October 30, 2024 • 43 mins
Disclosures are becoming increasingly important for podcasters, especially as the use of AI tools like Google Notebook becomes more prevalent. Dave Jackson and Daniel J. Lewis dive into the implications of using AI-generated content and the necessity of transparency for listeners. They explore various scenarios where disclosure is crucial, such as financial compensation and the potential for confusion over AI-generated voices. The discussion highlights the balance between utilizing AI efficiently and maintaining trust with the audience through clear communication. As the landscape of podcasting evolves, understanding what needs to be disclosed will help creators navigate ethical and legal considerations, ensuring that their listeners are well-informed.
Mentioned in This Episode
Disclosures 101 for Social Media Influencers
https://www.ftc.gov/business-guidance/resources/disclosures-101-social-media-influencers
Guides Concerning the Use of Endorsements and Testimonials in Advertising
Daniel's Proposal
https://github.com/Podcastindex-org/podcast-namespace/discussions/669
You MUST Disclose Whenever You’re Compensated! - The Audacity to Podcast
https://theaudacitytopodcast.com/you-must-disclose-whenever-youre-compensated/
FTC's Endorsement Guides: What People Are Askin
https://www.ftc.gov/business-guidance/resources/ftcs-endorsement-guides-what-people-are-asking
Your Hosts
Find Dave at schoolofpodcasting.com
Find Daniel at theaudacitytopodcast.com
New Podcast Apps
Listen to this podcast on any of the fine value-enabled app at www.newpodcastapps.com
Takeaways:
- The future of podcasting will likely require more clear disclosures regarding AI usage.
- Disclosures need to be prominent and cannot be relegated to the end of episodes.
- Listeners must know if content is AI-generated to make informed decisions about trust.
- AI can help repurpose content, but transparency about its use is essential for credibility.
- Financial disclosures are vital; podcasters should clearly communicate any compensation received.
- ...
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
Do we need disclosures in thefuture? This is the future of podcasting,
where we ponder what awaitsthe podcasters of today. From the
school of podcasting, here'sDave Jackson.
And from the Audacity topodcast, here's Daniel J. Lewis.
DANIEL Future of podcasting52. We're talking disclosures. So
(00:23):
we're gonna peel back thecurtain and talk about if you should
peel back the curtain a littlebit. And I think one of the things
I wanted to talk about, we'lljust jump right into it. I had somebody
that wanted to advertise on myshow called Ask the Podcast Coach.
I use Buzzsprout for thatparticular show. And I clicked on
it and I wanted to go hear theother show. And lo and behold, it
(00:46):
was everybody's favorite cohost and sidekick. I refer to them
as Kyle and Cheryl, betterknown as the Google Notebook twins.
Ooh, let's dive into that.
Let's do a deep dive. Let'sstrap in and talk about disclosures.
And I don't know why. I thinkbecause this person didn't disclose.
It was like, you know, theintro came on, and all of a sudden
(01:09):
here came Kyle and Cheryl. AndI was like, wait a minute, did I
miss a disclosure? And thefact that he was kind of not just
putting it out there. And Ijust thought, okay, I know these
two, and I know they're notreal, but I thought, but what if
I didn't, like, in the future,are Kyle and Cheryl going to win
a podcast award becausethey're so popular? I'm like, I don't
(01:30):
know. But I. It just. Itreally did rub me the wrong way.
I don't know what have you runinto this yet?
Not into AI stuff. I tend tostick with the same group of podcasts
I have heard. Well, I couldtake POD News Daily, for example,
with James Cridland. There wasa time when he was sick, lost his
(01:50):
voice, and he used AI to voicehis entire episode. And he disclosed
that at the beginning. And Icould understand now. It was obviously,
since we listened to James somuch, it could not replace him.
Right.
It sounded like AI to me. Icould tell other people might not,
especially if that was theonly thing they heard from him. Yeah,
(02:11):
they might not tell. Now, thebigger idea here, and there are some
interesting, specific thingsto get into, as well as something
I submitted as a proposal forpodcasting 2.0. But the bigger picture
here is what needs to bedisclosed and why. Right now, the
conversation is so much aroundthe use of AI but for years, we've
(02:33):
needed to disclose things inways that were compensated. That
is a big thing. So much thatthe FTC recently in the United States
came out with some newguidelines a couple of years ago
to say, here's when you needto disclose. Here are some ways you
need to disclose. You need tomake sure your disclosure is very
prominent. Talking aboutaffiliates or if you're compensated
(02:54):
in any way by talking aboutsomething or by including content.
And I did a whole episodeabout disclosures and such related
to that. But it has reallychanged from you can't just stick
something at the end of yourepisode or on the footer of your
website or on a separatedisclosures page on your website.
It needs to be prominentbecause. And this is the reason.
(03:15):
So think about this foranything along this line of disclosure,
because the person on theother side, the audience, needs that
information so they can takeit into consideration for how they
then take action or receivethe rest of the information. Because
if I'm being told by you thatyou're getting paid $1,000 per minute
(03:42):
to talk about the color blue,I'm going to think, well, wait a
minute. Does he really likethe color blue? $1,000aminute. That's
a lot. I mean, I talked aboutthe color blue for $50aminute. So
knowing that you'recompensated now, you don't have to
disclose how much you'recompensated, but it's that thing
(04:02):
that's giving me informationto make my own decision. So it's
putting the responsibilitythen in my hands. Because now that
I know that you are beingcompensated for talking about the
color blue, I can decide ifI'm going to believe you or not.
And then you can also do agood job of proving whether you're
(04:22):
saying good things onlybecause you're being compensated
or if you are just a shill forit because you're just. That's your
sponsor for the moment.
Well, I always thought it wasfunny when I would see articles about
podcast media hosting, and theonly people they reviewed just so
happened to be those companiesthat had affiliate programs. It was
(04:43):
like, wait, you're not goingto talk about this? Coming up. Just
Captivate, Buzzsprout,Blueberry, you know, Libsyn, that
was it. And that was like, oh,that's kind of odd that they're not.
Because there are other onesthey didn't talk about Spotify, didn't
talk about Red Circle or anyof the other ones that are out there.
And I was like, oh, I seewhat's going on here. These all have
Affiliate programs. And thatmay have been, you know, why they
(05:05):
chose those people to do that.I know for a while there were so
many websites, they'reprobably still out there talking
about web hosting.
Oh, yes, yes.
Because I remember when I wasmaking I think 50 bucks from Hostgator
back in the day, you know, youget a couple people sign up for that,
like, all right, dinner's onDave tonight. That's cool. Yeah,
the FTC actually had. It wasftc.gov I think, influencer for a
(05:30):
while. There was an actuallike video there and they explained
how it has to be at thebeginning. So like all my newsletters
now start off with my Amazon.I may earn whatever, you know, the
disclaimer for that. Andthey're saying in the, you know,
if you're doing videos andlike, hey, today's show is brought
to you by Dr. Rob. Blah, blah,blah. By the way, today's guest is,
(05:52):
you guessed it, Dr. Rob. Like,you have to say that up front. And
I just think the young kidsback in the day on MTV had a very
popular show called Catfish,which I never quite. I guess this
is a real thing where peopleget online and you think you found
the person of your dreams,only to find out that it's some 14
year old kid that's having aprank. And I'm like, well, in a way,
(06:16):
if you don't disclose that I'mnot real, that's catfishing. And
so somebody. Again, that's tome. And I think the difference is
the disclosure. Because look,we all watch TV and I know that Monica
Geller and Rachel Green andeverybody else on Friends, they're
not real. They're playingcharacters. But that's the whole
point. I know they're playinga character. It's weird. You kind
(06:38):
of. That's why you get sadwhen somebody dies or whatever, because
you somehow feel like youknow, this person that doesn't really
exist, it's a character. Butif you don't know that, then you
just feel like you're, youknow. Wow, I love. This is one of
my favorite podcasts. I just,I identify so much with this person.
Then you find out later it'snot. If you want to get on my bad
side, lie to me. Because youdon't have to lie to me. Like, I
(07:01):
can take lots of criticism.You don't. There's no reason to not
just tell me the truth. And ifyou don't and I find out, it's hard
for me to trust you again. Somaybe that's the other thing. Too.
I just have a. I am sensitiveto people lying to me, I guess.
Well, and especially in thisspace with AI taking over for so
much, or rather people givingover control to AI. I mentioned this.
(07:25):
I was a guest on thepodcasting 2.0 show recently with
Dave and Adam. And I broughtthis point up too. And I've mentioned
this in my own podcast thatthere's so much. I love Adam's term
for this. So I'm going to helpspread this. There's so much AI sloppy
out there that when you sharea genuine opinion, to me, it stands
(07:47):
out. The web hosting thing. Iknow that in the past I've looked
for things and Blue Host isalmost always near the top because
they pay out the biggestaffiliate program. And if you get
on YouTube, like looking upfor product reviews or something
like that, you'll typicallysee like this AI slop. That is, they
just found the top ratedproducts on Amazon and that's the
(08:09):
order of theirrecommendations. And they have the
audacity to say, in thesevideos, we tested the top 10. No,
you did not test, you liar.You did not test these things. You
just looked at reviews. Whichis actually. Some of. This is actually
a violation of the Amazonterms of service to run their product
(08:29):
pages through an AI in orderto promote your affiliate link. That
is a violation of their termsof service. But that aside, that
kind of stuff, there's so muchof it out there that a couple of
products that I've looked for,when I saw someone using the product
themselves, I see this forkitchen gadgets, for example. I enjoy
(08:50):
being in the kitchen. Therewas something. I was looking up all
this AI slop out there and Ifound a guy actually using the product.
His camera quality, his videoquality was okay, his audio quality
was okay. I think he mighthave been wearing a LAV mic, or maybe
not. But I could hear him andI could understand him enough. He
(09:10):
didn't have studio lighting,he wasn't in a studio, but he was
genuine, authentic and I couldunderstand him. So his quality was
good enough. And that sold meon the product because of all the
AI slop. I just wanted agenuine opinion. And if in those
videos, if they would say, thefollowing is generated by AI or the
(09:35):
following is a list of the topmost rated products. Like, they don't
always have to say, we didn'tactually review these products, although
that would be good for them todo. They could put it in more positive
language to say, this is aoverview of the most popular products.
That's better. And that Equipsme to make a better decision. Same
(09:57):
thing. Like what you weresaying is when they don't disclose
in that way, and I realizethis is just more AI Slop, then I
get upset, I thumb it down,and I leave. And I'm so tempted to
just comment on there, like,yet another AI Slop video, but I
try to be nice on the YouTubecomments, since everyone else isn't.
(10:17):
Well, we've all seen itbecause they have found some stock
images of the product,probably from their website, and
it's always slowly panned orslowly zoomed in because they don't
have any video and they'retrying to do. And then it's just
this AI Voice that isexplaining. And I've seen some that
are. Just before AI was evenremotely passable, that was just.
(10:39):
It was very robotic. As itexplains, the New Juicer 2000 will
do this. And you're like, holycow, is this horrible? And then it
dawns on you that this may bedone by somebody where English is
not their first language.They're just, like you said, they're
just trying to get thataffiliate commission, and they're
looking at what's popular. Andhere we go. So now to flip the coin.
(11:01):
I was doing a show, and JodyKringle, who's a voiceover artist,
who, as you might imagine,might have her eye on exactly what
AI Is doing. And we weretalking about people using Google
Notebook to do a show, and shesaid, if you don't want to do the
show, like, just don't do theshow. And when she said that, that
(11:23):
triggered it, because there'sa show I do that. I honestly don't
want to do the show. And it'sabout Akron, Ohio. It's called the
Akron Podcast. And I was like,ooh, that I wonder if this would
work. So I created an episode.And what was interesting, it was
a very sensitive subject inAkron, Ohio. Our police have a bad
habit of shooting people whenthey shouldn't, especially white
(11:44):
cops shooting people of color.And the actual people of Akron, Ohio,
now have to pay $4.8 million,not an insurance company, because
we shot a guy 46 times. Iguess the 45th, not enough or whatever.
But anyway, so I thought,well, this will be interesting because
it's a sensitive subject andAI is going to describe it. So I
(12:07):
was like, okay. And then therewas another. The second story was
yet another shooting that Ihad Kyle and Cheryl talk about it.
But I disclosed up front. Andafter the first story, I said, hey,
I have another story aboutthis Poor kid got shot because he
had a fake gun and he waspointing it at people. And I said,
so let's hear what Kyle andCheryl have to say about it. And
(12:29):
off they went. So I justexplained up front. I'm like, hey,
these are. I said, now I'velistened to what they had to say
and I've checked for accuracy.I said, but here you go. So I just
put that out today. It's avery small show because I. Again,
I don't really want to do thispodcast at this point. It's just
a test show for me, but I'llbe interested to see if I get any
(12:49):
kind of listenership. Theinteresting thing about Google Notebook
is in the past, if I hadthought about doing this, I would
have ran the story from thelocal newspaper, had some sort of
AI rewrite it again, check foraccuracy, and then throw that script
into Google Notebook. Well,Google Notebook kind of does that
automatically. You can copyand paste in text and then it just
(13:12):
straps in and does a deep diveas they delve into whatever. So it'll
be interesting to see. But Ido feel somewhat of a hypocrite that
I. But I'm like, no, no, I'mdisclosing. I'm not a hypocrite.
And there is a place for someof that. Not from. Don't think of
it completely as slop, becauseyou are taking actual content and
(13:34):
you're using AI to repurposeit. What I would call AI slop is
when you are creating thecontent almost from scratch or the
opinions almost from scratchwith AI, that is totally slop. But
if you're repurposing, I thinkAI is fantastic at repurposing. But
here's the thing. The point Iwant to make here, you are making
(13:56):
content more available topeople who want it in that format.
So, like, if I get to awebsite and I see a really long article
and they have an option or mybrowser has an option and plugin
or whatever to give me an AIgenerated summary. I'm clicking that
sometimes to see that.Actually, I'm using an email app
(14:16):
right now called Short Wave. Ido have an affiliate link, but I'm
not talking about them becauseI have an affiliate link. But they
have this neat feature that Ican press a button on an email and
it will summarize that emailor even an entire reply thread of
emails. It can summarize itfor me or given me action points
from it, or I can selectmultiple separate messages in my
(14:37):
inbox and have it summarizeall of those together. That's great.
That is saving me time. It'sgiving me information in a format
that's more digestible for me.Or you visit a website that also
has a lawn article, and youwant every word of that article.
You just don't want to read ityourself because you need to be copying
(14:58):
and placing or clicking thingsand doing other stuff like that.
But you want something tolisten to. So, I mean, that's. The
podcast solution is if there'sa button to turn that article into
a podcast where you can hearthe article read to you. If it's
an AI voice, I don't thinkthat needs to be disclosed because
it's just a voice. The sameway Mignon Fogarty with Grammar Girl
(15:23):
does this in her podcast,where so often she has guest writers
who contributed content, andshe discloses that she says, the
following is written by so andso. So when I say me, it's referring
to him or her. So then shestill reads it. And she reads it
(15:44):
in her genuine voice and withher own humor and inflection and
all of that to it, but it issomeone else's content. She's reading
it. I think that's disclosureenough. If an AI is simply reading
content that you've generatedyourself, not with AI, but you've
made it yourself, that mightnot need to be disclosed necessarily,
(16:05):
unless the AI is identifyingitself. So here's the question to
think about is where is theline of, at least with AI, where
you do and don't need todisclose your use of it?
That's a great question. Ithink you hit one of the points,
I think, are reviews. BecauseI remember one time I had a client
(16:29):
and I asked him, would youhave any gear yet? And he said, yeah,
I have a Podtrak P4. And thenI bought a Rodecaster back when the
first one came out. And Isaid, why do you have both of those?
And he said, well, when youtalked about the Rodecaster, you
said it was cool. And I waslike, wait. So he spent $500 because
I said something was cool? AndI was like, well, it is cool. I'm
(16:51):
like, but if you got the otherone, you kind of don't need this
one, or vice versa. And so Ithink that's why people like Tom
Buck on YouTube and AndrewScott, they always disclose up front
whether or not, like, hey, I'mgoing to talk about the new sure
Move Mike. And, like, Bandra.I just saw his video on that. And
he was like, hey, just so youknow, this is on loan from shore.
(17:13):
And then they always say, I'mgoing to give you my honest opinion
one way or another. And that'swhere it's always, for me, I'll listen
to it. And I always love thatboth of those guys were like, this
is what we really liked. Butthere is one thing, and the minute
they say something remotelynegative, I'm like, okay, I'm going
to believe this. Now, ofcourse, I always say you start off
(17:33):
with two things when you starta podcast. No audience and integrity.
And so the minute I find outthat you're lying, we're done. But
Tom Buck actually has a. Itsounds weird because he's. I look
at. Tommy's a YouTuber, but heactually has a podcast. Podcast.
It's an audio podcast. And hedid an episode and he was talking
about how companies have cometo him and go, hey, can you do a
(17:56):
review of our product? And youknow we're going to give it to you
for free, but if you could notdisclose that. And I'm like that.
It's just so weird that thereare companies going, yeah, can we
kind of keep that on thedownload? I'm like, well, it's a
little thing we like to callillegal, you know, But I'm sure there
are people like, man, I'mgoing to get free gear.
(18:17):
Sure.
And I'm like, oh, that's so bad.
I kind of feel like the lineis similar to a trademark where in
trademark stuff. And I've gonethrough this, and I'm going through
this with a trademark disputeright now. It's a matter of potential
confusion. Might someone thinkthat AI generated image is real?
(18:38):
If so, it needs to bedisclosed that it was AI generated?
Might someone think that thatvoice is a real person saying those
real things, giving their ownopinion that might need to be disclosed.
If it's simply a voice doingsomething for you, like reading something
(18:59):
for you, just so you have itin a different voice. And you've
never said, here's the voiceof so and so, or anything like that
that might not need to bedisclosed. And if you're repurposing
your own content that youalready created, like you, maybe
you used AI to generate abunch of social posts for you or
to summarize it. I don't thinkyou need to disclose the following
(19:21):
summary was generated withartificial intelligence. That's a
waste. I don't think you needto do that. But when there is the
potential for confusion, orespecially if you are trusting the
AI, blindly trusting the AI,then you definitely should Disclose
it. The illustration that Ilike to use in one of my episodes
(19:44):
about transcripts. If you useAI to do a transcript, you need to
be careful with certain thingsbecause sometimes the AI or the transcript
engine won't hear somethingcorrectly. And there are certain
cases where that could lead tomajor problems. An example I like
is you might say, I love twolittle puppies. Or you might slur
(20:08):
it, saying, I love two littlepuppies. The AI might hear that and
it thinks you said, I love tokill puppies.
Right.
And that can lead you to legalhot water. So if you don't audit
and edit, you could get in trouble.
Yeah, I'm with you. I thinkthe biggest mistake is a having it
(20:30):
generate the content. I alwayslike the way you said it was assisted
intelligence. So have it brushup on the content that you created
with and then from. I thinkthat's the best use of AI And. But
I also like your point. Like,for me, when I think about it, it's
just really just reading thefacts that I gave it. And so it's
(20:51):
not really throwing inopinion. And I was really glad to
see that both Kyle and Cherylthought killing people was bad. I
was like, yay, Kyle andCheryl. So that was good. So I think
that it's. When you, like yousaid, when you get into opinions,
and if somebody's going totake action on your words, it would
be good to know if it's someAI tool.
(21:11):
Yeah. With AI especially, itcan hallucinate, as they say, and
so either it can get somethingwrong, it heard something wrong,
it could get a fact wrong, itcan make up things. Like I loved
hearing in the no Agendapodcast recently, where I think it
was John C. Dvorak did this,where he took Notebook LM from Google
(21:31):
and he ran something throughit and then he got the transcript
and he ran that back throughNotebook LM again. I can't remember
how many times he did that.
It was twice.
Only twice? Yeah. And ended upwith some definitely hallucinated
information in there. It'slike playing the game of telephone.
This was about when the pagers exploded.
Oh, yeah.
(21:51):
And so they took that and hadit transcribed and then ran it through
back again. The final outputafter basically running it through
itself twice was they weretalking about exploding electronic
devices. And it wasn't verylike they lost some of the specifics.
And there was one. I don'tknow if they called it a Walkman.
(22:13):
Yeah. But it was something.It's like, oh, now we're in the realm
of not even close. And that'swhy Adam is all about, yeah, just
give us more AI slop. Becauseit's just going to be bad. And anybody
who does more than clicking abutton is going to stand out. So
I tend to agree with thatopinion because I've. I've just heard
some things. I'm like, ooh,that's. That's not good.
(22:35):
But then you look at thingslike, if you use an AI to generate
a title for you, I don't thinkyou need to disclose that. If your
AI made the chapters for you,I don't think you need to disclose
that. There are lots of thingsthat AI can do that I don't think
you need to disclose. And,like, also, you know, people are
using AI image makers to maketheir podcast cover art. I think
(22:59):
that's fine. You don't have todisclose that. That's AI. This is
all opinion, by the way. So adisclaimer here. We are not lawyers.
We have certain experienceanyways, but we are not actual lawyers.
So this is not legal advice.But my impression is that these types
of things, if it's obviousit's not real, you don't have to
(23:19):
disclose it. It's just like, Idon't have to say whether I used
Adobe Photoshop or GIMP if Iedited an image, but if I make that
image look real and it'ssomething obviously fake, obviously
not real, then I shoulddisclose that that is a parody in
some way. You look at, like,the Babylon Bee, one of my favorite
(23:42):
sites on the Internet rightnow, hilarious stuff comes from the
Babylon Bee. And they have allof these funny images that they make
with their articles, and it isobviously satire. So they don't have
to say this image in thisarticle is satire, because that's
what it is. That's their wholebrand. Same thing with the Onion.
(24:04):
And there is a case right nowwith, I think it's gone all the way
up to the Supreme Court. Andthe Babylon Bee has been involved,
The Onion has been involved.But, like, defending the right to
do parody because some peopleare saying someone might get misled
by this idea that thishappened. Or there's that. The example
I'm thinking of, there's theone where Jesus flipping over the
(24:27):
tables because there's anaccount in Scripture of when Jesus
cast out the money changers inthe temple, and he said, you turned
the temple of God into a denof thieves. And so he flipped over,
like, pushed over the tablesand cast them out. But then what
did the AI make?
It's a big picture of JesusBasically doing a backflip over a
table.
(24:48):
So it's like the AI understoodyour words, but it didn't understand
the idea. But if you didsomething like that, you know, pick
a politician, any politician.Politician steals candy from a baby,
here's photographic proofthat's going to get you in trouble.
Yeah. And I think it goes backto what you said earlier about confusion.
(25:09):
So when you make. I know, Ithink it was Taylor Swift, but they're
insert any female celebrity.And the minute you get popular, it's
probably a rite of passage,which is just sad that the minute
you get popular, nakedpictures of you are going to pop
up. And it's weird because Ithink everybody knows they're fake,
but it's naked people. So. Andit's just. So that's the again where
(25:33):
you get into trouble, whereyou're like, oh, this actually looks
real because he or she is on aboat and this might be that. And
you know, that's when you'regoing to get a trouble.
(26:00):
The trouble is more than justthe pixels. It's the ideas, it's
the connotations, it's thefeelings, the other things beyond
just the pixels themselves.And bringing it back to other kinds
of disclosures, too. Thesedays with financial disclosures,
whether you're beingcompensated for something or you
(26:21):
have some kind of associationwith something, like while I was
on the advisory board forCaptivate, I disclosed, whenever
I was talking about Captivate,I would disclose. I am an affiliate
and I'm also on the advisoryboard. And you have done this a lot,
Dave, while you were withLibsyn, whenever you would talk about
media hosting providers,whether positive or negative, you
(26:43):
would disclose that you workfor Libsyn. And yet the way you've
disclosed that and the waythat you talk about other things,
I think boosted yourreputation. So it didn't make people
think, oh, Dave is just sayingthis bad thing about them because
he works for Libsyn, or he'sjust saying this good thing because
he works for Libsyn. Butbecause you were honest, because
(27:04):
you shared the good and thebad on both sides of whatever you
were talking about, that madethe disclosure boost your reputation.
And I think we're going to seein the future more disclosures required.
I think that for sure, the useof AI, there's going to be some need
for disclosures in that from alegal perspective, I think that's
(27:26):
coming in probably the nextcouple of years. There will be some
kind of guideline or a lawthat you'll need to disclose. But
it will probably be somewherealong those lines of when someone
might confuse it or it mightmislead someone or something like
that. We also have the lawsalready about financial disclosures.
(27:47):
If you're being compensatedabout something and even if you're
not in the United States. TheFTC guidelines are interesting in
that they say basically, ifyou do business with anyone that's
in the United States or havethe potential to profit from anyone
in the United States, youstill need to disclose. But I think
they're good guidelinesanyway, and we'll have a link to
them in the notes so you canlook over them even if you're not
(28:09):
in the United States. Butthere isn't a completely good way,
a standard way to disclosethat works every time. And that's
why one of the things that Iproposed just recently in the podcast
Namespace Talk in Podcasting2.0 is a podcast disclosure tag that
(28:29):
would allow you to add thesedisclosures in your RSS feed. Now,
that will not be enough rightnow. What that would do is for the
apps that would someday readthat when that standard is accepted
and finalized and everythingif it is, that would allow the apps
to support that, to make thosedisclosures more prominent. And there
(28:50):
could be certain ways that youcould make that especially more prominent
and contextually relevant.Like maybe you could tie a disclosure
to a chapter. So while you aretalking about a product that you
earn commissions from, there'sa disclosure or during an ad, even
there's it says it in thechapter and such. This is an ad.
(29:12):
Now, ads, I think, are obviousthat they're advertisements. Most
of the time it's obvious. Andif it's not obvious in your own podcast,
then you need to make itobvious that it is an ad. You are
being paid to talk about thatthing. So you need to make that obvious
in some way. And there areproduction ways to do that or things
that you can say to do that tomake that obvious. But I think that
(29:33):
extra context can help people.And yes, this does open up the door
for systems someday in thefuture to detect when and where the
ads are in your episode. Butthat's going to happen anyway. That
is happening anyway. And in away that's actually going to help
you. Because as technology isgoing in the direction of being able
to detect when people hearads, if they're going to skip the
(29:57):
ad anyway, the sponsorshouldn't have to pay for it because
they're not getting whatthey're paying for. They're paying
for the actual impressions. Sothat could make Realistic impressions.
Even more valuable, I wouldhope, but. So I've proposed this
podcast disclosure tag thatit's a very rough proposal right
now. I'd love some comments onit from anyone who just sees it and
(30:18):
has some ideas of differentways that could be implemented. But
I tried to build it aroundthis idea of different types of disclosures
as well as different contextsfor those disclosures. So in my own
example that I gave in thecode, I gave the examples of three
different types of disclosure.Compensation, AI and association.
(30:40):
AI is obvious. It's if you'reusing AI for something that needs
to be disclosed. Compensationis if you're compensated in any way
for that thing, financialcompensation, you're earning points,
you were given something, youare paid, it's a sponsored ad or
something like that. Any kindof compensation would fit under that
kind of disclosure. Andassociation would be the thing where
(31:03):
you are associated like you'reemployed, you own stock, you're on
the advisory board, youfounded the company, anything like
that, you're related to thisperson, you're married to this person,
or they're your brother,sister, son, daughter, whatever.
Any kind of association likethat. Those are some types that,
that I thought of in theexamples I was making. And then you
(31:25):
could also think about addinga context to these which I put in
the code examples so you couldclarify where this is relevant for
compensation. The contextcould be links and in my example
I wrote some of the productsslash services in this episode use
affiliate links thatcompensate the podcaster for qualifying
(31:45):
purchases. It could be evenlike a context of content and that
could refer to the entirecontent of your episode or a portion
of content. Like if you have,you know, the tip of the day is brought
to you in part by orange juicedrink some today. Well, that could
be in the disclosure is thatyou are compensated for that piece
(32:06):
of content. Or it could bejust outright an ad is the context.
It could also be certainthings like the images that were
in there and to disclose thisepisode uses realistic looking AI
generated images, or thisepisode contains parody, anything
like that. These aredisclosures, not disclaimers. Disclaimer
(32:28):
is when you're saying I'm nota lawyer, this is not legal advice,
this is not financial advice,that's a disclaimer. But a disclosure
is if you are a lawyer or youneed to disclose that or you're related
to something, you're beingcompensated or in certain cases if
you're using tools, that kindof thing needs to be disclosed. Certain
(32:49):
kinds of tools like AI is whatI'm Talking about, or maybe if you're
using a product and talkingabout a product that was given to
you by the company in hopesthat you would do like right now
I'm podcasting with an electrovoice RE320 that goes into a original
first generation Rodecast Pro.Both the microphone and the Rodecaster
(33:12):
Pro were given to me by theirrespective companies. They were given
to me with no expectation thatI would do a review. They were given
to me out of appreciationbecause I already said nice things
about their products. And sothey said, hey, we want to give Daniel
one of our products so he canhave one. And when I talk about them,
especially if I talk about myuse of them, I try to make sure that
(33:35):
I disclose these were given tome. Nonetheless, I recommend things
that I truly believe inregardless of earnings. That's my
little thing that I say somuch. Now, it's a text expander snippet,
but this technical proposalfor a disclosure tag I think could
be great because then there'sless chance for someone to miss the
(33:55):
disclosure. And that's a bigimportant thing on the legal aspect
is the disclosures have to be prominent.
Well, that would be it becauseI mean, to play devil's advocate,
right, we already have anexplicit tag and that is completely
self regulated. And ifsomebody hears a show that should
(34:16):
obviously be explicit, you canreport that to Apple. The difference
is we all know what explicitsounds like. Not everybody may be
able to go, oh, that's Kyleand Cheryl, that's AI. And so that's
going to be the tricky part iskind of policing those people that
don't follow the rules. Butit'll be interesting because I know
already from the very firsttime I heard Google Notebook to now,
(34:40):
that's already better than itused to be. They breathe now they
occasionally chuckle. It'sinteresting how they've boosted that.
I'm sure. I know you can nowgo in and tell them the angle, what
you want them to talk about.And eventually you're going to be
able to pick somebody besidesKyle and Cheryl because not everybody
sounds like Kyle and Cheryl.And who knows, maybe someday you'll
(35:02):
be able to, you know, thiswhole show will be AI, just me and
Daniel and we'll say here makemake Dave kind of tired in this episode,
make Daniel super bounce offthe walls kind of thing. It'll be
interesting to see.
So when it comes to likecertain things, I know some people
have said, what does thismean? Like if I use AI Tic Descript,
for example, one of thebeautiful features that Descript
(35:24):
and a couple other tools offeris the overdubbing feature, as some
call it, or certain thingslike that, where you can use an AI
that's trained on your voiceto fix something that you said. I
think that's okay. That youprobably don't have to disclose it
because it was a mistake.You're simply correcting it. You
(35:44):
could do the same thing by rerecording it. Yeah, it might not
sound as smooth, but if you'revery skilled at what you do or if
you time it right, you canmake it sound as smooth. But that
doesn't have to be disclosed.You're not changing the content,
you're not misleading someonewith it. But if you were to use the
(36:07):
AI to make someone saysomething different, that's when
it gets into trouble. And evenwhen it comes to, like, the Kyle
and Cheryl voices, if you'resimply using them on your own content
and you fully edited thescript and all of that, should you
still disclose, probablybecause of this. Where else do Kyle
(36:30):
and Cheryl appear? And whatkind of impression is that going
to give the people who hearKyle and Cheryl, will they start
thinking, oh, man, this Kyleand Cheryl, they do great podcasts.
I want to find more podcastsby them. I hope I can meet them someday.
And then you discover they'refake. They were fake all along.
(36:51):
They're on the Wiccan show.
Yeah.
It's like, wait, what?Cheryl's a witch. I didn't know that.
Yeah. So that could be alittle bit of a problem in a weird,
spooky kind of way. It'll befun to watch and see where it goes
and see which lines it cross.And I think I know james cridlin@podnews.net
somebody. Oh, it was ListenNotes. Our good friends at Listen
(37:12):
Notes had come up with somesort of tool. And I want to say it
was already. It was either 208or 802. My brain's going dyslexic
on this. But either way, itwas A lot of shows were already using
Google Notebook and I waslike, oh, well, this is not unique
at all. So it'll beinteresting to see if they win Podcaster
of the Year next year. They'relike, these guys are productive.
(37:35):
They're, you know, I've doneprobably somewhere between 20 and
30 podcasts and tried and thenshut them down. These guys are on
70 shows and they're stillgoing. They're all still current.
Yeah.
And you mentioned some statsjust now. I think certain stuff around
that needs to be disclosed.More in the future, too, because
we are seeing so much, Ithink, especially more in the recent
(37:56):
years than before. But howpeople can lie with statistics and
the way that you tell thestory around the data. Sometimes
just the way a question isworded in a survey can affect how
people answer it. So I thinkwhen it comes to, like, survey data,
(38:19):
people should disclose whatthe exact question was and the exact
answers that were given, andthen you can summarize it from there,
you can abbreviate it orwhatever. But I think that's really
helpful to know because Isometimes call myself Mr. Caveat.
I start thinking of all ofthese caveats to things, and people
know that from the Audacity topodcast, I give lots of caveats and
(38:41):
alternatives and all of thatkind of stuff. That's kind of why
I started the show, because Ifelt like, hey, wait, wait, wait,
there's a caveat there to whatyou're not covering this caveat.
Some of that stuff you mightneed to disclose from content perspective.
So don't just think of it asfinancial or AI and that kind of
thing, but think about it fromthis perspective. And this is really,
(39:03):
I think, the crux of all ofthis. What does the audience need
to know that will help themmake a better decision of how to
process the content or takeaction? What is it that they need
to know? What is that extracontext that they need? That's the
kind of thing that adisclosure should cover.
(39:24):
Yeah. On a. An article onlistennotes.com they said just last
weekend. So this came out atthe beginning of the month. So it's
been more than that now. Butthey'd already deleted 500 fake shows
that were created withNotebook LM. I may have to realize
that, yes, I'm playing and I'mexperimenting, but the Akron podcast
(39:45):
may lose our listen noteslisted because I've done this. So
we shall see.
And I'm not so sure there's aplace to just cut things out because
they're AI. Because again, itcould be that by being an AI voiceover,
it is truly making contentaccessible, both from the perspective
of just, I don't have the timeto sit down and read this. I'd rather
(40:07):
listen to it. But alsoaccessible in the accessibility perspective
of someone who's visuallyimpaired, wants to listen, has to
listen to the article, read tothem, and maybe they just really
do not want to hear thearticle verbatim. They could retain
the information and takeaction on it. So much better if they
(40:29):
hear a conversation around thecontent. Like, that's what Notebook
LM is doing is it's turningbasically monologue content into
a dialogue that isn't usuallyadding much to it in terms of value.
It is adding fluff, we couldcall it. But it is fluff that is
more personable and because ofthat can make it more relatable,
(40:54):
more actionable for some people.
Yeah. And the part that's kindof scary is many podcasts with actual
people on them take reallywide tangents. And to their credit,
Google Notebook is somewhatstaying on topic. There are some
things we might actually learnfrom this. So maybe I don't care
(41:16):
that today is the two yearanniversary that you got your dog.
That's a real thing. Ilistened to six minutes of a show
this week and they, theyexplained what they were drinking.
They talked about theanniversary of the dog. I forget
what else, but I was justlike, they didn't. They barely announced
what the show was. So thereare things that maybe we could learn
from getting to the point andyou know, just the facts. Maybe so
(41:40):
we shall see.
And speaking of disclosures,we received a thousand sats from
someone in a booster grammar.So there's your disclosure for that.
Dave, what did we receive?
Yeah, from Creativity Foundcame a bit late to this episode holiday
listening and liked thediscussion around what or more differently
you can do with a podcastformat. So a your back catalog still
(42:04):
works. So keep that in mind.They continue on. I've planned a
step by step course formataround the features of podcasting
2.0. Yay. With explainerepisodes where guests tell me what
the thing is and why it'sgood. Then practical episodes teaching
the listener how to do thething. It's still a talk show, but
(42:26):
we'll use the 2.0 features toteach the 2.0 features inspired by
books I've worked on as a texteditor. And that is from Claire.
So thank you, Claire. That'sawesome. Very meta.
Yeah. And we both got tocontribute to her series as well,
so we might have a link tothat in the notes. I can't, I don't
know if it's published yet. Ithink it's publishing around the
(42:49):
same time this episode ispublishing, so we might be able to
link to that. Some of ourcontributions, like I got to talk
about chapters there andothers have been involved as well.
So love what Claire is doing.Thank you for those stats.
Yeah, thank you very much. AndI think that is going to do it for
this episode. Episode 52 is inthe can again.
(43:12):
Our websitefutureofpodcasting.net keep boosting
and keep podcasting.
Do we need disclosures in thefuture? This is the future of podcasting,
where we ponder what awaitsthe podcasters of today. From the
school of podcasting, here'sDave Jackson.
And from the Audacity topodcast, here's Daniel J. Lewis.
DANIEL Future of podcasting52. We're talking disclosures. So
(00:23):
we're gonna peel back thecurtain and talk about if you should
peel back the curtain a littlebit. And I think one of the things
I wanted to talk about, we'lljust jump right into it. I had somebody
that wanted to advertise on myshow called Ask the Podcast Coach.
I use Buzzsprout for thatparticular show. And I clicked on
it and I wanted to go hear theother show. And lo and behold, it
(00:46):
was everybody's favorite cohost and sidekick. I refer to them
as Kyle and Cheryl, betterknown as the Google Notebook twins.
Ooh, let's dive into that.
Let's do a deep dive. Let'sstrap in and talk about disclosures.
And I don't know why. I thinkbecause this person didn't disclose.
It was like, you know, theintro came on, and all of a sudden
(01:09):
here came Kyle and Cheryl. AndI was like, wait a minute, did I
miss a disclosure? And thefact that he was kind of not just
putting it out there. And Ijust thought, okay, I know these
two, and I know they're notreal, but I thought, but what if
I didn't, like, in the future,are Kyle and Cheryl going to win
a podcast award becausethey're so popular? I'm like, I don't
(01:30):
know. But I. It just. Itreally did rub me the wrong way.
I don't know what have you runinto this yet?
Not into AI stuff. I tend tostick with the same group of podcasts
I have heard. Well, I couldtake POD News Daily, for example,
with James Cridland. There wasa time when he was sick, lost his
(01:50):
voice, and he used AI to voicehis entire episode. And he disclosed
that at the beginning. And Icould understand now. It was obviously,
since we listened to James somuch, it could not replace him.
Right.
It sounded like AI to me. Icould tell other people might not,
especially if that was theonly thing they heard from him. Yeah,
(02:11):
they might not tell. Now, thebigger idea here, and there are some
interesting, specific thingsto get into, as well as something
I submitted as a proposal forpodcasting 2.0. But the bigger picture
here is what needs to bedisclosed and why. Right now, the
conversation is so much aroundthe use of AI but for years, we've
(02:33):
needed to disclose things inways that were compensated. That
is a big thing. So much thatthe FTC recently in the United States
came out with some newguidelines a couple of years ago
to say, here's when you needto disclose. Here are some ways you
need to disclose. You need tomake sure your disclosure is very
prominent. Talking aboutaffiliates or if you're compensated
(02:54):
in any way by talking aboutsomething or by including content.
And I did a whole episodeabout disclosures and such related
to that. But it has reallychanged from you can't just stick
something at the end of yourepisode or on the footer of your
website or on a separatedisclosures page on your website.
It needs to be prominentbecause. And this is the reason.
(03:15):
So think about this foranything along this line of disclosure,
because the person on theother side, the audience, needs that
information so they can takeit into consideration for how they
then take action or receivethe rest of the information. Because
if I'm being told by you thatyou're getting paid $1,000 per minute
(03:42):
to talk about the color blue,I'm going to think, well, wait a
minute. Does he really likethe color blue? $1,000aminute. That's
a lot. I mean, I talked aboutthe color blue for $50aminute. So
knowing that you'recompensated now, you don't have to
disclose how much you'recompensated, but it's that thing
(04:02):
that's giving me informationto make my own decision. So it's
putting the responsibilitythen in my hands. Because now that
I know that you are beingcompensated for talking about the
color blue, I can decide ifI'm going to believe you or not.
And then you can also do agood job of proving whether you're
(04:22):
saying good things onlybecause you're being compensated
or if you are just a shill forit because you're just. That's your
sponsor for the moment.
Well, I always thought it wasfunny when I would see articles about
podcast media hosting, and theonly people they reviewed just so
happened to be those companiesthat had affiliate programs. It was
(04:43):
like, wait, you're not goingto talk about this? Coming up. Just
Captivate, Buzzsprout,Blueberry, you know, Libsyn, that
was it. And that was like, oh,that's kind of odd that they're not.
Because there are other onesthey didn't talk about Spotify, didn't
talk about Red Circle or anyof the other ones that are out there.
And I was like, oh, I seewhat's going on here. These all have
Affiliate programs. And thatmay have been, you know, why they
(05:05):
chose those people to do that.I know for a while there were so
many websites, they'reprobably still out there talking
about web hosting.
Oh, yes, yes.
Because I remember when I wasmaking I think 50 bucks from Hostgator
back in the day, you know, youget a couple people sign up for that,
like, all right, dinner's onDave tonight. That's cool. Yeah,
the FTC actually had. It wasftc.gov I think, influencer for a
(05:30):
while. There was an actuallike video there and they explained
how it has to be at thebeginning. So like all my newsletters
now start off with my Amazon.I may earn whatever, you know, the
disclaimer for that. Andthey're saying in the, you know,
if you're doing videos andlike, hey, today's show is brought
to you by Dr. Rob. Blah, blah,blah. By the way, today's guest is,
(05:52):
you guessed it, Dr. Rob. Like,you have to say that up front. And
I just think the young kidsback in the day on MTV had a very
popular show called Catfish,which I never quite. I guess this
is a real thing where peopleget online and you think you found
the person of your dreams,only to find out that it's some 14
year old kid that's having aprank. And I'm like, well, in a way,
(06:16):
if you don't disclose that I'mnot real, that's catfishing. And
so somebody. Again, that's tome. And I think the difference is
the disclosure. Because look,we all watch TV and I know that Monica
Geller and Rachel Green andeverybody else on Friends, they're
not real. They're playingcharacters. But that's the whole
point. I know they're playinga character. It's weird. You kind
(06:38):
of. That's why you get sadwhen somebody dies or whatever, because
you somehow feel like youknow, this person that doesn't really
exist, it's a character. Butif you don't know that, then you
just feel like you're, youknow. Wow, I love. This is one of
my favorite podcasts. I just,I identify so much with this person.
Then you find out later it'snot. If you want to get on my bad
side, lie to me. Because youdon't have to lie to me. Like, I
(07:01):
can take lots of criticism.You don't. There's no reason to not
just tell me the truth. And ifyou don't and I find out, it's hard
for me to trust you again. Somaybe that's the other thing. Too.
I just have a. I am sensitiveto people lying to me, I guess.
Well, and especially in thisspace with AI taking over for so
much, or rather people givingover control to AI. I mentioned this.
(07:25):
I was a guest on thepodcasting 2.0 show recently with
Dave and Adam. And I broughtthis point up too. And I've mentioned
this in my own podcast thatthere's so much. I love Adam's term
for this. So I'm going to helpspread this. There's so much AI sloppy
out there that when you sharea genuine opinion, to me, it stands
(07:47):
out. The web hosting thing. Iknow that in the past I've looked
for things and Blue Host isalmost always near the top because
they pay out the biggestaffiliate program. And if you get
on YouTube, like looking upfor product reviews or something
like that, you'll typicallysee like this AI slop. That is, they
just found the top ratedproducts on Amazon and that's the
(08:09):
order of theirrecommendations. And they have the
audacity to say, in thesevideos, we tested the top 10. No,
you did not test, you liar.You did not test these things. You
just looked at reviews. Whichis actually. Some of. This is actually
a violation of the Amazonterms of service to run their product
(08:29):
pages through an AI in orderto promote your affiliate link. That
is a violation of their termsof service. But that aside, that
kind of stuff, there's so muchof it out there that a couple of
products that I've looked for,when I saw someone using the product
themselves, I see this forkitchen gadgets, for example. I enjoy
(08:50):
being in the kitchen. Therewas something. I was looking up all
this AI slop out there and Ifound a guy actually using the product.
His camera quality, his videoquality was okay, his audio quality
was okay. I think he mighthave been wearing a LAV mic, or maybe
not. But I could hear him andI could understand him enough. He
(09:10):
didn't have studio lighting,he wasn't in a studio, but he was
genuine, authentic and I couldunderstand him. So his quality was
good enough. And that sold meon the product because of all the
AI slop. I just wanted agenuine opinion. And if in those
videos, if they would say, thefollowing is generated by AI or the
(09:35):
following is a list of the topmost rated products. Like, they don't
always have to say, we didn'tactually review these products, although
that would be good for them todo. They could put it in more positive
language to say, this is aoverview of the most popular products.
That's better. And that Equipsme to make a better decision. Same
(09:57):
thing. Like what you weresaying is when they don't disclose
in that way, and I realizethis is just more AI Slop, then I
get upset, I thumb it down,and I leave. And I'm so tempted to
just comment on there, like,yet another AI Slop video, but I
try to be nice on the YouTubecomments, since everyone else isn't.
(10:17):
Well, we've all seen itbecause they have found some stock
images of the product,probably from their website, and
it's always slowly panned orslowly zoomed in because they don't
have any video and they'retrying to do. And then it's just
this AI Voice that isexplaining. And I've seen some that
are. Just before AI was evenremotely passable, that was just.
(10:39):
It was very robotic. As itexplains, the New Juicer 2000 will
do this. And you're like, holycow, is this horrible? And then it
dawns on you that this may bedone by somebody where English is
not their first language.They're just, like you said, they're
just trying to get thataffiliate commission, and they're
looking at what's popular. Andhere we go. So now to flip the coin.
(11:01):
I was doing a show, and JodyKringle, who's a voiceover artist,
who, as you might imagine,might have her eye on exactly what
AI Is doing. And we weretalking about people using Google
Notebook to do a show, and shesaid, if you don't want to do the
show, like, just don't do theshow. And when she said that, that
(11:23):
triggered it, because there'sa show I do that. I honestly don't
want to do the show. And it'sabout Akron, Ohio. It's called the
Akron Podcast. And I was like,ooh, that I wonder if this would
work. So I created an episode.And what was interesting, it was
a very sensitive subject inAkron, Ohio. Our police have a bad
habit of shooting people whenthey shouldn't, especially white
(11:44):
cops shooting people of color.And the actual people of Akron, Ohio,
now have to pay $4.8 million,not an insurance company, because
we shot a guy 46 times. Iguess the 45th, not enough or whatever.
But anyway, so I thought,well, this will be interesting because
it's a sensitive subject andAI is going to describe it. So I
(12:07):
was like, okay. And then therewas another. The second story was
yet another shooting that Ihad Kyle and Cheryl talk about it.
But I disclosed up front. Andafter the first story, I said, hey,
I have another story aboutthis Poor kid got shot because he
had a fake gun and he waspointing it at people. And I said,
so let's hear what Kyle andCheryl have to say about it. And
(12:29):
off they went. So I justexplained up front. I'm like, hey,
these are. I said, now I'velistened to what they had to say
and I've checked for accuracy.I said, but here you go. So I just
put that out today. It's avery small show because I. Again,
I don't really want to do thispodcast at this point. It's just
a test show for me, but I'llbe interested to see if I get any
(12:49):
kind of listenership. Theinteresting thing about Google Notebook
is in the past, if I hadthought about doing this, I would
have ran the story from thelocal newspaper, had some sort of
AI rewrite it again, check foraccuracy, and then throw that script
into Google Notebook. Well,Google Notebook kind of does that
automatically. You can copyand paste in text and then it just
(13:12):
straps in and does a deep diveas they delve into whatever. So it'll
be interesting to see. But Ido feel somewhat of a hypocrite that
I. But I'm like, no, no, I'mdisclosing. I'm not a hypocrite.
And there is a place for someof that. Not from. Don't think of
it completely as slop, becauseyou are taking actual content and
(13:34):
you're using AI to repurposeit. What I would call AI slop is
when you are creating thecontent almost from scratch or the
opinions almost from scratchwith AI, that is totally slop. But
if you're repurposing, I thinkAI is fantastic at repurposing. But
here's the thing. The point Iwant to make here, you are making
(13:56):
content more available topeople who want it in that format.
So, like, if I get to awebsite and I see a really long article
and they have an option or mybrowser has an option and plugin
or whatever to give me an AIgenerated summary. I'm clicking that
sometimes to see that.Actually, I'm using an email app
(14:16):
right now called Short Wave. Ido have an affiliate link, but I'm
not talking about them becauseI have an affiliate link. But they
have this neat feature that Ican press a button on an email and
it will summarize that emailor even an entire reply thread of
emails. It can summarize itfor me or given me action points
from it, or I can selectmultiple separate messages in my
(14:37):
inbox and have it summarizeall of those together. That's great.
That is saving me time. It'sgiving me information in a format
that's more digestible for me.Or you visit a website that also
has a lawn article, and youwant every word of that article.
You just don't want to read ityourself because you need to be copying
(14:58):
and placing or clicking thingsand doing other stuff like that.
But you want something tolisten to. So, I mean, that's. The
podcast solution is if there'sa button to turn that article into
a podcast where you can hearthe article read to you. If it's
an AI voice, I don't thinkthat needs to be disclosed because
it's just a voice. The sameway Mignon Fogarty with Grammar Girl
(15:23):
does this in her podcast,where so often she has guest writers
who contributed content, andshe discloses that she says, the
following is written by so andso. So when I say me, it's referring
to him or her. So then shestill reads it. And she reads it
(15:44):
in her genuine voice and withher own humor and inflection and
all of that to it, but it issomeone else's content. She's reading
it. I think that's disclosureenough. If an AI is simply reading
content that you've generatedyourself, not with AI, but you've
made it yourself, that mightnot need to be disclosed necessarily,
(16:05):
unless the AI is identifyingitself. So here's the question to
think about is where is theline of, at least with AI, where
you do and don't need todisclose your use of it?
That's a great question. Ithink you hit one of the points,
I think, are reviews. BecauseI remember one time I had a client
(16:29):
and I asked him, would youhave any gear yet? And he said, yeah,
I have a Podtrak P4. And thenI bought a Rodecaster back when the
first one came out. And Isaid, why do you have both of those?
And he said, well, when youtalked about the Rodecaster, you
said it was cool. And I waslike, wait. So he spent $500 because
I said something was cool? AndI was like, well, it is cool. I'm
(16:51):
like, but if you got the otherone, you kind of don't need this
one, or vice versa. And so Ithink that's why people like Tom
Buck on YouTube and AndrewScott, they always disclose up front
whether or not, like, hey, I'mgoing to talk about the new sure
Move Mike. And, like, Bandra.I just saw his video on that. And
he was like, hey, just so youknow, this is on loan from shore.
(17:13):
And then they always say, I'mgoing to give you my honest opinion
one way or another. And that'swhere it's always, for me, I'll listen
to it. And I always love thatboth of those guys were like, this
is what we really liked. Butthere is one thing, and the minute
they say something remotelynegative, I'm like, okay, I'm going
to believe this. Now, ofcourse, I always say you start off
(17:33):
with two things when you starta podcast. No audience and integrity.
And so the minute I find outthat you're lying, we're done. But
Tom Buck actually has a. Itsounds weird because he's. I look
at. Tommy's a YouTuber, but heactually has a podcast. Podcast.
It's an audio podcast. And hedid an episode and he was talking
about how companies have cometo him and go, hey, can you do a
(17:56):
review of our product? And youknow we're going to give it to you
for free, but if you could notdisclose that. And I'm like that.
It's just so weird that thereare companies going, yeah, can we
kind of keep that on thedownload? I'm like, well, it's a
little thing we like to callillegal, you know, But I'm sure there
are people like, man, I'mgoing to get free gear.
(18:17):
Sure.
And I'm like, oh, that's so bad.
I kind of feel like the lineis similar to a trademark where in
trademark stuff. And I've gonethrough this, and I'm going through
this with a trademark disputeright now. It's a matter of potential
confusion. Might someone thinkthat AI generated image is real?
(18:38):
If so, it needs to bedisclosed that it was AI generated?
Might someone think that thatvoice is a real person saying those
real things, giving their ownopinion that might need to be disclosed.
If it's simply a voice doingsomething for you, like reading something
(18:59):
for you, just so you have itin a different voice. And you've
never said, here's the voiceof so and so, or anything like that
that might not need to bedisclosed. And if you're repurposing
your own content that youalready created, like you, maybe
you used AI to generate abunch of social posts for you or
to summarize it. I don't thinkyou need to disclose the following
(19:21):
summary was generated withartificial intelligence. That's a
waste. I don't think you needto do that. But when there is the
potential for confusion, orespecially if you are trusting the
AI, blindly trusting the AI,then you definitely should Disclose
it. The illustration that Ilike to use in one of my episodes
(19:44):
about transcripts. If you useAI to do a transcript, you need to
be careful with certain thingsbecause sometimes the AI or the transcript
engine won't hear somethingcorrectly. And there are certain
cases where that could lead tomajor problems. An example I like
is you might say, I love twolittle puppies. Or you might slur
(20:08):
it, saying, I love two littlepuppies. The AI might hear that and
it thinks you said, I love tokill puppies.
Right.
And that can lead you to legalhot water. So if you don't audit
and edit, you could get in trouble.
Yeah, I'm with you. I thinkthe biggest mistake is a having it
(20:30):
generate the content. I alwayslike the way you said it was assisted
intelligence. So have it brushup on the content that you created
with and then from. I thinkthat's the best use of AI And. But
I also like your point. Like,for me, when I think about it, it's
just really just reading thefacts that I gave it. And so it's
(20:51):
not really throwing inopinion. And I was really glad to
see that both Kyle and Cherylthought killing people was bad. I
was like, yay, Kyle andCheryl. So that was good. So I think
that it's. When you, like yousaid, when you get into opinions,
and if somebody's going totake action on your words, it would
be good to know if it's someAI tool.
(21:11):
Yeah. With AI especially, itcan hallucinate, as they say, and
so either it can get somethingwrong, it heard something wrong,
it could get a fact wrong, itcan make up things. Like I loved
hearing in the no Agendapodcast recently, where I think it
was John C. Dvorak did this,where he took Notebook LM from Google
(21:31):
and he ran something throughit and then he got the transcript
and he ran that back throughNotebook LM again. I can't remember
how many times he did that.
It was twice.
Only twice? Yeah. And ended upwith some definitely hallucinated
information in there. It'slike playing the game of telephone.
This was about when the pagers exploded.
Oh, yeah.
(21:51):
And so they took that and hadit transcribed and then ran it through
back again. The final outputafter basically running it through
itself twice was they weretalking about exploding electronic
devices. And it wasn't verylike they lost some of the specifics.
And there was one. I don'tknow if they called it a Walkman.
(22:13):
Yeah. But it was something.It's like, oh, now we're in the realm
of not even close. And that'swhy Adam is all about, yeah, just
give us more AI slop. Becauseit's just going to be bad. And anybody
who does more than clicking abutton is going to stand out. So
I tend to agree with thatopinion because I've. I've just heard
some things. I'm like, ooh,that's. That's not good.
(22:35):
But then you look at thingslike, if you use an AI to generate
a title for you, I don't thinkyou need to disclose that. If your
AI made the chapters for you,I don't think you need to disclose
that. There are lots of thingsthat AI can do that I don't think
you need to disclose. And,like, also, you know, people are
using AI image makers to maketheir podcast cover art. I think
(22:59):
that's fine. You don't have todisclose that. That's AI. This is
all opinion, by the way. So adisclaimer here. We are not lawyers.
We have certain experienceanyways, but we are not actual lawyers.
So this is not legal advice.But my impression is that these types
of things, if it's obviousit's not real, you don't have to
(23:19):
disclose it. It's just like, Idon't have to say whether I used
Adobe Photoshop or GIMP if Iedited an image, but if I make that
image look real and it'ssomething obviously fake, obviously
not real, then I shoulddisclose that that is a parody in
some way. You look at, like,the Babylon Bee, one of my favorite
(23:42):
sites on the Internet rightnow, hilarious stuff comes from the
Babylon Bee. And they have allof these funny images that they make
with their articles, and it isobviously satire. So they don't have
to say this image in thisarticle is satire, because that's
what it is. That's their wholebrand. Same thing with the Onion.
(24:04):
And there is a case right nowwith, I think it's gone all the way
up to the Supreme Court. Andthe Babylon Bee has been involved,
The Onion has been involved.But, like, defending the right to
do parody because some peopleare saying someone might get misled
by this idea that thishappened. Or there's that. The example
I'm thinking of, there's theone where Jesus flipping over the
(24:27):
tables because there's anaccount in Scripture of when Jesus
cast out the money changers inthe temple, and he said, you turned
the temple of God into a denof thieves. And so he flipped over,
like, pushed over the tablesand cast them out. But then what
did the AI make?
It's a big picture of JesusBasically doing a backflip over a
table.
(24:48):
So it's like the AI understoodyour words, but it didn't understand
the idea. But if you didsomething like that, you know, pick
a politician, any politician.Politician steals candy from a baby,
here's photographic proofthat's going to get you in trouble.
Yeah. And I think it goes backto what you said earlier about confusion.
(25:09):
So when you make. I know, Ithink it was Taylor Swift, but they're
insert any female celebrity.And the minute you get popular, it's
probably a rite of passage,which is just sad that the minute
you get popular, nakedpictures of you are going to pop
up. And it's weird because Ithink everybody knows they're fake,
but it's naked people. So. Andit's just. So that's the again where
(25:33):
you get into trouble, whereyou're like, oh, this actually looks
real because he or she is on aboat and this might be that. And
you know, that's when you'regoing to get a trouble.
(26:00):
The trouble is more than justthe pixels. It's the ideas, it's
the connotations, it's thefeelings, the other things beyond
just the pixels themselves.And bringing it back to other kinds
of disclosures, too. Thesedays with financial disclosures,
whether you're beingcompensated for something or you
(26:21):
have some kind of associationwith something, like while I was
on the advisory board forCaptivate, I disclosed, whenever
I was talking about Captivate,I would disclose. I am an affiliate
and I'm also on the advisoryboard. And you have done this a lot,
Dave, while you were withLibsyn, whenever you would talk about
media hosting providers,whether positive or negative, you
(26:43):
would disclose that you workfor Libsyn. And yet the way you've
disclosed that and the waythat you talk about other things,
I think boosted yourreputation. So it didn't make people
think, oh, Dave is just sayingthis bad thing about them because
he works for Libsyn, or he'sjust saying this good thing because
he works for Libsyn. Butbecause you were honest, because
(27:04):
you shared the good and thebad on both sides of whatever you
were talking about, that madethe disclosure boost your reputation.
And I think we're going to seein the future more disclosures required.
I think that for sure, the useof AI, there's going to be some need
for disclosures in that from alegal perspective, I think that's
(27:26):
coming in probably the nextcouple of years. There will be some
kind of guideline or a lawthat you'll need to disclose. But
it will probably be somewherealong those lines of when someone
might confuse it or it mightmislead someone or something like
that. We also have the lawsalready about financial disclosures.
(27:47):
If you're being compensatedabout something and even if you're
not in the United States. TheFTC guidelines are interesting in
that they say basically, ifyou do business with anyone that's
in the United States or havethe potential to profit from anyone
in the United States, youstill need to disclose. But I think
they're good guidelinesanyway, and we'll have a link to
them in the notes so you canlook over them even if you're not
(28:09):
in the United States. Butthere isn't a completely good way,
a standard way to disclosethat works every time. And that's
why one of the things that Iproposed just recently in the podcast
Namespace Talk in Podcasting2.0 is a podcast disclosure tag that
(28:29):
would allow you to add thesedisclosures in your RSS feed. Now,
that will not be enough rightnow. What that would do is for the
apps that would someday readthat when that standard is accepted
and finalized and everythingif it is, that would allow the apps
to support that, to make thosedisclosures more prominent. And there
(28:50):
could be certain ways that youcould make that especially more prominent
and contextually relevant.Like maybe you could tie a disclosure
to a chapter. So while you aretalking about a product that you
earn commissions from, there'sa disclosure or during an ad, even
there's it says it in thechapter and such. This is an ad.
(29:12):
Now, ads, I think, are obviousthat they're advertisements. Most
of the time it's obvious. Andif it's not obvious in your own podcast,
then you need to make itobvious that it is an ad. You are
being paid to talk about thatthing. So you need to make that obvious
in some way. And there areproduction ways to do that or things
that you can say to do that tomake that obvious. But I think that
(29:33):
extra context can help people.And yes, this does open up the door
for systems someday in thefuture to detect when and where the
ads are in your episode. Butthat's going to happen anyway. That
is happening anyway. And in away that's actually going to help
you. Because as technology isgoing in the direction of being able
to detect when people hearads, if they're going to skip the
(29:57):
ad anyway, the sponsorshouldn't have to pay for it because
they're not getting whatthey're paying for. They're paying
for the actual impressions. Sothat could make Realistic impressions.
Even more valuable, I wouldhope, but. So I've proposed this
podcast disclosure tag thatit's a very rough proposal right
now. I'd love some comments onit from anyone who just sees it and
(30:18):
has some ideas of differentways that could be implemented. But
I tried to build it aroundthis idea of different types of disclosures
as well as different contextsfor those disclosures. So in my own
example that I gave in thecode, I gave the examples of three
different types of disclosure.Compensation, AI and association.
(30:40):
AI is obvious. It's if you'reusing AI for something that needs
to be disclosed. Compensationis if you're compensated in any way
for that thing, financialcompensation, you're earning points,
you were given something, youare paid, it's a sponsored ad or
something like that. Any kindof compensation would fit under that
kind of disclosure. Andassociation would be the thing where
(31:03):
you are associated like you'reemployed, you own stock, you're on
the advisory board, youfounded the company, anything like
that, you're related to thisperson, you're married to this person,
or they're your brother,sister, son, daughter, whatever.
Any kind of association likethat. Those are some types that,
that I thought of in theexamples I was making. And then you
(31:25):
could also think about addinga context to these which I put in
the code examples so you couldclarify where this is relevant for
compensation. The contextcould be links and in my example
I wrote some of the productsslash services in this episode use
affiliate links thatcompensate the podcaster for qualifying
(31:45):
purchases. It could be evenlike a context of content and that
could refer to the entirecontent of your episode or a portion
of content. Like if you have,you know, the tip of the day is brought
to you in part by orange juicedrink some today. Well, that could
be in the disclosure is thatyou are compensated for that piece
(32:06):
of content. Or it could bejust outright an ad is the context.
It could also be certainthings like the images that were
in there and to disclose thisepisode uses realistic looking AI
generated images, or thisepisode contains parody, anything
like that. These aredisclosures, not disclaimers. Disclaimer
(32:28):
is when you're saying I'm nota lawyer, this is not legal advice,
this is not financial advice,that's a disclaimer. But a disclosure
is if you are a lawyer or youneed to disclose that or you're related
to something, you're beingcompensated or in certain cases if
you're using tools, that kindof thing needs to be disclosed. Certain
(32:49):
kinds of tools like AI is whatI'm Talking about, or maybe if you're
using a product and talkingabout a product that was given to
you by the company in hopesthat you would do like right now
I'm podcasting with an electrovoice RE320 that goes into a original
first generation Rodecast Pro.Both the microphone and the Rodecaster
(33:12):
Pro were given to me by theirrespective companies. They were given
to me with no expectation thatI would do a review. They were given
to me out of appreciationbecause I already said nice things
about their products. And sothey said, hey, we want to give Daniel
one of our products so he canhave one. And when I talk about them,
especially if I talk about myuse of them, I try to make sure that
(33:35):
I disclose these were given tome. Nonetheless, I recommend things
that I truly believe inregardless of earnings. That's my
little thing that I say somuch. Now, it's a text expander snippet,
but this technical proposalfor a disclosure tag I think could
be great because then there'sless chance for someone to miss the
(33:55):
disclosure. And that's a bigimportant thing on the legal aspect
is the disclosures have to be prominent.
Well, that would be it becauseI mean, to play devil's advocate,
right, we already have anexplicit tag and that is completely
self regulated. And ifsomebody hears a show that should
(34:16):
obviously be explicit, you canreport that to Apple. The difference
is we all know what explicitsounds like. Not everybody may be
able to go, oh, that's Kyleand Cheryl, that's AI. And so that's
going to be the tricky part iskind of policing those people that
don't follow the rules. Butit'll be interesting because I know
already from the very firsttime I heard Google Notebook to now,
(34:40):
that's already better than itused to be. They breathe now they
occasionally chuckle. It'sinteresting how they've boosted that.
I'm sure. I know you can nowgo in and tell them the angle, what
you want them to talk about.And eventually you're going to be
able to pick somebody besidesKyle and Cheryl because not everybody
sounds like Kyle and Cheryl.And who knows, maybe someday you'll
(35:02):
be able to, you know, thiswhole show will be AI, just me and
Daniel and we'll say here makemake Dave kind of tired in this episode,
make Daniel super bounce offthe walls kind of thing. It'll be
interesting to see.
So when it comes to likecertain things, I know some people
have said, what does thismean? Like if I use AI Tic Descript,
for example, one of thebeautiful features that Descript
(35:24):
and a couple other tools offeris the overdubbing feature, as some
call it, or certain thingslike that, where you can use an AI
that's trained on your voiceto fix something that you said. I
think that's okay. That youprobably don't have to disclose it
because it was a mistake.You're simply correcting it. You
(35:44):
could do the same thing by rerecording it. Yeah, it might not
sound as smooth, but if you'revery skilled at what you do or if
you time it right, you canmake it sound as smooth. But that
doesn't have to be disclosed.You're not changing the content,
you're not misleading someonewith it. But if you were to use the
(36:07):
AI to make someone saysomething different, that's when
it gets into trouble. And evenwhen it comes to, like, the Kyle
and Cheryl voices, if you'resimply using them on your own content
and you fully edited thescript and all of that, should you
still disclose, probablybecause of this. Where else do Kyle
(36:30):
and Cheryl appear? And whatkind of impression is that going
to give the people who hearKyle and Cheryl, will they start
thinking, oh, man, this Kyleand Cheryl, they do great podcasts.
I want to find more podcastsby them. I hope I can meet them someday.
And then you discover they'refake. They were fake all along.
(36:51):
They're on the Wiccan show.
Yeah.
It's like, wait, what?Cheryl's a witch. I didn't know that.
Yeah. So that could be alittle bit of a problem in a weird,
spooky kind of way. It'll befun to watch and see where it goes
and see which lines it cross.And I think I know james cridlin@podnews.net
somebody. Oh, it was ListenNotes. Our good friends at Listen
(37:12):
Notes had come up with somesort of tool. And I want to say it
was already. It was either 208or 802. My brain's going dyslexic
on this. But either way, itwas A lot of shows were already using
Google Notebook and I waslike, oh, well, this is not unique
at all. So it'll beinteresting to see if they win Podcaster
of the Year next year. They'relike, these guys are productive.
(37:35):
They're, you know, I've doneprobably somewhere between 20 and
30 podcasts and tried and thenshut them down. These guys are on
70 shows and they're stillgoing. They're all still current.
Yeah.
And you mentioned some statsjust now. I think certain stuff around
that needs to be disclosed.More in the future, too, because
we are seeing so much, Ithink, especially more in the recent
(37:56):
years than before. But howpeople can lie with statistics and
the way that you tell thestory around the data. Sometimes
just the way a question isworded in a survey can affect how
people answer it. So I thinkwhen it comes to, like, survey data,
(38:19):
people should disclose whatthe exact question was and the exact
answers that were given, andthen you can summarize it from there,
you can abbreviate it orwhatever. But I think that's really
helpful to know because Isometimes call myself Mr. Caveat.
I start thinking of all ofthese caveats to things, and people
know that from the Audacity topodcast, I give lots of caveats and
(38:41):
alternatives and all of thatkind of stuff. That's kind of why
I started the show, because Ifelt like, hey, wait, wait, wait,
there's a caveat there to whatyou're not covering this caveat.
Some of that stuff you mightneed to disclose from content perspective.
So don't just think of it asfinancial or AI and that kind of
thing, but think about it fromthis perspective. And this is really,
(39:03):
I think, the crux of all ofthis. What does the audience need
to know that will help themmake a better decision of how to
process the content or takeaction? What is it that they need
to know? What is that extracontext that they need? That's the
kind of thing that adisclosure should cover.
(39:24):
Yeah. On a. An article onlistennotes.com they said just last
weekend. So this came out atthe beginning of the month. So it's
been more than that now. Butthey'd already deleted 500 fake shows
that were created withNotebook LM. I may have to realize
that, yes, I'm playing and I'mexperimenting, but the Akron podcast
(39:45):
may lose our listen noteslisted because I've done this. So
we shall see.
And I'm not so sure there's aplace to just cut things out because
they're AI. Because again, itcould be that by being an AI voiceover,
it is truly making contentaccessible, both from the perspective
of just, I don't have the timeto sit down and read this. I'd rather
(40:07):
listen to it. But alsoaccessible in the accessibility perspective
of someone who's visuallyimpaired, wants to listen, has to
listen to the article, read tothem, and maybe they just really
do not want to hear thearticle verbatim. They could retain
the information and takeaction on it. So much better if they
(40:29):
hear a conversation around thecontent. Like, that's what Notebook
LM is doing is it's turningbasically monologue content into
a dialogue that isn't usuallyadding much to it in terms of value.
It is adding fluff, we couldcall it. But it is fluff that is
more personable and because ofthat can make it more relatable,
(40:54):
more actionable for some people.
Yeah. And the part that's kindof scary is many podcasts with actual
people on them take reallywide tangents. And to their credit,
Google Notebook is somewhatstaying on topic. There are some
things we might actually learnfrom this. So maybe I don't care
(41:16):
that today is the two yearanniversary that you got your dog.
That's a real thing. Ilistened to six minutes of a show
this week and they, theyexplained what they were drinking.
They talked about theanniversary of the dog. I forget
what else, but I was justlike, they didn't. They barely announced
what the show was. So thereare things that maybe we could learn
from getting to the point andyou know, just the facts. Maybe so
(41:40):
we shall see.
And speaking of disclosures,we received a thousand sats from
someone in a booster grammar.So there's your disclosure for that.
Dave, what did we receive?
Yeah, from Creativity Foundcame a bit late to this episode holiday
listening and liked thediscussion around what or more differently
you can do with a podcastformat. So a your back catalog still
(42:04):
works. So keep that in mind.They continue on. I've planned a
step by step course formataround the features of podcasting
2.0. Yay. With explainerepisodes where guests tell me what
the thing is and why it'sgood. Then practical episodes teaching
the listener how to do thething. It's still a talk show, but
(42:26):
we'll use the 2.0 features toteach the 2.0 features inspired by
books I've worked on as a texteditor. And that is from Claire.
So thank you, Claire. That'sawesome. Very meta.
Yeah. And we both got tocontribute to her series as well,
so we might have a link tothat in the notes. I can't, I don't
know if it's published yet. Ithink it's publishing around the
(42:49):
same time this episode ispublishing, so we might be able to
link to that. Some of ourcontributions, like I got to talk
about chapters there andothers have been involved as well.
So love what Claire is doing.Thank you for those stats.
Yeah, thank you very much. AndI think that is going to do it for
this episode. Episode 52 is inthe can again.
(43:12):
Our websitefutureofpodcasting.net keep boosting
and keep podcasting.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!