We has podcast

It is easy to produce if you write - take your content and feed it to NotebookLM (or similar), and you'll have your own podcast!

Nearly a year ago, when Manu and I were having early discussions on what has now become Babbage Insight, one of the things we spoke about was about who does what. Around the time we incorporated, we even drew up a formal document (a Google Sheet) that tagged different responsibilities to both of us. The idea here is that in whatever is my “area of responsibility”, I can veto Manu. And vice versa.

In any case, one of the things I remember telling him then was that “I’m happy to delegate everything. However the only thing that I’m going to control is corporate communications”. Having been a blogger for 20 years, someone else controlling or censoring what I want to put out is something I’m not going to be particularly comfortable with.

Now, nearly a year later, I’m not so sure. Maybe the first rule of running an AI company is that you need to use AI yourself. And so when we figured out about Google’s NotebookLM, we had to try it out.

First I fed it our homepage, and asked it to generate a podcast, and the output wasn’t bad at all. Listen to this and then read on.

It’s actually great. I was having some doubts about our own website, whether we’re getting the right message across. On the evidence of what is there in the podcast, our message is spot on - at least the way the AI has interpreted it.

Now, you human (I assume you’re one), can you please look at our homepage and tell us if what the AI has gotten is right or not?

In any case, we weren’t done. As you might know (I really don’t know if anyone actually reads this newsletter), we recently announced that we’re building out Babbage as a service. We have a few ongoing conversations as we speak. There were two blogposts I wrote in this regard - one on what we are offering, and the other describing who the ideal customer will be.

You know what I did - I fed both these posts together to NotebookLM and asked it for a podcast. Again the results are impressive:

The AI has got our messaging fairly accurately, I would think (if you are confused, please let me know!).

A few random pertinent observations to round off:

• Please subscribe to our YouTube channel. Soon we’ll figure out a way to add faces to this as well (if you know an AI that offers that, please let us know)

• One thing I’ve heard is about cultural differences - that Indians (and Brits) are conservative about tomtomming their own accomplishments and abilities, and that Americans tend to be far more aggressive. AI is a great workaround for this - it can help you praise yourself much more than you would yourself! It can help you make your tone “more neutral”.

• Once the novelty value of these NotebookLM generated podcasts wear off, I don’t know how sustainable it will be. Unless you are driving or working out or doing dishes, audio is a suboptimal means of consuming content - you are far better off consuming text instead (also audio needs much more redundancy, as a function of the medium itself, which makes things long).

• That said, podcast is only one of the possible outputs of NotebookLM. It can generate FAQs, answer questions based on the content, etc. If our requirement is that we’re not able to beat our own drum particularly well, maybe we can just use it to beat ours instead? Maybe ask it to remake the copy etc.

• AI is likely to be a “product management game”. A lot of cool tech has already been built, and more is going to be built for sure. The way to monetisation is on how you package this text to serve particular needs of customers. I know this sound random coming at the end of this blogpost, but whatever.

• The bar for humans creating has gotten higher - your writing / speaking / doing needs to be far more idiosyncratic if people have to pay attention. I suppose “An AI could have written this” has already started becoming an insult.

• The best thing about AI handling your corporate communications is that it won’t try to boss over you. Or so I hope!