Podnews Weekly Review

Interview with Ken Miller from Fathom; Spotify buys an AI voice service, Google Podcasts slips up in Pride Month, Sam looks different to what Adam Curry thinks

June 16, 2022 James Cridland & Sam Sethi Season 1 Episode 79
Podnews Weekly Review
Interview with Ken Miller from Fathom; Spotify buys an AI voice service, Google Podcasts slips up in Pride Month, Sam looks different to what Adam Curry thinks
PoWeR Supporter
Become a supporter of the show!
Starting at $3/month
Support
Show Notes Transcript Chapter Markers

Special Guest

Show Notes

Support the show

Connect With Us:

You know, there there's a couple values we have at fathom and it's. And bafflement, if you want a successful startup, what you have to create is hot, impossible, magic. You know, like anything below hot, impossible magic is not gonna be a big one, but that first time you hit that button and a car came and picked you up.

You're like, ah, that's hot and possible magic. Well, to Poland's the last word in podcasting news. It's Thursday, June the 16th, 2022. I'm James Cridlin the editor. Pod news.net. And I'm Sam Setti, the MD of river radio. Hi, I'm Ken Miller co-founder of fathom, and I'll be on later to talk about our new app. He will.

Pod land is sponsored by squad the remote recording tool that your audience will love. Squad version five is coming later this month. There's not much of this month left, uh, with new features and a new look, squad.com for more. And we're sponsored by buns, brow podcast hosting made easy last week.

Thousand 374 people started a podcast with buzz brown and now there's buzz brown adds to grow your podcast. Wherever it's hosted, you can find out more@buzzbrow.com slash ads. Pod land is where James and I review the latest news from the podcast industry. James, the, uh, first story up this week. Is Spotify it's acquired, uh, semantic an AI voice platform, uh, used to simulate Val Kilmer's voice.

So the first thing I have to ask is, have you seen top gun Maverick? I have not seen top gun Maverick. No. Um, uh, no, in fact, I, I, I normally have a little bit of a joke that, uh, I don't watch movies. And the last movie that I watched was Superman three. That's a slight line, but it's, it's not too far away from the truth, to be honest.

So, uh, no, but is it. Uh, top gun Maverick is absolutely brilliant. I actually have to say it, it most sequels fail. This one actually, uh, exceeds and does very well. The end is inevitable. Maverick. You kind of headed it for extinction, maybe so. Sorry. And the bit within it, where Val Kilmer who's got throat cancer actually then talks, um, Till I knew about this.

I thought it was Val KMA talking. So, um, amazing what they've done, but who are symantech tell me more. Well, they are a, uh, AI voice platform. What does that mean? It means that they make, um, some, uh, really interesting. Sounding voices. Um, I played this little clip on the pod news podcast. The other day, what you're hearing me say was never said by a human, it was generated by a computer.

I'm not real. I was never born and I will never die because I do not exist. I know sounds like something out of a movie. Right? . It's real. So, yeah. So as you can tell from that, uh, it sounds just really real. Um, and, uh, you know, you can make, uh, these, uh, artificial voices that just sound completely realistic, which is, uh, really clever.

A question is what are they gonna use it? Uh, four at Spotify. Have you any idea? I thought about it cuz they have their incar hardware device. Don't they now? And I wondered whether they were going to use it for voice commands for their incar, uh, device. Oh, okay. Yeah, maybe that would be, um, well, I'm gonna ask Alice Mitchell, he's one of the investors, uh, from EQT ventures.

Uh, he's an old friend of mine, so I'm gonna ping him and see if he'll come on pod land and tell us what he thinks they're gonna do at Spotify with his, uh, investment. Yeah. Well, that sounds really interesting. Another person who, you know, uh, Sam says  is just like, you know, you know, everybody is very, no, no, no.

So, yes. So that was one of the, uh, of the acquisitions, uh, global has made an acquisition as well. Yeah. Um, they've bought a company called OD, which, uh, I have to admit I've not come across. So tell me more about them. Yes. Well, not that OD for a start, not the OD that, uh, um, was a podcast platform back in 2005, that then turned into Twitter, not that one.

Uh, it's audio with two E and it puts audio ads into mobile games. So if. A game and, uh, you know, while you are, I don't know, shooting the battery or whatever, then you hear an ad for, I don't know, um,  something for, you know, headache pills or something who knows. But anyway, um, what global have done is they've made a strategic investment, so it isn't, um, it isn't an acquisition quite yet.

Um, but they're also going to become exclusive sales partner in the UK and the us as well, which is. Clever. Uh, so you know, more audio ads, audio ads in a slightly different place. Uh, Global's a very clever company because it's basically it's selling audio ads in podcasts through DS. It sells audio acts in radio stations.

It owns capital and heart and other ones in the UK. Um, it sells. Outdoor advertising, uh, on places like the, the London underground and, you know, and, and across the UK in terms of billboards, and now it's getting into, uh, online games as well. It's doing some very clever things in terms of, in terms of advertising.

So certainly a company to, um, keep a watch on, I think, yeah. Daks needs to extend its reach and I, I can see why this is just another interface where they can put ads into. I wonder what else they're gonna look at. Um, you know, if you think about. The smart speakers are beginning to get a lot of ads into them.

I've noticed. Um, again, I wonder whether DS will be doing anything in that space as well. Yeah. And of course, they're doing a little bit of that in terms of the global player, which is, um, a, uh, podcast app that they own. Um, but, uh, yeah, it's, uh, I, I think it's a fascinating company. I think it's one of those companies, there was a talk about a year or so ago of them buying iHeartRadio.

Um, so whether or not they do. It, you can certainly see that they are a company with, um, quite a lot of aspirations. So, um, you know, worthwhile taking a peak, they also of course, own captivate podcast hosting company. Uh, so, um, you know, who knows what they're going to do next? Now let's move on. Uh, the good, the bad and the very ugly, I'm sorry.

A cast. Look, let's start off with a good, we do like a cast. They are. Good company and they do some very good things. And one of the things they've done recently, James is they've announced the virtual audio pride parade, uh, to celebrate L G B Q I a plus events over the next three months. Um, so those are the sorts of things they do very well, James, but they do a lot of, uh, naughty little things.

What have they done this time, James? Oh, they've done well. So they they've announced a partnership with spring. Um, a company that, uh, you might know, uh, used to be called Teespring and, uh, they allowed people to make their own, um, you know, t-shirts basically, um, they've expanded out to merchandise and they've expanded out to NFTs.

Now I actually missed this in the press release that, that they sent, I looked at it and I thought it's it's merch it's tote bags. It's t-shirts, it's that sort of.  but, uh, what spring also sells is they sell NFTs as well on your behalf as a content, uh, creator. And as soon as you mention NFTs, uh, there's, uh, parts of the internet that gets very upset with you.

Um, and so, um, you know, perhaps that's, uh, perhaps that's a bad thing, perhaps that's a good thing. I don't really understand enough about NFTs to really make a decision. Well, I look at NFTs as trying to create scarcity in a, in a world of abundance. Um, I'm not sure right now, some of the things that they're applying NFTs to actually make sense, um, some feedback from Twitter, uh, T said, uh, I'm once again, reminding podcasters that the day you sell NFTs off your podcast is the day I stopped listening to your show.

So, uh, not a positive. And on the back of that, though, of course, there was an announcement this week that you wrote about from a company called uncut.fm, which is a new podcast platform that lets you min share and sell content as NFTs. So. It seems that other people apart from just Acast are looking to get into trying to monetize podcasts using NFTs.

Yeah. And, you know, uh, I, I think again, interesting. I, I mean, you have a look@uncut.fm and. You know, there's lots of things that I don't, again, I don't fully understand, or perhaps it's just because I'm too old. I don't know. But, uh, I, I look at it and I go, sorry, what are people buying again? I don't really understand it.

Um, but you will have the CEO of uncut.fm on this very show next week. So I'm looking forward to, um, uh, beginning to understand a little bit more about the wide world of NFTs. Yes. We'll have Carlos Diaz, uh, on the show. So. Moving on to the other things that ACAS have been doing that aren't very good. Um, ACAS has been, misleadingly claiming that one of its companies has public benefit corporation status.

That means legally defined goals for positive impact on society workers, the community. And the environment James, tell me what have they done? Yeah. So this is, um, radio public that Acast bought in February and radio public initially was a public benefit corporation, but, uh, obviously as soon as Acast bought it, then they turned not.

To be that, but we're actually still claiming that they were a public benefit corporation, which is, you know, it's a legal thing that you have to, uh, support society and workers in the community and the environment along with profit. Um, that's what you are supposed to be doing as a public benefit corporation.

and radio public was saying that they were a, a public benefit corporation, but they actually ceased to exist legally at the end of this year, December the 31st, uh, 2021. Uh, so the end of last year, um, and, uh, basically they've still been claiming that they are a public benefit corporation when they're not, um, I spotted this in March and I asked Acast and I said, Hey, just, just as a matter of interest, are you still a publicly benefit?

Uh, are you still a public benefit company? Because I'm not sure you are. And they said, oh, well, you know, we'll, we'll, we'll get back to you on that. Um,  and here we are in June, weirdly, as soon as I publish it that very same day, they went in and edited the website and got rid of any mention of public benefit corporation, at least from their front page and from their privacy policy.

So. You know, do you really have to go public in order to get something that's misleading and incorrect changed? That's kind of not great in terms of ethics. And the other thing that's not great in terms of ethics is what a cast is continuing to do in terms of spaming other customers of, uh, other podcast hosts.

Um, captivate has just been spamed with, you know, a, a, a bit of a cheeky. Email which contains even logos of captivate or logos of transistor or logos of, you know, various other people as well. And, uh, it's just, uh, you know, it's not the ethical thing. It's also not the right thing to do, because if you are there trying to grow podcasting as a whole, then you don't do that by cannibalizing other people's businesses.

You go out and you get new podcasters. That's what Acast should be all about. And it's. A shame. I think that Acast has turned from being a company, which is, you know, quite a well thought of, you know, professional company to a company, which, you know, if you, if you look on, uh, Twitter for a search for Acast, all you see is negative comments about, uh, them spaming about them, you know, fighting.

Podcast hosting companies, you know, et cetera, et cetera, those, uh, emails that they were sending out, uh, earlier on this year were blatantly illegal. Um, I actually forced them to make some changes to those emails for compliance. Now I think that they're probably legal if not particularly ethical, but just in Jackson who runs transistor has reported them to Canadian law enforcement and to a cast's email service provider.

Because Justin reckons that they're not legal, um, in, uh, Canada. And, uh, he, um, is, you know, obviously a little bit miffed because they've been, uh, trying to target his customers as well. Um, and it's just, you know, I don't know, maybe I'm making too much of this, but, uh, I'm just there thinking it's not, it's not the right thing to do.

Is it? Well, I dunno if you are making too much, it cuz clearly. The CEOs and founders of other companies have equally miffed. I mean, mark cast was tweeted, came back to work, seeing Acast spaming captivate customers, where they move to Acast email based on scraping email addresses from the RSS feeds. Um, he said of Ross Adams.

It just seems to ignore it publicly. Like it's never happened from all of the feedback I've heard about it. I'm pretty embarra. It's pretty embarrassing. As a CEO, the buckle stops with a person highest. And if that person hides, it's a character reflection. Woo. That's a, that says a lot about Ross. Yeah, no, indeed.

A spokesperson told me, uh, earlier on in the week that the emails are quote part of our continuing marketing strategy, which we don't currently plan to change. Mm-hmm. So, um, basically showed me the finger and that's where Acast is gonna be, I guess. Yeah. Mark Deadman said I can. O I see. Not only are Acast spaming podcasters, they're also doing the gross.

We're just following up email. Uh, he said, if you are hosting on Acast, I think it's time to switch. If they're using these tactics, they must be desperate. Yeah. And they're actually sending two follow up emails. Um, they sent one to this very podcast. And, uh, it ended up saying, you know, move away from buzz sprout, um, because you know, uh, we'll give you more, more for your money and all this kind of, uh, stuff we're quite happy with buzz sprout, uh, for a start.

And, uh, secondly buzz sprout sponsor us. So, um, uh, yeah, so, no, thanks.  um, but, uh, yeah, so, uh, I. They would stop. I wish they would stop. Let's move on to Google podcasts who have also done something a little bit, weirdly inexplicable haven't they? Yeah. Uh, it seems James Google don't like lesbians. No. Tell me more.

it's pride month, right? It's pride month. And so during pride month, what better to do, uh, if you are a podcast, uh, directory, then block and hide some podcast episodes because they use the word lesbian in the title. Uh, so this has happened to, um, a podcast called RPG realms of peril and glory, which has, uh, an episode which, um, uh, is, uh, called something.

I can't quite remember what it's called, but it's called something like the sword shrunk lesbians or something. Anyway, uh, there's nothing adult. In it, um, it's, uh, a, um, you know, it's a, it's a play by play, uh, podcast, uh, which includes a same sex relationship. Ooh. Uh, and that's about as far as it goes, you can see that episode on apple podcasts.

You can see that episode on Spotify. You can't see that episode on Google podcasts. Um, even if you're logged in, even if you're over 18, So don't really understand what's going on there. Why is Google censoring that they don't censor the Alex Jones podcast? Why are they censoring this, um, podcast, which is about, about two people, loving each other, who knows?

now, uh, new platforms. Uh, we talked about uncut launching, uh, another new platform. That's launches called FM. Uh, it's a new podcast. App. It's a search engine that uses AI to discover parts of your podcast within the episode, it's launched on iOS and in the browser. I do know that an Android version James is coming out very shortly.

Um, I should Johnny, I hope so. Um, , it piqued my interest for two reasons. One cuz of the use of AI for, uh, transcription, uh, discovery. Uh, and secondly, because Jason Cahan is the, uh, famous VC, uh, investor over in California is also part of their investment. Uh, Roster. So I thought I'd reach out to Ken Miller.

Who's the co-founder and CTO to ask him to tell me all about fathom. I really like the metaphor that goes along with fathom, which is a form of depth. I think that one of the things that podcasts offer is a certain depth of conversation that comes along with nuance that you don't find in other forms of media.

And certainly working it as a verb, people often say, oh, just Google it. You can fathom. What was the itch or what was the em, embryo of the idea that led to fathom? It's something that's been cooking up for a while? Actually, there was a conversation. My co-founder Paul block, and I have known each other since high school and have worked at multiple startups together in companies together and years ago, this must have been like eight or nine years ago.

We were having a conversation about, uh, this philosopher Allen Watts, and Paul was really into Allen Watts and he thought it would be fun to make an app where you just press a button. And Allen Watts starts talking, cuz anything that comes out of his mouth is typically pretty good. We didn't build that app.

At least until years later, I actually ended up creating an Alexa app called Sage Allen Watts. And in the process of creating that application, I thought it would be super interesting. If you could actually ask this dead philosopher questions, like almost resurrecting him in a sense. I'm sure Allen Watts is rolling in his grave that never really happened.

It actually did. In the sense, since there is an Allen Watts podcast and you can ask it questions. Fast forward a few years. And I had started to listen to a lot of podcasts. I got really into L Friedman and as anybody who gets into podcasts comes to discover discoverability, searchability. These are serious problems in the space.

And at the time I was working with, um, a lot of AI technology, not so much in the. The media space I was working, I was the chief research officer at the startup called in trio. And I was working with a lot of natural language processing, but in the financial sector, so processing 10 Ks and 10 QS and whatnot.

And, and as a result, I was really heavily embedded in NLP and AI. And I decided to begin to take a look at what would it look like if you could actually use the latest advancements in AI to ask a podcast, a question. And actually have the AI deliver you the answer from the podcast, almost appropriating that intelligence.

And I got a prototype going, and of course it like so many engineers out there. I got the prototype going and then it went into a folder and I forgot about it, but I can credit Lex. Friedman actually for prompting fathom again, in that I was listening to one of his podcasts and I had fallen asleep to it about an hour, goes by and I get shocked, awake outta my, out of my sleep.

And he, uh, at the, and the podcast was still playing and he was talking about searchability of podcasts and how. Game changing. It would be if you could search podcast. So I saw that as a union synchronistic sign from the universe that I should dig that prototype out of the folder and sure enough, three months later we were, uh, we were accepted into the launch incubator.

Yeah. So launch incubator for those that don't know is run by Jason Cal canis at the famous venture capitalist over in California. So what you just applied, went through the process with Jason and got accepted. We had some connections. So once I just decided to really take the project seriously and take it all the way, something beyond a, a prototype that I can show my wife, I reached out to my co-founder Paul.

He had just come back from Japan. He actually spent his one year stint as a Zen monk in a monastery in Japan. And so I was like, Hey, Paul, do you. Put away your monk, like ways and become a capitalist with me and do this startup. And he was like, sure, Ken sounds great. What do you got? And I was like, I have an AI search engine for podcast.

He's like, awesome. So we worked on it a bit. Paul came at it with his product eye and his design eye. And once we really had something to show, Paul had been at 1.1 of the designers at yam. Prior to its acquisition. So he knew David Sachs and was connected with some of the other associates over at craft.

He reached out to them, they referred us to Jason and Jason's crew and we gave one of Jason's people, a demo, and we were shortly accepted immediately. Afterwards. Now let's unpack, fathom a bit. How does it. AI machine learning requires volumes of data and training and learning. So there are 4.3 million podcasts out there.

What are you doing? Are you indexing everything and then unpacking it all? Or are you taking a subset? How does it work? I guess, especially for this particular audience, some of the important mechanics to understand, no, we are not using AI transcription for 4 million different podcasters. Unfortunately, that's just not tenable from a cost perspective right now.

I think we're doing about 3000 podcasts. You can access all of the, it's not really four. Once you cleanse the data. I think it's closer to, you can access over 2 million podcasts on fathom, but we're only. Actively transcribing and then actively AI processing about 3000 podcasts in the next six months.

I'm hoping to grow that. We built our own internal speech to text engine, which was quite an, an endeavor, but that's gonna allow us to get up to about hopefully. 25 30,000 podcasts in the next six to eight months. If you would like to have your podcast actually transcribed and processed by our AI, the easiest way to get that done is to get your audience, to follow the podcast on fathom because I've built in mechanisms to where as soon as people begin to follow something, we begin to transcribe it.

And then if you wanna take it one step further, if you put a link to your pod on fat, In your show notes, I'm building something into our ingestion pipeline that will check the show notes. If a link to fathom is in they're right along with Spotify and apple, we will basically put you in a higher tier of AI processing.

We're gonna use a more expensive AI transcript generation engine and do more in depth, AI processing and search indexing. I'll be doing that in our show notes. We. Very in-depth share notes. And so anyone listening to this podcast, please, please, please download fathom and go and follow pod land so that we can get an index for it.

Now, once we've done that, what does that give us the capabilities to do? Because one of the things I found very difficult, and I guess I found this with the Alexa, when I first got them was I didn't know what to ask. And I've seen some reviews where people have asked some crazy stuff and stuff's come back and some stuff hasn't come back.

And so how do you go about improving what you've got? And what's the mechanism when it comes to question. Fathom really excels at very high level questions. What kind of supplements could I take to improve my cognitive function or what's the best way to begin meditating? What are the keys to start up business success?

So I think it can take a little bit of reeducation cause people aren't even used to searching inside of a podcast at all. And then you're going one step further than that. And being like, don't talk to it. Like Google, don't talk to it like a robot, ask it questions like you would ask your friend questions.

So I think there's a bit of reeducation that's gonna take some time to see, but yeah, so we wanted to do the hard things first with the search engine and question answering with audio it's incredibly novel and it's. Challenging to do from an AI perspective. So we wanted to do that first. I would say there's certain aspects of the search engine.

That don't work as well. If you're just like throwing keywords at it currently, but we're working on those. We will be improving the surgeon engine. This is very nascent technology. So over the next year, you're going to see like a 50% improvement in the kind of results you can get. So that's one aspect of it, the algorithms themselves, the other aspect is the amount of content right now.

There are only. 120,000 or so individual episodes that have been indexed by a search engine. And that is across a wide variety of topics, but it's heavily focused on the kind of like science technology, business categories of content. So if you're gonna ask something about. Kayaking any like leisure stuff, you might not get that great of results.

But if you do ask something in the business technology, science categories, you will get good results. Again, that is something we're gonna be growing the library of content that we're processing over this next year. The other thing that fathom does is. Our AI is actually able to comprehend the conversation itself.

So when a new episode comes in first, we transcribe it. That's gonna take it out of the audio domain. Now we're in the text domain, which opens up the whole, uh, swath of natural language processing, AI advancements that have come into play over the last five years. One of the things we do is we also extract highlights.

Episodes. We use our AI to find interesting or funny or even profound highlights. And then for our users, we give them a feed of recommended episodes with highlights from all of those episodes so that they can get a sense of what the conversation was about and decide whether or not they wanna jump in and listen to the entire conversation.

And what's fun about this is that. Our AI finds multiple in interesting parts within an episode, it actually selects the interesting part that it thinks will be most interesting for the user. That's listening based on what they like. They listen to and they follow. It's a fairly complicated content recommendation algorithm, and it's getting better every day.

So that's pretty powerful. Now are the clips that come out, are they uniform in length or are they variable based on the conversation? Right now they're all uniform length. I would say that one of the areas we're investing heavily in right now is actually that highlight generation functionality. We have some new models and some new novel approaches with any of these AI startups.

You're always running into the issue where you have to go through this R and D process to get something that's good. And that takes time because AI. And machine learning, it's less like regular software engineering. It's not like framing a house. It's more like picking a lock. And, uh, once, once you got the door open, then you actually have to go through the door and get into the kitchen that can take some time.

So it's interesting because we always have a lot more technology that we know that we have, but that nobody else knows that we have. And in. There's actually a couple different models that we have that are gonna help us have variable length clips that have very on point start points and very on point end points.

And so that's something that we're actively working on. So where's the monetization of all of this. So how do you make the money? I love Jason, but Jason's not a benevolent person. He'd be looking for a 20 X return. So where's the money in this. There's all the obvious routes. This is a consumer play typically with a deep tech consumer play.

You, you can potentially be looking at years before you, you decide to monetize. You're really looking for explosive user growth, which is not to say that we are in any way, opposed to beginning to test for monetization. Ads are obviously something you can do. And then all of the typical business models, then you can have premium accounts that remove the ads.

I think that there may be some opportunities in the creator space. There's a ton of AI analytics. About the episodes that we have that I think creators would be very interested to see. So there may be some like premium accounts for creators to give them access to some of these AI analysis tools. But yeah, a ads is gonna probably be first thing that, that we begin experimenting with.

Acast have started something called conversational advertising. They haven't really revealed how they're doing it in terms of, are they using AI or just transcribing it and then looking for interesting keyword, but you can see how they're gonna map, I guess, brands and people who want to advertise against the transcribed content of the podcast to try and match somebody who might be, for example, talking about business, but happens to be talking about cars within that specific podcast.

And so they'll find a car manufacturer to map that against the transcribed conversation. Yeah. With where we're at with our technology today, we can do that immediately. We already have that capability, but I don't think the first place we would be looking to experiment with ads would involve ad insertion in the audio itself.

Mm-hmm  that really. Requires having a relationship with the podcaster that they, I should be getting a cut of that revenue. We can certainly do it. I think that it's one thing, obviously. So somebody's talking about cars and now we're gonna insert and add about cars. I think what's interesting about fathom as a platform is that we know a lot about the listener, right?

We're basically able to look at everything. You listen to everything you like, everything you follow and that we could take all those transcripts and paste them together. And now our AI can understand all of that. So that means that our AI fundamentally understands our listeners. And we use this for in depth recommendation and what I've found, especially in experimenting, like with our search engine algorithms and whatnot, is that it's very important to not take just.

The content itself into account, but you need to take the listener and their preferences into account as well. And that's where I think that a lot of the sort of AI based dynamic Audi ad insertion may run into a little bit of trouble. Now, given your background, your expertise and where AI is today, what's the challenge that you've still got to overcome.

Is it the amount of data that you need? Is there better AI algorithm that you need? When do you pass the alluring test? When does the intelligence for the AI that I ask questions of forming, meaning to believing that it could be a human answering it. I think that there's a gradient there. You might be able to fool a 12 year old.

Can you fool a 50 year old, there's certainly a gradient in that test and you have to take a look at a lot of cases. I think we're right at the brink of basically passing the turn test. In fact, there's, the AI community has been buzzing about this recent news headline. I guess there's this engineer at Google who raised some flags about their Lambda model, claiming that he believes that it's beginning to exhibit signs of se.

And I've actually looked into, into this particular engineer and, and he's well qualified to make such claims. This is not somebody who's flippantly making these claims. I, I don't necessarily agree with him, but I can certainly empathize in, in certain ways with his position. And sentience is a fundamentally, a difficult thing to prove outright, but I think that it's, it may be an easier thing to disprove.

And I think that we can do that through prompt engineering, but given the way. These large language models are operating. We're almost right there. You can have a conversation even with something that's more publicly available, like G P T three. Yeah. And you might be hard pressed to tell that it's an AI.

It may seem to you just to be a very intelligent human being. So have we moved out of narrow AI into general AI in again, everything's like to what degree? I think there's some really interesting things, but what we're generally seeing in the world of AI is that in somewhat unsurprisingly, is that the more parameters you have, and, and when I say parameters, you can think of that in terms of like synaptic connections between neurons in your brain, as it turns.

With the right architecture, the more of these parameters you have new. Skills begin to emerge in these large language models. And a great example of this is a new model from Google called Palm. And they do provide some examples on their AI blog, but it's so fascinating. This thing is actually able to reason.

A chain it's reasoning together. I think that's SU super fascinating and it really raises the question. Okay. Right now we're looking at these 500 billion to 1.5 trillion parameter models. What does it look like when we start getting up into the 20 trillion or a parameter range? I think some very interesting cognitive functions will begin to emerge.

And I think what's most fascinating for me. Is what functions will emerge that human beings fundamentally don't have, which is a question, but when it comes to sentient, I think there's a huge difference between intelligence cognition, reasoning. Understanding sentience and ultimately consciousness. Last question, internationalization.

Obviously this is English language. One of the things that I know a lot of people ask for is when, or how will international language be part of this? If I wanted to listen to a Chinese podcast or a podcast out of India, or maybe Africa, could I use the AI still? Yeah. And. Kind of goes back to, to one of your questions about what are the real challenges.

So yeah, data is a challenge. Compute is a challenge. And, and really when you start to talk about internationalization data, is that challenge, right? It's most, there is a huge bias in the space for the English language. And when we take a look at things, I think the two languages that we would like to really go after are Spanish and Hindi since India is the, if you look at podcast, listening by country, I think it's like us, China and then India.

And, but the problem that we face is that there, there are some neural network models, which are trained to recognize and even translate from audio, into text multiple languages while that's the case that they tend to have lower accuracy. Than then models, which only have to deal with a single language.

And obviously it's because you're, you have a, we have a limited amount of compute in number of parameters these models can have, and you're, you're making it do more work with the same size brain, essentially. So one of the reasons why fathom. Works. It does. It's kinda, it's nascent it when it hits, it really hits it's mind blowing.

Part of the reason for that is we decided early on to use the most accurate, most cutting edge AI models possible. So when, and we set aside the compute problem and said, ah, we'll solve that later. We'll solve that further down the line. But in order to do that, all of the models that we use are single language.

And we now use over a, a dozen. I think, I think once we get everything we have outta R and D into production, it'll be O over, probably over a dozen neural networks to pull off the tricks that, that Fathom's able to do. So since those are all single model networks, it means that in order to expand into other languages, we basically have to train.

12 new neural networks all in, in another language. So there may be some ways around that, or even to enable some functionality, but maybe not everything. And we're also actively exploring those routes as well. Ken, thank you so much before you go. Can you please tell people where they can get a hold of the beater?

So you can just go to fathom, do FM and there's links there to download the app. We, you can download it on the iPhone. That's actually been out of beta for a while now. And we also have a desktop version and a mobile web version. Android is on its way most likely before the end of July. So we're very excited about that.

And then if you are a creator, you can go to hello dot fathom fem slash. Claim dash podcast. In fact, Sam, I'll give you the link for that one there to put in the show notes and you can claim your podcast with us right now. We're working on a lot of those creator flows. If you do claim your podcast with us, you will be the amongst the first to get access to all of the AI, uh, episode analysis that we are performing.

We have some really cool stuff I would love to get in the hands of creators. I think we're talking about here's 10. Interesting. Clips ready to share on social media. Autogenerated from our AI here's autogenerated chapters. Here are all the other similar episodes across the entire podcast O sphere to yours.

Based on the conversation you had, some of which could give you potential like future guest ideas. So we have a lot of interesting stuff in store, and by claiming your podcast, you will get access to all of that. And again, have your audience follow your podcast on fathom that will begin to kick off AI transcription and AI search indexing.

And if you link to fathom in your show notes, we will notice that, and you will be upgraded to kind of like our latest and greatest, most compute expensive AI. Ken Miller from fathom. I bumped into Paul block from, uh, fathom at, uh, podcast movement evolutions as well. Thrust a card into my hand. Uh, and, uh, yeah, so it's, um, you know, great to, uh, see them launching and great to, to, uh, see them doing some interesting things, to get more people, to have a listen to podcasts.

I think that's a good thing. Yeah. I mean, they are going to, uh, from what Kim was saying, they are looking at some. Deep technology to, uh, expose what's within the transcript a again, bringing it back round to Acast unfortunately, um, the conversational AI that Acast announced a couple of weeks ago. Um, I dunno how Acast are doing it specifically, but I mean, if they are using AI, they're on good on them, but.

This is exactly what fathom wants to do. It wants to use very deep AI technology. And they've got some very smart people there and some very deep pockets as well. So I think they will do very well with it. How they're gonna monetize it though. I'm still not clear. Yeah, no, indeed, indeed. So worthwhile keeping a watch.

On that now Edon research and NPR are going to present the smart audio report today. James, come on, tell me you've got a sneak peek. I do not have a sneak peek. Uh, it'll be, uh, later on, uh, today as we record this. Um, but, uh, what the smart audio report has done in the past is measure. Use of smart speakers, voice assistant technology, and so on and so forth.

Um, some new data that's, uh, just come out of the UK shows that 10% of all radio listening happens on a smart speaker, 10%, which is quite a thing. Uh, so, um, the question is, is that happening in the us is podcasting a big thing now in terms. Uh, smart speakers. Um, we should, uh, find out over the next couple of days with the smart audio report from Edison research and, uh, NPR and Edison research is also doing some work with, um, sounds profitable.

Um, they've put together what they call the first credible study of the profile of podcast creators in America. So who makes podcasts? What are they like? What are they. Uh, you can find out all of that. There's a, um, uh, free webinar that you can, uh, you can register, uh, to, uh, take part, uh, in, if you want to, uh, find out more about that.

It sounds profitable.com/the creators. Now for a little bit of quick news, uh, WordPress promise, you can automatically convert the text. Any post into a podcast using anchor the world's largest podcasting platform. It says that is part of Spotify. And you can do that at no cost to you, they say yes. So basically WordPress, uh, offering you the tools to fill anchor with more crap.

So that's great. A well done, WordPress. Um,  new version of, uh, watch OS for your apple watch, uh, will let you, uh, search for podcasts directly on your apple watch. Yay said absolutely. Nobody.  actually funny enough. I've been using my, uh, Android, uh, wear OS watch, uh, for the first time in a while I've been wearing a Gar in for the last year or so.

But I went back to the, uh, Android, uh, one and, um, it's reminded me of, uh, actually how useful it is to have, um, a little device on your wrist that, uh, you know, buzzes at you when, uh, you know, something goes hideously wrong on a website that you run. Uh, so that's been, uh, quite useful thing. Wow. Um, hopefully it doesn't buzz too much for you.

Um, Now in breaking news that you, you posted, uh, apple podcast is published details of how search in it app works. Um, one of the things they clarified is that ratings and reviews aren't factored into search results. I think you've said this consistently for about two years now, James. Well, I think we've said that ratings and reviews aren't factored into the chart.

Um, I think this is the first time that they've AC they've actually said that it's. Also factored into search results either so useful to know that, um, what they've also clarified is what the search function searches. So it only searches the show, name, channel name, and the episode title. And that in case you are ever wondering is why the episode titles to pot land are so long.

Um, so that we get lots of things into the episode title so that people can find us. Uh, easily. Um, so we actually did, uh, on, on pod news, Mike Stedman, and I did some research in July last year, looking at what, um, uh, things, different podcast apps search through, and we spotted back then that apple podcasts doesn't search the descriptions of podcasts.

It's either the show description or the episode one. Um, and, um, it's good to, uh, see that apple agree with us.  so, so that's a good thing. So, yeah. So if you are interested in, uh, getting your SEO right for your, uh, podcasts in the, uh, search in, in a podcast app, then, um, this new posting from apple podcast is well worth having a look@movingonrss.com.

Our friends have added automatic submission of pods to boom play. Now I can't say I've come across boom play. So tell me more. No, it's, it's a African, uh, podcast app. Uh, it's got over 50 million users. Apparently I had. Quick play with it, this, this, uh, morning and it looked, uh, you know, pretty good. Um, and so rss.com have, uh, added automated distribution into that from, you know, from, uh, all of those, um, podcasters there.

I think, uh, you know, it's very clear that, uh, podcasting is growing quite fast in many African countries. Um, and, uh, anything that makes life easier to get your podcast into these specific podcast apps that people use there. Which isn't likely to be apple podcasts and is much more likely to be, you know, more local stuff than, uh, all the better.

So, um, yes, I'm gonna have a closer look at, um, boom, play a little bit later and see if I can make sure that our shows are in there after being purchased by Spotify. At the end of last year, podcast, host ger is to close on August the 31st. Yes. Uh, everybody is, uh, being invited to migrate over to, uh, megaphone doesn't, uh, your radio station use, uh, wash.

Yeah. When I got on email from Rob Loha at about three in the morning, my time, uh, pinging me to tell me that ska was closing. I had a little bit of a hot sweat, right then thinking, oh my God, what do I need to do next? But Rob's been very kind and said he will personally make sure my radio station, uh, feeds and moved over.

And actually megaphone are offering a full 12 months, uh, for free on the platform during the migration. So actually a little bit of a bonus for us. Well, there you go. A little bit of, uh, a little bit of saved money. Uh, and there's an interesting piece from, uh, Jonas woot from Pacific content friend of the show.

He's been on the show before, uh, talking about the true measure of podcast success, which is not downloads it's attention time. That's what he reckons. Uh, and so it's, uh, it's an interesting piece that you should go and, uh, have a look at, I caught up with Dan Meisner at, uh, Pacific content. When I was in Canada a couple of weeks ago, and he did a great talk for Canadian music week around how he's managed to work out, um, how many Canadian podcasts are in the top charts in Canada.

And basically it turns out to be about 10%. So the rest is from the us and the UK. Uh, so Canadian podcasts not doing particularly well, he, he ended up, uh, showing a couple of categories that, uh, really work. Uh, so, um, you know, there are. Company worthwhile keeping an eye on, I think talking about people on the move next, James, uh, congratulations to Helen Arnold for starting a new position as head of UK podcast sales at Sony music entertainment.

Yeah. Uh, Sony music getting, uh, more into, uh, podcasting in the UK as well at Tony Mory. Uh, who is the nicest man in radio has been appointed bow media, audio UK's content director of digital platforms. Now this is a really intriguing move. He was basically the boss. Of magic radio, which is a large national radio, uh, station in the UK and absolute radio, which has 10 national radio stations in the UK.

Um, he's moving on from that to oversee the audio and visual content that consumers can access on Bauer owned digital platforms, uh, having followed Bauer for many years, um, Bauer appeared to always be short of a strategy, uh, for pretty well. Anything. So, uh, it does look as if there might be a strategy here.

Uh, so many congratulations, uh, Tony, and, uh, maybe who knows, maybe we might get him on the podcast at some, uh, point in the future. He is a very lovely man, as I may already have mentioned. And, uh, Matt Stewart has left ma media labs. Um, he is to take his jam street media back. He's gonna work with more companies.

He used to work for iHeartRadio a long while.  and he's a, a decent chap. And I saw him at, uh, podcast movement evolutions as well. Mm. Moving onto some tech stuff. Now, uh, Brian of London, one of the guys behind pod ping, uh, on the hive network has written a great post about how notifications ares. Bringing free of the regulated gatekeepers.

Have you had a read of that piece? Yes, he has. Yeah. And it's, uh, all about how pod ping works and how pod ping is essentially making life easy and simple for all kinds of things, including for, uh, live podcasts as well. Um, Adam and Dave on podcasting 2.0 last week, did a very good demonstration of, um, doing a live show and getting some live, um, Uh, feedback from their audience.

So, uh, they did a, a good job there pointing out how useful it is. So it's a great, uh, article, which is, uh, worthwhile having a look at. You've also written something about, uh, Firefox, uh, in here I noticed as well. Yeah. I just noticed something that came through. I mean, you know, occasionally you see stuff.

Uh, the Firefox has rolled out something called total cookie protection, and they've turned it on by default to all users worldwide. The reason I mention it is cuz it's the cookie Armageddon that's starting where, um, you know, I guess this is why people like Acast are looking at. AI conversations rather than looking at, uh, the individual user of a podcast, because at some point Google will and others turn on, uh, cookie blocking, I guess.

Uh, and, and this will mean that you won't have the same access to the same information, but it looks like five Fox have started. So, uh, The starting guide. Yeah. Five Fox is basically starting already, uh, blocking third party, uh, cookies. Google are getting rid of them in late 20, 23. It doesn't necessarily impact podcasting because we've never used cookies.

We don't use cookies for, um, any podcast analytics, but, you know, if, uh, given that we, we all. Um, uh, you know, websites as well. Uh, it's probably a good, uh, plan just to keep an eye on that pod news. Doesn't use cookies, uh, at all. So, um, uh, so there we are, I'm, I'm a bit militant in terms of what I allow onto that website.

So, uh, yeah, but, uh, good on, uh, five Fox for, uh, sticking that going. Mm. Now your favorite part of the week, James it's booster Graham corner. Now it's time for the boost boost boost boost it's boost to Graham corner. It's too much. Oh yes it is. And, uh, 1000 sat from Dave Jackson talking about nice people. Uh, he's a, he's a terribly nice person.

Uh, he's uh, sent us a thousand sets using castomatic and says, love the show. Thanks for the global coverage. Thank you, Dave Jackson, that's very kind. We should get Dave on one of these days. He's a, he's a top man. Or maybe when, when I'm away in a couple of weeks, time, maybe he might be your other co-presenter unless you've already filled those spots already.

No, I haven't yet. And we will be discussing that at the end of this show.  but let's move on. Oh, here we go. Um, yes. Well, thank you. Uh, so much, Dave. Uh, one of, one of the good guys never has a bad word to, uh, to say about anybody. Um, and so thank you for the sat. I appreciate that. Uh, thank you also for the 5,000 anonymous sat from, uh, somebody who was using, uh, Curio caster and all they left, uh, as a message was x-ray Yankee Zu.

Uh, so, uh, who knows who that might be? Todd Cochran, maybe, maybe Mike Dell, who knows? Yeah. Anyway, uh, thank you for those. And thank you for another person who just left a random amount of 1000 SATs and didn't even bother writing. Anonymous in, uh, curer, maybe one of the things that Curio caster could do is just remind you if you haven't set a, uh, a username to go and set a username prior to sending a boost, um, that would be a good plan, but thank you so much, uh, for, uh, doing that, even though the.

Uh, even though Bitcoin is down by 34%, uh, year on year, it's still a very important thing and a very useful thing, uh, for us. So if you like the show, uh, then please do, uh, hold down that post button as send us a booster Graham. And if you're not using a new podcast app yet, then you should be@newpodcastapps.com.

Now, uh, let's have a look at some of the events that are going on around the. Um, we, we keep hinting and reminding people, um, in August in Dallas, it's the next podcast movement. It is. And, uh, they've just made some speaker announcements actually. And Dallas Taylor from 20,000 Hertz, which is one of my favorite podcasts.

He will be talking about how they make the sausage. Uh, the sausage in this case, being that the, the podcast, not, not a, uh, you know, ground up meat, uh, Robert Riggs from true crime reporter will be talking about, uh, doing, uh, true crime, uh, reporting. Uh, he's a proper, proper reporter. I went out for dinner with him at.

Podcast movement to evolutions, and he was even holding one of those reporters' notebooks, you know, with the spiraly thing at the top and, and everything else. He's a proper old school reporter. Um, he will be very good to watch and also, uh, excellent news, Sam fireside chats, mark Cuban and Fallon for Timmy.

Our returning a for Temi are returning to speak about the power of web three and metaverse so I'll be sure to be, um, Going to that one.  I think the bouncer might have your name on the list not to let you inmate that's more likely. Yes. Well, I'm very much looking for great teeth. Great teeth. I'm very much looking forward, uh, to seeing mark Cuban and Fallon.

For Tammy for another 45 minutes of, uh, sales talk. Well, it looks like they're pivoted. Then if they're talking about now, their keywords are metaverse and web three. Uh, so I'm sure that they'll be selling NFTs as well as part of the, uh, fireside chat, uh, platform. Yeah. Well, apparently they're talking about the future of podcasting media and entertainment.

Uh, so. Very much looking forward to be sold to for 45 minutes. Um, but, uh, but there are plenty of good things to, uh, go and see a podcast movement, 20, 22 and mark Cuban. Um, and, uh, I will be, uh, there as well. So in case you, um, just like a podcast movement to evolutions, if you want to, um, uh, pin me against a wall and shout at me for 10 minutes about something that I may have written four months ago.

Then you've got another chance in Dallas in August and just point out he's taller than he looks, James. He is taller than he looks  indeed. Uh, yeah. So let's move on. Uh, international women's podcast festival, uh, is in London, uh, on, uh, Saturday, they've got a number of, uh, high profile speakers. Haven't they, uh, in partner.

Pinterest, uh, which is, uh, worthwhile doing, uh, the British podcast awards have, uh, nominations, um, uh, coming up on Monday, uh, which will be fun. And, uh, they have a, a thing, a new thing called grow, which is a specific event. If you have just started a podcast and you're looking at, uh, techniques to grow your podcast, that's in July, July the 22nd, you can find out more information at, uh, British podcast awards.

Do. Com I think it's dot com. Isn't it? Yes. Yes. It's probably.com. Uh, so you can find out more information about that. That's the day before the British podcast awards itself, that we will not be winning. No,  no, we won't win any awards for this thing because we didn't enter. That's one of the reasons, uh, so yes, there we go.

And the podcast awards, uh, which, uh, Todd Cochran works on are open for nominations from July. The first, uh, pod news was a nomination last year. Uh, so, uh, who knows we might be entering, uh, again, uh, but well worth, uh, entering too. So what's happening for you? This in, uh, pot land, Sam. Well, uh, I'm glad to say I finally installed my new roader PROI so they does it work.

It works very well. Yeah. Uh, all the lights came on the, uh, the firmware upgraded properly. Uh, we've plugged it all in and I'm doing final testing before it goes, live on the radio station. Yeah. Excellent. Well, very nice. Yes. Well, I'm looking forward to, I, I, I. Been using U using this mug in your, in your honor?

Uh, uh, Sam. Thank you. Yes. So, uh, I shared that photo forum wide. Yes. I, I posted a beautiful photograph of, of, uh, of a mug for Sam's radio station on the Twitter the other day. And, uh, yes, that was all. Good. Um, and also I've been enjoying, I was listening to a bit of, uh, the podcasting 2.0 podcast with, uh, Adam and, uh, Dave.

And it, uh, I very much enjoyed it, Sam. It started with this bit, they had an experience today that I can't get out of my head and it has to do with audio versus video. So this morning I, uh, I got a ping from Sam Sethy. He wanted to talk about something. So, uh, we hop on, uh, a video. Now I've been listening to pod land.

I've been listening to, uh, uh, Cridlin and Seth for, I don't know, two years maybe. And, and so we're talking and he start and he starts talking about something. And I, I say, I gotta interrupt you for a second. It, my brain cannot wrap around the voice. I've been hearing the image in my mind of what you look like and what you actually look like.

Yeah. He, he has visions of me being a blonde blue-eyed Brit or, or, or nerd. Yes. Yes. Yes. And that's not quite what's going on there, is it? No, it was one of the funniest things I've ever heard. I have to say  but thank you, Adam and Dave, so good on, then I'm looking forward to meeting, uh, Adam and Dave, who I've not yet met in, in person.

Uh, they, I gather will be at podcast movement as well. So that should be good. So James, what else has been happening for you this week in pod land? Well, I've been basically, uh, I've been enjoying being at home. Uh, I've been enjoying, uh, playing around. I did some, uh, fidling around with the QR codes on pod news, podcast pages the other day.

Uh, and, uh, you know, a few other little site enhancements as well. Uh, one of the things that I've just, just done, which is good for you, if you're looking for a brand new podcast to go and, uh, have listened to is a new. Podcast trailers, uh, feed, uh, which I've enhanced, uh, this, uh, afternoon. So you can easily get it on all of the, all of your favorite podcast apps, as long as it's not Spotify.

Um, pod news.net/trailers, uh, is where to go to find out more about that. But it's basically full of, uh, trailers for new shows that you. Be having a listen to, uh, so that was fun. Mm. And that's it for this week. If you like pod land, tell others to visit, you can tell your friends on Twitter, LinkedIn, Facebook, uh, MySpace, uh, or Bebe.

Yes. You can't do those last two. But you never know, they might come back. Uh, you can also email comments, pod land.news, or send us booster grounds that we love to read out on the show. You'll also find all our previous shows and interviews@podland.news. Yes. And if you want daily news, you should get pod news.

The news S free of pod news.net. The podcast can be found in your podcast, apple by asking your smart speaker sweetly. Uh, all of the stories we've discussed on pod land today are in the show notes and we use chapters and transcripts too, cuz Buzzsprout lets. Our music is from ignite jingles, and we're hosted and sponsored by our good friends, bus sprout and squad cast.

Keep listening.

Podcasts we love