seem like in observe, we constructed our personal toolchain that provides us rather more management over the outcomes. It’s a multistage pipeline:
- We use AI to generate a abstract for every chapter of a guide, ensuring that every one the essential subjects are coated.
- We use AI to assemble the chapter summaries right into a single abstract. This step basically offers us an prolonged define.
- We use AI to generate a two-person dialogue that turns into the podcast script.
- We edit the script by hand, once more ensuring that the summaries cowl the proper subjects in the proper order. That is additionally a chance to appropriate errors and hallucinations.
- We use Google’s speech-to-text multispeaker API (nonetheless in preview) to generate a abstract podcast with two contributors.
Why are we specializing in summaries? Summaries curiosity us for a number of causes. First, let’s face it: Having two nonexistent folks talk about one thing you wrote is fascinating—particularly since they sound genuinely and excited. Listening to the voices of nonexistent cyberpeople talk about your work makes you’re feeling such as you’re residing in a sci-fi fantasy. Extra virtually: Generative AI is definitely good at summarization. There are few errors and nearly no outright hallucinations. Lastly, our customers need summarization. On O’Reilly Solutions, our prospects often ask for summaries: summarize this guide, summarize this chapter. They need to discover the knowledge they want. They need to discover out whether or not they actually need to learn the guide—and if that’s the case, what components. A abstract helps them do this whereas saving time. It lets them uncover rapidly whether or not the guide will probably be useful, and does so higher than the again cowl copy or a blurb on Amazon.
With that in thoughts, we needed to suppose by means of what essentially the most helpful abstract could be for our members. Ought to there be a single speaker or two? When a single synthesized voice summarized the guide, my eyes (ears?) glazed over rapidly. It was a lot simpler to take heed to a podcast-style abstract the place the digital contributors had been excited and enthusiastic, like those on NotebookLM, than to a lecture. The give and take of a dialogue, even when simulated, gave the podcasts vitality {that a} single speaker didn’t have.
How lengthy ought to the abstract be? That’s an essential query. Sooner or later, the listener loses curiosity. We might feed a guide’s total textual content right into a speech synthesis mannequin and get an audio model—we might but do this; it’s a product some folks need. However on the entire, we count on summaries to be minutes lengthy quite than hours. I would pay attention for 10 minutes, perhaps 30 if it’s a subject or a speaker that I discover fascinating. However I’m notably impatient after I take heed to podcasts, and I don’t have a commute or different downtime for listening. Your preferences and your scenario could also be a lot completely different.
What precisely do listeners count on from these podcasts? Do customers count on to study, or do they solely need to discover out whether or not the guide has what they’re on the lookout for? That relies on the subject. I can’t see somebody studying Go from a abstract—perhaps extra to the purpose, I don’t see somebody who’s fluent in Go studying program with AI. Summaries are helpful for presenting the important thing concepts offered within the guide: For instance, the summaries of Cloud Native Go gave an excellent overview of how Go might be used to deal with the problems confronted by folks writing software program that runs within the cloud. However actually studying this materials requires taking a look at examples, writing code, and training—one thing that’s out of bounds in a medium that’s restricted to audio. I’ve heard AIs learn out supply code listings in Python; it’s terrible and ineffective. Studying is extra seemingly with a guide like Facilitating Software program Structure, which is extra about ideas and concepts than code. Somebody might come away from the dialogue with some helpful concepts and presumably put them into observe. However once more, the podcast abstract is barely an outline. To get all the worth and element, you want the guide. In a latest article, Ethan Mollick writes, “Asking for a abstract shouldn’t be the identical as studying for your self. Asking AI to unravel an issue for you shouldn’t be an efficient solution to study, even when it feels prefer it must be. To study one thing new, you will must do the studying and pondering your self.”
One other distinction between the NotebookLM podcasts and ours could also be extra essential. The podcasts we generated from our toolchain are all about six minutes lengthy. The podcasts generated by NotebookLM are within the 10- to 25-minute vary. The longer size might enable the NotebookLM podcasts to be extra detailed, however in actuality that’s not what occurs. Fairly than discussing the guide itself, NotebookLM tends to make use of the guide as a leaping off level for a broader dialogue. The O’Reilly-generated podcasts are extra directed. They comply with the guide’s construction as a result of we supplied a plan, an overview, for the AI to comply with. The digital podcasters nonetheless specific enthusiasm, nonetheless herald concepts from different sources, however they’re headed in a route. The longer NotebookLM podcasts, in distinction, can appear aimless, looping again round to choose up concepts they’ve already coated. To me, a minimum of, that appears like an essential level. Granted, utilizing the guide because the jumping-off level for a broader dialogue can also be helpful, and there’s a steadiness that must be maintained. You don’t need it to really feel such as you’re listening to the desk of contents. However you additionally don’t need it to really feel unfocused. And in order for you a dialogue of a guide, it’s best to get a dialogue of the guide.
None of those AI-generated podcasts are with out limitations. An AI-generated abstract isn’t good at detecting and reflecting on nuances within the authentic writing. With NotebookLM, that clearly wasn’t below our management. With our personal toolchain, we might definitely edit the script to replicate no matter we wished, however the voices themselves weren’t below our management and wouldn’t essentially comply with the textual content’s lead. (It’s controversial that reflecting the nuances of a 250-page guide in a six-minute podcast is a shedding proposition.) Bias—a form of implied nuance—is a much bigger subject. Our first experiments with NotebookLM tended to have the feminine voice asking the questions, with the male voice offering the solutions, although that appeared to enhance over time. Our toolchain gave us management, as a result of we supplied the script. We gained’t declare that we had been unbiased—no one ought to make claims like that—however a minimum of we managed how our digital folks offered themselves.
Our experiments are completed; it’s time to point out you what we created. We’ve taken 5 books, generated quick podcasts summarizing every with each NotebookLM and our toolchain, and posted each units on oreilly.com and in our studying platform. We’ll be including extra books in 2025. Take heed to them—see what works for you. And please tell us what you suppose!
