For Immediate Release

FIR #459: AI Transforms Content from Passive to Interactive


Listen Later

In this episode, Shel Holtz and Neville Hobson discuss the evolving landscape of podcast consumption, particularly in light of Satya Nadella’s innovative approach to engaging with audio content through AI. They explore the significance of transcripts, the potential for AI to facilitate interactive experiences, and the challenges that come with adopting these new technologies. The conversation highlights the future of podcasts as a medium that can be both passive and interactive, reshaping how audiences engage with audio content. Neville and Shel also examine how these same generative AI tools can make other content interactive and the ease with which users will be able to take advantage of it as LLMs become multi-modal.

Links from this episode

  • The surprising way Microsoft CEO Satya Nadella uses AI to consume podcasts on his commute
  • Podcast Transcription: How & Why You Must Transcribe Podcasts
  • The next monthly, long-form episode of FIR will drop on Monday, February 24.

    We host a Communicators Zoom Chat most Thursdays at 1 p.m. ET. To obtain the credentials needed to participate, contact Shel or Neville directly, request them in our Facebook group, or email [email protected].

    Special thanks to Jay Moonah for the opening and closing music.

    You can find the stories from which Shel’s FIR content is selected at Shel’s Link Blog. Shel has started a metaverse-focused Flipboard magazine. You can catch up with both co-hosts on Neville’s blog and Shel’s blog.

    Disclaimer: The opinions expressed in this podcast are Shel’s and Neville’s and do not reflect the views of their employers and/or clients.

    Raw Transcript:

    Shel Holtz (00:03.168)

    Hi everybody, and welcome to episode number 459 of Four Immediate Release. I’m Shel Holtz.

    @nevillehobson (00:10.742)

    And I’m Neville Hobson. One of the more thought provoking stories I came across recently was a short piece on Geekwire about Microsoft CEO Satya Nadella and how he listens to podcasts, or rather how he doesn’t. According to Nadella, the best way for him to consume podcasts these days isn’t by listening to them in the traditional sense. So what does he do? We’ll discuss that in just a minute.

    @nevillehobson (00:38.72)

    Instead, during his commute, he interacts with the transcript of a podcast using his personal AI co-pilot. He speaks to it, asks questions, interrupts when needed, essentially turning what would normally be a passive listening experience into an active conversational one. Nadella describes it as a full duplex conversation, a two-way interaction, which until recently would have seemed futuristic. This kind of back and forth modality, he says, is more convenient and powerful than traditional listening.

    His comment was, there’s no going back. This shines a spotlight on a crucial, but sometimes overlooked asset for podcasters, the transcript. Providing a transcript isn’t just about accessibility, although that’s a critical benefit. It’s also about discoverability, repurposing, and now enabling new ways of engaging with content. Transcripts can enable a number of things, such as improving SEO, helping your podcast get found in search.

    makes your content more accessible to people who are deaf or hard of hearing, enable content repurposing into blog posts, social media or newsletters, and as Nadella demonstrates, allow AI tools to analyze, summarize or even hold conversations about your content. In a world where attention is scarce, transcripts are no longer just a nice to have, they’re becoming essential infrastructure for how audiences, including high level business leaders, interact with audio content.

    So here’s the primary discussion point for our conversation today. Could Satya Nadella’s model be the future for busy professionals where AI acts as a bridge between long form audio and actionable insights? And if so, what does that mean for podcasts and communicators in how we produce a packaged content? What do you think, Cheryl?

    Shel Holtz (02:23.167)

    I think there will be a fair number of people who will take advantage of this and the activities that you’ll be able to engage in like it that will emerge as these tools continue to evolve. I think there will be a majority of people who will continue to listen to podcasts because it is the lowest friction way to listen to a podcast. There’s no work. just go to the podcast app, pick the one you want to listen to. Latest episode play.

    And a lot of people don’t want to work in their car. They’re driving, they’re relaxing. Asking questions is not on their bingo card for their for their drive to or from work, for example, or or during a road trip. So I think people will continue to take the easy road for the most part. But for people who are trying to learn a lot or glean information.

    related to current events or emerging science or whatever the theme of the podcast they listen to is. Yeah, absolutely. I think this is going to be a popular approach and there are more than one way to do it. For example, you could put that transcript into Google’s Notebook LM. I don’t think you can talk to it yet, but you can. can you? OK, great. Well, there you go.

    @nevillehobson (03:40.267)

    Yeah, yes, you can. Yes, you can. That’s just just just been introduced. But it’s different because there you’re talking to the two AI hosts and you then having a conversation you can’t script.

    Shel Holtz (03:49.226)

    Right.

    Yeah, exactly. But if the transcript has been loaded into the notebook, then the conversation, obviously, you can listen to the podcast with the two hosts talking about it. But at that point, you might as well just listen to a podcast, one or the other listen to the original. But the whole idea of Notebook LM, even before they introduced that podcast feature, was your ability to query the notebook based on everything that’s in it. You could have 50 episodes of a podcast or you could load

    the transcripts of the most recent episodes of 20 different podcasts on the same theme and just start querying it. And it’ll give you answers to your questions. This was what makes it so powerful a tool. think there’s also, by the way, I’m using a tool now. This is not for podcasts, but along the same lines, it’s called Drip Drippp, three P’s. what I do here and it’s a paid service is

    I take all of the AI email newsletters that I subscribe to, and I subscribe to probably 12 or 13 of them, and it gets all of them and sends me one email summary of what’s in all of them. So I’m now reading my one daily drip rather than all 13. Email.

    newsletters, which is particularly useful because there are days I don’t have time to read any of those newsletters, but I find the time to read that one drip. this being able to repurpose, which is kind of what we’re talking about here, content to make it more consumable for you in the circumstances that you’re in. This is, I think, one of the powerful uses that we’re seeing a lot of people start to adopt AI for.

    @nevillehobson (05:39.17)

    Yeah, I mean, in this specific case of what the Sati Nadella did with co-pilot, it’s not so much reading it out. It’s the way in which it interacts. You interact with it. And literally it’s random. It has no foresight of what you’re going to ask, but it finds the content you’re looking for. So you could ask it, for instance, tell me a bit more about the topic that Shail and I talked about on

    Shel Holtz (05:49.931)

    Mm-hmm.

    @nevillehobson (06:08.333)

    XYZ topic or at about the 18 minute mark. Can you quickly summarize the key points of what we discussed? And it will do that. And that is wow to me. The first one’s easy. You could tell it as I did in an experiment I did, which you’ve got a clip we’re going to have. You can be able to listen to that in a minute is to summarize the podcast. And what I did was upload the transcript of our previous episode for 98 for 58. Sorry.

    And that was the one we talked about on AI being a part of your team, an AI chatbot and all that. So I uploaded the whole thing and asked it a couple of things and you’ll hear that in the transcript in a minute. But it got me thinking that the interview with Satya Nadella is worth a read. It’s quite concise, but you can project out your own thinking as to what this might mean and what Satya Nadella talks about his experience with it and particularly the multimodal element.

    Your point about driving got it entirely, although if you’re commuting on a train or a bus, no big deal. You can just do it like that. But I think the idea of being able to tell the chat bot, give me a summary of the episode 458 and it’s got, it’s uploaded and it tells you that that might prompt further things. Tell me, you you and I talked about such as a topic. Give me the key takeaways that we discussed.

    So it’s literally as fluid as that. It’s not more structured than that even. And the experiment I did, which may or may not be an indicator of wow or not, worked extremely well. There are some big downsides with this. And this I wondered about what Satya Nadella is doing about this, because the work in preparing the transcripts and uploading it to the chat bot, each time for each episode is severely not.

    good from ease of use and all that. There’s big barriers that you’ve got to be really keen to do this. I did it. One thing I found, Nadella used Copilot. I knew few people using Copilot actively, but I know tons of people using ChatGPT. So I asked ChatGPT, can you do this? To which it replied, seriously, confidently, yes, of course I can.

    @nevillehobson (08:28.755)

    And it then actually asked me, what do want to do? Upload the transcript and tell me how we can have a conversation. I said, well, that’s exactly what I want to do. So I followed the advice on ChatGPT’s page about this, which is to use the mobile app on my Android phone. But when I tried to upload the transcript, ChatGPT told me on the app, sorry, the app doesn’t support file uploads for this use. You can upload files, but not for this use, using the audio.

    engagement function. But I did discover that the web app, the PWA on Windows 11 does support that. And so I engaged with the chatbot via the PWA, the personal web app on Windows 11 on my desktop computer, not the mobile. But it was an interesting experiment in my engagement with chat GPT that way, very conversational. And I was actually very impressed indeed with with how it performed.

    So let’s include the clips so people can listen to it, Shell, because it was well done. It’s lightly edited to edit out some of the gaps in it to make for a better listening experience. And I’ve amplified the chatbots audio a bit because I was doing this on my microphone, the desktop, and the sound was coming out of the desktop speakers, which had to amplify a bit. But I think you’ll get the idea here. When I asked it to summarize, then I asked it specific questions about one of the segments we discussed. So let’s take a listen.

    @nevillehobson (10:01.899)

    So what do you think of that, Cheryl? I mean, that’s a simple instance. I honestly couldn’t imagine me going through all the faff of uploading transcripts for each individual episode and then listen to it on my car and asking it questions. I just simply couldn’t imagine that. But as you mentioned earlier on in our conversation, things like this are only going to get easier, I would imagine, don’t you think?

    Shel Holtz (10:24.711)

    absolutely. For a number of reasons. One of the big ones is that what’s coming is the large language models, the big frontier models are all going to be multimodal before too long, which means that you’ll be able to upload the audio podcast or even the video version of the podcast from YouTube or just point it to it. And it’ll be able to do exactly the same thing without having to go through the rigmarole.

    of the transcript. this is just going to get easier and easier as time goes on. The other thing that you mentioned having to create the transcript, a lot of podcasters are including a transcript in their show notes for SEO purposes, and we are among those. I don’t edit it. I don’t go in and fix all of the errors. I don’t have time. It’s raw. So that at least somebody searching for some keywords might find us. But all you would have to do with

    @nevillehobson (11:14.391)

    For this you’d have to.

    Shel Holtz (11:20.743)

    any podcaster who’s doing that is copy and paste that transcript. You don’t have to go through any hoops to create one.

    @nevillehobson (11:27.241)

    no, no. When I said you have to do something with the transcript, I didn’t mean that at all. You don’t write it out. But I’ve noticed this in the past on other experiments I’ve done. If you upload a raw verbatim transcript with lots of Ams and As in and some things that didn’t quite catch and so it does it wrong, that will seriously impact the quality of what you get back. So for something like this, I would definitely edit lightly the output. And of course, on a 20 minute episode such as we had, that would be an easy thing to do.

    but that doesn’t scale. this isn’t, in my view, this is not a prime time tool for everyone to think, wow, I can use this. This is if you’re keen, if you’ve got the patience and the time to go through the prep. So you prep 10 episodes, let’s say, or even your transcripts. Let’s say you speak, you’ve got the ideal audio environment in which you’re recording, in which case the recording software will pick up everything about 99 % correctly, like Riverside that we do, for instance.

    And that’s got a method to smooth some of the stuff. Or you’ve got Descript. I mean, there are tools you can do this. The point, though, is there are still separate tools you’ve got to use to get this into the state before you can upload it and share it with the chatbot. That may well be a barrier too much for people. The idea is fabulous, I think. I love this idea, which is why I was so keen on trying it out myself. And I think it is going to be part of the landscape pretty soon, just without all the barriers, hopefully.

    Shel Holtz (12:52.009)

    Yeah, and the easier it gets, the more people will do it. But to your initial question, I don’t think podcasters have anything to worry about. think people, by and large, are just going to continue listening to podcasts. After all, a lot of people listen to podcasts because they like the hosts. They like the segments. They like the vibe. And you won’t get that out of summaries and the ability to query. But on the other hand, I listened to an episode of a podcast that was over five hours long.

    @nevillehobson (12:59.925)

    No, no.

    Shel Holtz (13:20.595)

    It was interviews with three people. think they were from Anthropic, but it was a five hour podcast. Boy, would I have loved a summary and the ability to query that because a lot of it was very over my head.

    @nevillehobson (13:29.335)

    Yeah. I mean, I think.

    Sure. I think this is in that area, of course, this is just another method of engaging with the content. It isn’t intended to do this instead of listening to the podcast. Not at all. If you want to do that, like you said, get a summary, whether it’s 20 minutes or five hours, that could be handy. It might work good for if you’re the kind of person who likes trying out new things. There is a new podcast. don’t know. Let’s get it. Let’s upload the transcript and get a summary. mean, that’s maybe not, you know.

    daily activity. But all of this will be part of the landscape to make things a lot easier. And I could even imagine that one of the chat bots, you’re not going to go through all this stuff of uploading files and editing, you’re going to tell the chat bot to do that. Look at the file, get the transcript, see if it’s okay, share it with your colleague at ChatGPT. We’ll take it from there. But it’s a neat idea. I’d keen to know anyone listening, one of our listeners has done this with

    Copilot or are you thinking about it? Who’s using Copilot? ChatGPT were hurdles. I wonder how that would be with Copilot. I did ask ChatGPT to compare this feature with Copilots, which it did. And it came back and Copilot excelled in areas such as complete integration with all your typical software apps you’ve got on your PC. Whereas ChatGPT, this is not integrated in the same way. That’s today. I suspect that’ll change.

    Shel Holtz (15:02.582)

    So what we’re talking about here is taking a passive medium and extending its utility to make it an interactive medium, which is amazing, frankly. And it’ll be interesting to see what other kinds of media can be made interactive with the aid of artificial intelligence. And that’ll be a 30 for this episode of for immediate release.

     

     

    The post FIR #459: AI Transforms Content from Passive to Interactive appeared first on FIR Podcast Network.

    ...more
    View all episodesView all episodes
    Download on the App Store

    For Immediate ReleaseBy Neville Hobson and Shel Holtz

    • 5
    • 5
    • 5
    • 5
    • 5

    5

    20 ratings


    More shows like For Immediate Release

    View all
    TED Talks Daily by TED

    TED Talks Daily

    11,256 Listeners

    Fresh Air by NPR

    Fresh Air

    38,086 Listeners

    Planet Money by NPR

    Planet Money

    30,891 Listeners

    Freakonomics Radio by Freakonomics Radio + Stitcher

    Freakonomics Radio

    32,108 Listeners

    Hidden Brain by Hidden Brain, Shankar Vedantam

    Hidden Brain

    43,250 Listeners

    Pivot by New York Magazine

    Pivot

    8,814 Listeners

    The Daily by The New York Times

    The Daily

    111,470 Listeners

    Today, Explained by Vox

    Today, Explained

    10,061 Listeners

    The Spin Sucks Podcast with Gini Dietrich by Gini Dietrich, Founder of Spin Sucks

    The Spin Sucks Podcast with Gini Dietrich

    64 Listeners

    The Bulwark Podcast by The Bulwark

    The Bulwark Podcast

    11,386 Listeners

    The Journal. by The Wall Street Journal & Gimlet

    The Journal.

    5,886 Listeners

    Think Fast Talk Smart: Communication Techniques by Matt Abrahams, Think Fast Talk Smart

    Think Fast Talk Smart: Communication Techniques

    772 Listeners

    Hard Fork by The New York Times

    Hard Fork

    5,338 Listeners

    The 7 by The Washington Post

    The 7

    1,140 Listeners

    The Headlines by The New York Times

    The Headlines

    416 Listeners