Original video in Chinese.
Key Takeaway
- Google NotebookLM now supports generating Chinese podcasts, which is a huge help for studying and for independent creator businesses.
- By combining mind maps, AI podcasts, and AI conversations, NotebookLM provides a three-dimensional learning experience and improves knowledge absorption efficiency.
- Although AI podcasts still lack a “human” feel, their quality is already good enough, and they have a huge advantage in volume, so they may give rise to new content formats.
- The article predicts that AI podcasts may be integrated into platforms like WeChat Official Accounts, breaking down barriers to content creation.
- It emphasizes that the power of AI lies in breaking down barriers between tracks and product distinctions, and organizations should actively embrace AI.
NotebookLM finally supports generating Chinese podcasts! This will be extremely helpful for our Chinese readers’ learning, and even for independent creator businesses.
Just yesterday, Google announced that NotebookLM’s language summary feature now supports more than 50 languages, including Chinese. Like the people in my community, I had been eagerly waiting for this for a long time. Because it can turn the documents you upload into a podcast-style conversation. Let me show you the result.
This is a piece by the two big shots, David Silver and Richard Sutton, about how the evolution of AI will enter the “experience era.” All along, AI has been progressing by using human data and imitating human abilities — this is called the “experience era.” But this approach is starting to hit its ceiling.
First, there isn’t enough high-quality data anymore. Human data accumulated over so many years has basically already been fed in. Second, if AI uses human experience as learning material, it becomes very hard for AI to go beyond the boundaries of human understanding and evolve into a new intelligence that surpasses human intelligence.
So, AI is going to enter the “experience era” next. This paper is exactly about that content, and it’s excellent. I strongly recommend that everyone take a look.
This paper is 11 pages long in total. With the help of the Gemini model, NotebookLM generated an 8-minute podcast based on the content. I’ll play a short segment for everyone to hear.
I don’t know what everyone thinks, but I was quite pleasantly surprised.
From the perspective of Chinese speech, aside from a few places in the middle where you can hear a dubbed-film kind of tone, or rather a foreigner speaking Chinese — a bit too exaggerated — overall, it’s already very good.
In terms of content quality, I listened to the entire audio, and I had also read the paper before. I can say that its extraction of the key points is quite on target. The core takeaways are all there.
So now, when you import a document into NotebookLM, you can first look at the mind map it generates to get an understanding of the overall logic and the core conclusions.
Then you can listen to the AI podcast. You can listen carefully, or just listen casually. Anyway, after listening, the parts you’re interested in will automatically sink into your mind, and you’ll remember them.
Finally, you can talk to the AI. You can ask questions proactively, or use the suggested questions it provides — either way works. The AI has already thoroughly digested the content. You just let it “navigate,” and have it take you through the entire “logic road network.”
With this whole setup, in addition to text-based communication, you also get auditory input from the podcast and visual logical input from the mind map. So you can see that with NotebookLM, the input you get while learning will be extremely three-dimensional, and your brain will receive much more stimulation, which helps you absorb new things better.
Believe me, this will definitely work much better than just “reading books”!
Lastly, one more thing. I’m guessing people are definitely already trying to make “AI podcasts.” Before, NotebookLM didn’t support Chinese, so you had to look for alternatives. Now that it does, you just drag the document in, wait three or four minutes, and a podcast comes out. Then you download the audio file and that’s it.
Compared with traditional podcasts, AI podcasts really do lack a human feel. But they’ve already reached a 70-point level, the quality is already passable, and they’re even better than many podcasts made by people. At that point, their advantage in volume can be fully brought into play.
If I were at WeChat, I would definitely try something like this on Official Accounts. There’s already a “listen to the full text” feature, so why not take it one step further and add an “AI podcast”? Maybe in a few months we’ll see a gray-release test of AI podcasts on Official Accounts.
You see, this is the power of AI. The barriers between tracks and the distinctions between products are completely broken down. No organization can escape it; the only option is to actively embrace it.
OK, that’s it for this episode. If you want to learn about AI, want to become a super individual, and want to find like-minded people, come join our newtype community. See you next time!