Podcasting's AI Risks and Opportunities - Podcast Tips Ep 36

Welcome. I'm Rob Greenlee, and tonight we're going to discuss AI's transformative effects on podcasting and new media creation on the show. So we're going to examine the incredible opportunities and its significant challenges and risks tonight on the show in the area of podcasting as well as video and audio mediums just in general out there. I think a lot of what we're going to talk about tonight applies to a lot of different kind of content creation and work and kind of like our our daily lives and how we consume content too. But it's really gonna impact the creation side of podcasting. In this episode 36, I'm gonna emphasize AI's transformative nature and I'm gonna suggest some significant shifts in podcasting and its methods. And this is all evolving right now. So let's go ahead and jump in here. I'm gonna turn down this music and actually turn it off. But well, yeah, so I wanted to start expanding on this. And while listeners still expect, I think, full human-created content, I do think that audiences will react long-term to AI's growing use in a positive or negative way. And I'm not sure exactly which way it's gonna go. I think it depends on what we as content creators decide. And also because content creators are also content consumers too, as we see this growing use of AI technology to impact how we produce content, how we market content, how we package it, and market it. And this applies, like I said earlier, to all content creation, whether it's short form, long form, audio, video, everything that's coming to the world and actually could have a significant impact on how we consume content too. Just think about the concept of an autonomous vehicle that you're driving around in that drives itself. That's AI technology giving you the potential ability to, while your car's driving down the road, watching video. So you could be watching a YouTube video or listening to a podcast while you're driving down the road. Now, granted, you can listen to a podcast right now if it's audio only, but you potentially, at some point here, you'll be able to watch a full video on your big screen on your car. And it doesn't necessarily have to be an electric vehicle. It can be any kind of a vehicle that may have this ability in it, but But it's just a matter of, you know, it's gonna change things. And AI is definitely getting set up. It's not a huge impact quite yet, I don't believe. I think what we're seeing is the very early stages of it. And it could go very positively or it could go negatively. And I kind of wanna dive into that tonight. I think that on the positive side, so I would prefer to kind of start out on the positive spectrum of this because there are quite a bit of negative or downsides of us evolving into this AI kind of environment when it comes to this topic of content creation and consumption. But I think that the big positive changes are the impact on each of us time efficiency and productivity. Now, obviously, those are probably pretty inherent, and those are really the first line of benefit that we see today is on the impact on all of us who are content creators creating content. But I do think that over time, it's going to help increasingly on the consumption side too. The AI technology will enable us to, as consumers, to better find content that is aligned with hopefully not exactly what we want to see, but maybe what we should be seeing. And should may take a different kind of a spectrum here than what you might expect. I think what we've been through over the last 10 or 15 years is, and I got caught up in it too in the early days when I was doing the nationally syndicated radio show that I did, about the development of the internet was being all proactive about what's called personalization, right? And I think that concept seemed very attractive at the time, but I'm not sure that it was as attractive when it was fully deployed. And I'm going to go into that here too. You know, AI can be used to streamline editing right now, can be streamlined to create show notes, it can be streamlined it can streamline your editing process of the content as long as you don't go too far down that path. I think some of the tools in the marketplace right now give probably— I'm not sure we found kind of that balance. And I think that the tools give you the ability to do a lot more editing and a lot more processing than maybe we should be doing. And that's kind of the process I've gone through too. I've played with a lot of the tools and have processed audio, and maybe I've gone down that path too far. So I've had to kind of pull back on that and say, well, you know, it's great that it does some of these things, but maybe what it's creating is an unhuman-like experience. So, you know, if you cut out too many like spaces where you have dead air, or you cut out too many filler words like the ums and the ahs and, and the wheeze, and there's like 47 kind of filler words that have been identified in these tools that you have the option to hit one button and they're all gone, poof. But if you're processing video and audio, it can work pretty good on the audio. So these tools can work fairly well on the audio, but when you do it to the video and the audio, it can create kind of a jerky kind of video experience. So you kind of have to, you know, maybe do the audio separate from the videos, kind of what I'm learning here. So anyway, I didn't want to get too far into the weeds on this, but the other big benefit in this efficiency is transcriptions and summaries. And I think I mentioned that a little bit of being able to use these tools to summarize, create chapters, create markers in the content that can help you with your post-production publishing process. And I think that's a really powerful tool for doing this kind of content in a more efficient way and taking advantage of the distribution technology that exists now around best ways to consume. I mean, some of these things are being done automated in platforms like YouTube and other places, but if you're in the podcast area, you might want to, you know, do these things and put it in your show notes so you can have like markers in there at what point I'm talking about this topic at one, you know, like minute 2:30 or something like that to be able to have those jump links. So there is some time savings that can be had from this, and I think that's kind of where the major benefit is of these things right now. And then the other big area is enhancing your ability to create content and coming up with like new ideas or enhancements on the ideas that you already have. I think AI can help you with that, can really give you a bunch of ideas or that's really kind of almost like a brainstorming kind of partner of sorts that can come up with So if you're trying to figure out what title you want to give to your episode or something like that, you can just prompt the AI to say, give me 5 options of what you think are the top keyword phrases that we can use to tag our episodes with to get clicks, right? Because we all want clicks. And then some of the tools are building in the ability to analyze your audio, analyze your your transcript and come up with artwork that would be custom to your episode. So I think that there's some significant time savings and some of these processes are still a little bit rough right now, especially on the artwork side. It's not as great as it could be, but I think over time it's gonna get there and it's really going to be a powerful thing. To kind of speed up and to enhance the marketing of your show as well. And the whole kind of personalization part of it is one thing that I think can be utilized at some point, but I'm not as convinced on the personalization side of it too much because, I mean, I think maybe at some point the AI on the consumption side can personalize messages to the listener of what kind of content that might match somehow, but I do, and I mentioned this earlier, I think we have to be careful about that whole kind of suggestion stuff that tends to be algorithmically focused on a certain limited spectrum of what our interests are and only delivering to us what the algorithm thinks we want. And I think that is, I've just kind of learned this over being a technologist for all these years, is that I think that's driven a certain amount of division in our world, in our society, because we're only seeing like almost like a bubble of what is out there, right? We're not seeing the full spectrum because of these algorithms, and algorithms are really, were really an early form of AI technologies, what they were. I think in the early days, these were based on what they termed machine learning capability, so it's learning about your behaviors and your habits, so those were early AI algorithms, or those personalization tools. And so I'm a little kind of dubious of those, and I think we may need to kind of stray away from that and maybe get a little more intelligent on that kind of technology to keep it, you know, kind of aligned with what our interests are, but open to sharing things that maybe potentially could maybe upset us, make us, And this gets back into the later discussion I'm gonna have about kind of AI's impact potentially on free speech and our ability to take in information from the world that is maybe more complete than maybe the algorithms currently wanna deliver to us. And then the other benefit of AI technology is the improved ability for us to get access to the content for people that are maybe hearing impaired, visually impaired. It can also improve your SEO, which means that you can get discovered more based on deeper levels of data that you've been able to extract from transcripts and summaries and things like that that are much richer and maybe metadata. And keywords, those are important, or keyword phrases, because I think a lot of podcasters get caught in this trap of they do an episode, they make an episode, and then they get caught up in this deep well of post-production and publishing that can take longer to do than to produce the episode in the first place. So that's where this technology can help with that, is to speed up that process, maybe make it better, maybe find better keywords to use because the AI is smart enough to know what keywords are drawing people's attention. So I don't know that we're entirely there yet, but I think that is probably going to come as we merge these sources of data and technology in the platform. So, and then really the next really big one that I think we're gonna see increasingly happen is, Language translation. So AI is gonna be able to take any piece of content, could be audio, could be video, and could translate it. I mean, so YouTube has already said that they're planning on doing this. I don't know if they fully rolled it out yet, but were they gonna take any specific language content that could be from, you know, it could be English, it could be Japanese, it could be whatever, and they will automatically translate that into all the other languages of the world. And be able to actually change the video too. So if you're doing a video, it'll actually align your lips with this new translation. It'll sound like your voice from what I understand, but it won't— not everything that goes out to the world will be in your native language. So that has really the ability to just unlock global consumption of content. But that also opens up, you know, if you're producing content in the United States, that also opens up the Pandora's box to people in the United States that just speak English, maybe, being given access to content that's produced in other languages around the world. So it's going to create a much bigger pool of content that we can all consume from. And we already have kind of like closed captioning experiences on these platforms already, so that can be a translation as well, but I do believe that the audio and the video is going to be manipulated to create a better kind of experience that really makes all of our content global, really, and not required if you're an English-only speaker like I am. That wouldn't limit the consumption of your platform just to English viewers or watchers. However you want to say it. And the other big benefit of AI is data-driven insights. So as you look at the metrics that you get around audience consumption and how far they listen or watch, where they're listening from, who they are, if we can start matching up these sources of information about our content where people drop off. Now granted, we have access to a lot of this information right now, but most of it isn't being run through an AI algorithm to create like a report for us. That's happening to some degree in YouTube today. I get an email from the YouTube platform giving me like a weekly report of what the activity of my channel is on the platform. I think that's terrific, but it it still doesn't have the deep, I don't think, interpretation of what the data means, right? And doesn't help us make better choices around maybe what we're doing in our content or what we're not doing in our content to make us better content creators. So I think that is the power. So it's in-depth viewer insights and that helps the content creator, but it also helps me or you as a content creator be a better content creator as well and cater to the needs of the viewer, which we all want to do. And then also AI could give us the ability to do almost like audience research to some degree to find out who these people are. Now, granted, we're starting to traipse into areas that inflict on privacy. So I think we, we have to be conscious of, of going too far down this path. Now granted, if you're working with a proprietary platform like a YouTube, they get deeper user metrics than, let's say, a podcast hosting platform that doesn't get that data from the consumption side of podcasting. All they get is basically downloads, where the downloads came from, what countries, what devices, and what kind of consumption platform. So it can tell you if it was watched on Spotify or is watched on Apple Podcasts or something like that, or listened to, but that's about as deep as it goes. Where some of the proprietary platforms, this is what we're seeing happen, and I believe that's probably where the AI is going to come in soonest, and to be able to analyze a much deeper set of data that's been able to be be picked up from the actual player side of the consumption. So because podcasts really don't have access to really good quality player information, like clicking, starting, stopping, how much did they watch down to the second, which has been around for a long time on the streaming side of distribution of content. And I do think that we're kind of to some degree, maybe heading back to an increased amount of, you know, you're seeing it with YouTube and you're seeing it with these short-form content platforms like Instagram or whatever. Those aren't downloads, those are streams. So we're seeing kind of this growth of that kind of distribution that we're gonna have the access to AI that can coalesce and give us deep reports on that. And then the last couple ones here are around interaction, right, with audiences and, you know, AI chat boxes, so, or virtual assistants or personal agents that can be deployed to either interact with your audience, have, you know, they can almost simulate like they're talking to the host, you, if you're the content creator, and you could potentially give that chatbot, you know, your personality or your kind of AI profile or something like that. As you look at that kind of moving into a virtual assistant, an AI virtual assistant or an AI personal agent, Now, those technologies are starting to be deployed now. You can actually get an app now that will run on your mobile phone that will analyze your digital life, right? And it will give you recommendations, it'll help you book things, it'll look in your email, it has the potential to respond to emails in a certain way. So there's gonna have to be a certain amount of training that goes into these things. So, they can do things maybe while you're sleeping. Now, granted, the question gets back in is, do you want it to be doing things for you is the bigger question. And that's, do you feel confident in it? And this gets back to maybe some of the downsides that I'll get into a little bit further in the show here. But the AI agent automation, you know, like I was saying, they could schedule your calendar based on email communications, based on your priorities. They could write text for you. They could do art creation based on what is asked for of you in your email or in any of the other types of ways that people communicate on the business side to ask for things like, you know, like an email that's asked for like your profile picture and maybe a bio or something like that. This AI agent could automatically respond to those requests as long as they know where that information they can get that from. And is it current? Is it accurate? And that would be up to the AI to help that happen. And then to give you a report on what it's interacted with, what decisions and what things did it take action on. So you're kept in the loop on those processes. So that's a fairly good summary of all of the positive sides of AI technology as we look to, let's say, the next 2 or 3 years. I think much beyond about 2 or 3 years, we get into spectrum of potential with AI that is probably a little hard for us to comprehend and predict right now. I think as we move into the next level of what's called artificial general intelligence, which is AGI, that is when the AI becomes maybe pushing the envelope towards SDI, which is kind of super artificial intelligence. So, you know, the difference between AGI and SDI if that's even the right acronym, I'm not sure, but it's this ability of the AI to have more knowledge than humans do, have more capabilities than humans do. I think we're approaching that right now, but I don't think it has human characteristics quite yet that would be part of the AI once we move into this AGI aspect of it. And I'm certainly not an expert on AGI, 'Cause none of us have had any, as far as I know, unless you're a high-level engineer at OpenAI or Google or something like that, they're probably playing with these models trying to come up with a model that will actually fit into this definition of AGI. So I think we'll see it, but how soon we see it, I think is a huge question mark very few people want to give a definite answer to. Could be a lot sooner than we realize. It could be in the next couple years, um, or it could be next year. Who knows? Who freaking knows, right? Um, all, all we know is AI is coming quickly and, and it's evolving. And I do think that OpenAI is going to be announcing, announcing kind of like a new upgrade 4.5 or 5 or something like that very soon. And it's going to have enhanced abilities that kind of takes things to another level. So I'm going to talk about some of the downsides now of AI and that— take it with a little bit of a grain of salt because it's— these are kind of speculations to some degree and things that we're seeing signs of. In the market right now of things that could happen here. These are just speculative concerns is what they are. And the number one one in the context of podcasting and new media creation is quality, right? Quality and authenticity. Those are the two— I'm not necessarily saying that they're risks,, but there is a concern that we as humans get too dependent on the AI to do too much that then gets in the way of this perception of what's quality. Now, I did do an episode, I believe it was last week, on the perception of quality. So it's, it's this vector too as we get into more artificially created content. How much, how far down that path do we feel comfortable as creators? And how comfortable do we feel as consumers? Do we trust these things? I think it's really easy to have an overreliance on AI-created content and could cause us to get into, at least in the short run, this, you know, this challenge of originality, right? Are we creating very creative human capacity type of content that will align better with human audiences, right? Once we start going down this path of creating AGI or AGI-driven content, does it become too perfect? Does it become less trustworthy? Is the messages that are in them. I'm going to go into this too. It's kind of like the next phase of this. It kind of loses that personal touch, right? It kind of loses that human component to it of expression. Now granted, AI is going to have, I mean, you're already seeing it in some of the cloning, the voice clones that are out there and the video cloning that's being done of humans. And they're getting pretty darn good. I mean, they sound like they're humans. I don't know that they'll ever sound like me, but I have trained a clone of my voice in the Descript tool. I created a clone of my voice over there and it sounded pretty close to my voice, but it still sounded a little bit robotic to some degree. It wasn't really as inflective as my real voice, and it doesn't take into account fluctuations of my throat and my mouth and how I feel at the time, my energy level. It's— that's the variability of humans, is the energy level and our ability to express ourselves based on how we feel at that moment. And it's not always an easy thing for an AI to be able to replicate that unless they have a huge dataset of the things that you've created in the past. And I do believe over time it will have that. But oftentimes these clones are built off of like 15 seconds of your audio or something like that. They can be generated very quickly. And I don't think it has the spectrum quite yet that it would need to be able to really replicate what, you know, using me as an example, my vocal spectrum of talking about various topics under different emotional situations, how I feel at the time, different energy levels, and to be able to navigate that. I'm not quite sure. So that authenticity is really, I think, a big question mark right now. And the other part that I'm going to go down the path is the potential AI bias that could come into play here. Now, granted, if you're feeding it what you want it to say, then there's less likelihood that there's going to be bias in there. But I have seen already in my use of OpenAI and Google's Gemini, that there can be, if you ask it open-ended type questions, or you prompt it to generate something that is reliant on it to provide depth and context, that it will kind of give you a flowery response, right? So if you ask it something that may be a little controversial, it will today, all these platforms will give you an answer that is, I guess, probably a good way to say it is it would be more politically correct, right? I think is the right way to say it. But then if you prompt it, if you question it a little bit deeper and say, well, what about this part of the answer? It's been shared that there is this other component of what you said shared, I think you should add that to, to what your output is. It'll go in and it'll pull a different set of information and put that in there. But you had to specifically ask for that difference of opinion of sorts. So it'll give you the output that it's— I guess it's been prioritized to give. And I don't know if it's based on training data. And that's where a lot of this stuff comes from, is from training data. Right? So training data comes from all sorts of sources, but I think that's really the key. And I know this is a big issue with getting AI to be usable is it has to be trained on information, right? And the question you have to ask is where does that information come from? And is it really is it really information that is true, or is there bias built into the training information that goes in? And I think that there's been some examples over the last year or so where we've seen AI results come back with very biased results, and you have to wonder how it did that. And I think how it did it is because of its training data. It was given data that pointed in a certain direction. And AI is really, at this point, is still kind of like a very complex algorithm that's based on machine learning, but it can unintentionally as well as intentionally introduce bias into information. And like I said, that comes from training data,, but it can also just be coming from what it sees in the world. So if it's going out and it's analyzing all the information that's online, right, the same kind of thing that you would do if you do a search in Google or whatever, it'll give you a bunch of search results. And I'm going to pull up one here in a minute to show you what I mean. But there can be a bunch of different viewpoints shared in a Google search result based on pulling up links to a webpage. And those can take different positions on things. And if the data sources that are out there are coming back with results that lean one way or the other, and maybe those content sources are not telling the truth, right? Maybe some of them are telling the truth, maybe some of them are exaggerating, maybe some of them— so it's gonna have to analyze all those and it's gonna have to prioritize them on what is significant. So, you know, who's actually behind the scenes making content choices here around what's true and what's false? And also, I think the big picture question that I have as I move down this path is I hope that AI will be kind of like a content tribe trust factor here, right, or a tiebreaker, right? Because we live in a very divided information world right now, and that is, there needs to be something that breaks the tie, right, to help us all better understand what's true and what's false, what's the truth and what's a lie. How do we as humans today do that? I think we have to rely on our own intuition. We have to rely on our own ability to research, our own ability to sense human body language and human communication and what's believable and what's not. There's a lot of filters in our— well, there's a lot of filters that we have to apply to the world today. To keep our sanity because of what's happening in the world and how much division is in the world and how much people fundamentally disagree with each other on very important topics. And so how do we break that tie, or how do we break that division to say, okay, you have this opinion, this other person has this opinion, Who's gonna say, well, you're wrong and you're wrong, or you're right and you're wrong, right? And is it possible that AI could do that for us, especially AGI, you know, the superintelligence? My concern is that, are we going to see AI not have bias? Now, granted, what is bias? I guess, is it fair? Is it a lack of fairness? Is it a lack of truthfulness? Is it a lack of inclusivity? It's essentially putting your thumb on the scale and saying, what's true is this over here, and that's a human construct based on our own interests internal bias and our own wish for something to be true. Maybe there's an agenda there. Maybe there's a, you know, I hate to say, but there's a political campaign that needs to be swayed one way or the other. And the creator of the AI, they have a particular bias where they want one candidate to win. So does that mean that they put the thumb on the scale to have everything come back with a result that is more favorable to one side of that divide. So one of the things that has come up here in the last week or so is, and this affects kind of podcasting more directly and online media creation, is I guess a testimony that was given by a host of a Daily Wire program talking about, um, there are algorithms out there that are, um, in their term, kind of brand safety and suitability models that are being applied to content now. And, and these are being applied to YouTube and being applied probably to Spotify, and increasingly we're hearing rumors that it's being applied to potentially Apple as well, where they're analyzing the content for advertising. I don't— I mean, Apple's not really involved too much in advertising, but this technology could be used as a way to determine the type of content that is being produced and ranking it in almost like a nutritional label. If you think about what a nutritional label is, is that there's like a spectrum of how many carbohydrates are in this. So if the carbohydrates are too high, then maybe it's not healthy for you. That's essentially what this GARM does is it ranks content based on what types of content it sees as unhealthy for you, right? And what content is seen as good for you to hear., right? And that's the justification for it. And GARM is a, it's an organization, uh, that stands for Global Alliance for Responsible Media, and it's an initiative of the World Federation of Advertisers. And so we've, there's a lawsuit that, that's been put out there, um, against GARM here recently, um, by X and Rumble, basically alleging that this GARM platform that's being utilized increasingly with the brand safety and suitability platforms has been specifically targeting certain content platforms to demonetize based on content on those platforms. And so there's a whole controversy around this, And I can actually bring up what that GARM is. Let me bring it up on the stage here and I can share that with you. And actually, I wrote an article about this and it says GARM is a moderation rating and ranking of podcast content from more of a historical and past episode perspective. Now, that's what was going on in the past. Now they're looking to build these technologies into podcast hosting platforms, which means that episodes could be scanned with this technology before they're actually published. So this comment that I have in here, the historical and past episodes may not be the case for much longer. So it could get in the way of you even publishing these episodes based on the output that these suitability platforms have for your content. So, and it's, and it could, it's currently only associated with advertisers right now, but I do think that it could come to analyzing if your content is actually suitable to be available on a given platform. So here, let me click on, this website. This is the World Federation of Advertisers. It is basically the global alliance for responsible media launches to address digital safety. So this is kind of what is going on is that they are using these tools to identify content that maybe a brand, a brand advertiser typically is the is the people that are interested in this, could say, well, this content has this and this type of discussion, or they use these kind of words, and I don't feel safe running my ad on those platforms. So that's, that's what those are. And then let me go, go back to this and scroll down to Where— okay, so the source of a lot of this comes from a platform called NewsGuard, and it's actually a, a platform that is, um, this article is about a year old. I don't know if it has a link to it, but this is a kind of a global platform that is, that is working with these content moderation platforms, um, to, to get this technology, the GARM technology, into all, all these content platforms. So, so let me hop out of this and, um, and I can go back to, um, what were we talking about here? GARM's primary goals stated on their their website is to improve the safety, trustworthiness, and sustainability of media environments by reducing harmful content online. So the question that we have here is, what is deemed harmful content? And I think that's the core issue. A lot of platforms, hosting platforms and listening and viewing platforms, have what they call terms of service, right? So when you sign up or you publish your content into those platforms, you basically have to approve their terms of service, right? And the terms of service specify in them what type of content is allowed, how it's distributed, and are there any limitations to what kind of content can be included in that platform. Typically, terms of service have two types of categories of content that are typically not allowed on the platform, and those are content that is deemed hate speech and content that is deemed as violent— violence and provocative content, right, that basically is threatening or potentially harmful to another person, right? So you are basically it very much aligns with the current laws that exist on the books around hate speech that can be linked up with violence of sorts, is what you're— what they're basically doing here. So those are the major categories that are being used to create all of these different kind of targeting type of concepts for with the brand safety and suitability and moderation tools is that there's been an expansion of what the definition of hate speech. So hate speech now in our society, because there's been so much division and weaponization and harm in the perception of various groups of information that now a much broader spectrum of topics are now deemed as hate speech, right? So these fall under platforms' terms of service. Now, what is going on here with GARM being sued by X and Rumble is what's being sued about is that X and Rumble have found evidence that there is an illegal monopoly practice to remove advertising from entire platforms based on free speech in a coordinated kind of conspiracy. Now granted, this has not been proven in court, but this does raise the issue around GARM being used as a content moderation tool, not so much an advertising tool. So this is— they're getting linked up with each other. And so this is only AI-related because some AI tools are being used to analyze transcripts. So basically, the GAARM platform and these tools basically utilize transcripts, which typically come from, you know, AI-generated analysis of the content that's being produced. It's a way to look deep into the content to determine what is actually in the content, what keywords are used, what phrases, what's the context. And that's really where the rubber hits the road here, is this term context. Does the AI that's being used with this GARM rating system, is it basically a, is it utilizing deep context, right? Does it understand the difference between a news program referring to a gun in the context of a news story or a podcast using the term gun to refer to the intent of violence, right? So that's the contextual reference that I'm talking about, that sometimes these algorithms can confuse that and block advertising to— like, the New York Times put out some press releases here in the last few months saying that these brand safety and suitability models have limited their ability to monetize their news programs because they were triggered by keywords that were in the news that triggers these algorithms to say, this is unsafe content because they use the word gun. And there's a bunch of other words that can potentially trigger that. And I do think that the more advanced brand safety platforms are getting more contextually relevant because they're feeling pressure to get better at doing this. So they don't create a situation where the New York Times gets demonetized because they're reporting on a story and they feel like they can actually report the story accurately and not use terms that maybe they have to in order to tell the story, um, where right now some of them may feel like they can't use the term of a weapon or a gun or something like that to, to refer to something in a news story, uh, because they may have fear that they're going to get demonetized. So that's the risk here, um, and it's, it's, it has a slippery slope to it too. I mean, this is a path towards censorship, and it may limit and stifle free speech, people being able to talk about things freely. If we're talking about things like gun control or something like that, you should be able to talk about those things without being flagged as a hateful piece of content or something like that that's dangerous or threatening violence. So it needs to be able to analyze that information and make the proper choice here. So, you know, it can also be used to disproportionately target viewpoints and political orientations too based on that same analysis, right? So if a person uses the term Biden or they use the term Trump or they use the term Harris or something like that, that could trigger the algorithm to say this is not a safe place to run your ad. If the company, the brand, is actually buying advertising on that, that they may align as a corporation with one party or the other. That's the real danger here. The other danger of this is this concept of self-censorship on the part of the content creator. So, you know, after a while, once you get into utilizing these technologies and these platforms, you start to understand because you've seen the output that comes from it. And, and I wanted to play a clip, um, from Conan O'Brien who actually commented on this brand safety stuff because he, he's gotten caught up in this as well. So let me go ahead and pull that up and I can play that for you here so you can hear him talk about how he feels like he needs to maybe modify his content because of these brand safety platforms. So let me pull this up. So let's see here. Okay, I need to get to it to start it here.

Okay, here we go. All right, so this is, uh, this— I think this is mostly— I, I believe— what is this now? Explain what's happening. Can you take a second? Yeah, so this is our brand brand safety analysis that's done, I think, primarily by like AI and transcription. So transcription is done on the podcast and then some algorithm decides how brand safe the show is. This is real. This is real. These are real ratings on the show. And so brand advertisers, when they're deciding whether they want to advertise on a specific podcast, will look at these ratings and decide if the show is safe for them to be associated with. Okay. All right. So medium adult sexual explicit content. That seems low. Is that— that seems very low.

It says it's only a 4%.

Dicks and stuff. Oh, but down here, obscenity and profanity, it's in the red for 83% and 71% high risk. Yeah. So we're killing it with obscenity and profanity. Maybe this is just all our clinical talk. Whenever we say, you know what, I don't like vagina instead of cockaroo. Hate speech, 7%.

That's you.

Acts of aggression. Oh, acts of aggression. That's you. I don't like— because there's no hate speech here, but there is me routinely trying to take a swat at someone.

You're also crime and violence because you say the word murder on a regular basis.

Yeah, I mean, you're 11% high risk. Wait a minute. Military conflict. Do you want to just run down what the sheet says to give us kind of an overview? Well, adult sexual explicit content, 4%. Alcohol, 2%. These are risks. High risk, 0%. For suitability. Crime and violence, 12%. Wow. Death or injury, 3%. Relatively low. Remember one time we did a segment about robbing a bank and what would our masks be? So yes, things like that are factored in. Okay, uh, death or injury, 3%. Epidemic, 3%. Hate speech and acts of aggression, 7%. It's a little high for me. Illegal drugs, 1%. We don't really talk about drugs a lot. Yeah, we do. Yeah, we do.

But those aren't— smoking that—

well, they're not illegal. Everything you do is legal. Yeah, that's true. Uh, obscenity and profanity, 83%.

Wow.

Geez. Geez. Christ, that's a high number. Fuck. Tobacco, e-cigarettes, vaping, 0%. Well, now it's going to be in there though. Does it— will it factor? Oh yeah, now it's going to be in there. This is a category violation of human rights. Oh, that's just us working for you. Yeah, that's— that's you. It's a violation of human rights to work here, and that's 1%. Um, I don't understand military conflict. Yeah. What is military content? Well, we talk about World War II.

Yeah.

I mean, the term Nazi will pop up every once in a while. I think it's, I think it is like just talking about World War II, basically World War II facts, trivia. Stalin comes up a lot.

Yeah.

But is this a concern that us discussing these things right now will bump these numbers up? I hadn't thought of it, but yeah. And then next year when we do this again, it'll compound. I think so. Boobies, boobies.

Nazis.

Vaginas. But I think so. The truth is like, this is more. Sona made a really, really important point, which is like, you were on TV for 28 years. Your brand is established. Yes. You had advertisers on that show. Yes, I did. It was a brand-safe show. You're brand-safe talent. This really is more done for talent that is maybe lesser known to, you know, advertisers if they're— when they're making the decision if they want to advertise on a specific show. I see. But I did think that you would just find it interesting that we're being analyzed. What do you take away from this? I actually take away that, um, there is a high sensitivity among advertisers now, higher than ever, it seems, to be associated with brand-safe content. So I think for us, we should just be monitoring your language primarily, everyone's language here. Um, and thinking of these categories every time we go into an interview or do a segment.

No, but, but, um, Uh, I'm told not to do something, I just want to do it more.

Adam, you're coming dangerously close to being the man censoring the artist here. So also, I don't think I use profanity. I mean, cockaroo, is that a— is that really profanity? I don't know. I love that answer. I love the way you said it.

I don't know.

Well, there was talk of gerbils. Is all kind of— unless you really get into what it's all about. I don't think this thing understands context either. Yeah, it doesn't understand context. So talk of gerbils is quite a lot— that could be a children's story, but it, it might take gerbils and think Joseph Goebbels and make it a Nazi thing. No, that's Goebbels. Well, they know it's written in a transcript. It doesn't know. It does. It knows the difference. They're spelled completely differently. It's ridiculous. I think the state of the con— uh, the state of the conflict— I think the state of the show Numbers are going up on military conflict. I think the show is doing well. We are marching forward. We're like the allies pushing.

All right, well, let's hop out of that. I thought that would be pretty instructive to all of us to see that clip. I think it really gives good context to what I've been talking about here. And it's not just me that's talking about this. It's, it's how a content creator, specifically a comedian, would look at this situation presented to them. And I think it was really interesting that they created this episode because it really showed clearly what is, what is happening here and how it can impact a content creator or a podcaster, right? And these are things that, you know, it's not like I'm trying to take down any platform or I'm trying to slander anybody for doing something. That's not my intent here. I think what my intent is, is let's make sure that we all as an industry know what's going on here and know its potential risks, its potential advantages, there is a useful use case for this. I just wonder if we're fully factoring in where we want to go as a medium, right, and don't allow this to become something that basically, you know, takes away the ability of this medium to be unique and to offer a spectrum of content that attracts listeners because they view it as authentic. If we're put in a position where we have to censor everything that we say or we have to be careful about what words we use, you know, it's going to be very difficult to create a show that is appealing to anyone. We can wind up neutering the shows so much that it's not consumable content, and is that Is that what we want? I think is at the end of the day. So, so you could see there that gets back to what is also saying about self-censorship as a content creator. Do we, as we think about their own shows that we do, I mean, I don't use profanity in the shows that I do. I don't use any kind of speech, but that's because that's not the type of content that I produce. So that's not what I'm trying to do here. But if you're a stand-up comedian or you're a comedian that's known for that type of content, to have this come in and say, boy, all these things that you've been doing in your content for all these years is all of a sudden now not okay to do anymore. That's, that's a— I don't think that's a good thing, especially when those many of those advertisers have been perfectly fine with what he's done in the past. So what's changed is the question that you have to ask, and why is this necessary? So anyway, I didn't want to spend the whole show on this, but it kind of gives you an idea of what I'm talking about here. So to kind of move on from this topic, there are some ethical and legal issues that come into play when you're utilizing AI technology in podcasts, and one of them is transparency around clear guidelines on disclosing the use of AI in the creation process. Now I think we are being faced with an increasing challenge of deciding what type of use of AI in our content creation process is okay to not disclose and what other types of use cases for AI to create content must be disclosed to the audience. So I think we're going to run into this challenge increasingly as AI gets better and as people increasingly use AI to generate their content and create every aspect of it. And I think that day is coming. I'm not, I'm not kind of thinking that it's not. I think that there are early platforms that are already starting to come out that would But basically you type into some sort of a box in the software that you're using to say what kind of show you want to create. And it will say, you know, it'll generate the artwork, the show art, it'll generate the description, it'll generate potentially if you've trained it with your voice, you just tell it what topic you want to talk about. Maybe give it a paragraph, say I want to produce an audio episode that talks about XYZ, and you know, you can probably give it a little deeper prompt about what you want to talk about. Maybe you come up with an outline or almost like an article, a short article on what topics you want to talk about, and that prompts the AI to go and say, well, I'll just use your voice. I'll generate a 30-minute episode just like what it does today with ask, asking AI to write an article for you. It's reasonable to think that the AI will automatically generate that audio for you. And then once it generates it, then maybe you can review it, make sure it sounds all right, it's okay with you, and all you click is a button and it does everything that's necessary to publish it. It'll write your title, it'll write your show notes, or it'll do your chapters. It'll do your, all the kind of sample clips that you need to have to get out to the short-form video content platforms, 'cause it could potentially generate all the short-form video and the long-form video that's necessary to publish out to YouTube and Spotify and into a podcast feed or whatever, and it's all generated based on the AI. And I, I see that day coming. And as a listener and as a consumer of this content, um, are, are we going to trust it? Are we going to want that to happen? Um, I can definitely see some advantages there. I think we're starting to see it happen a little bit more on the YouTube side where there's these faceless videos that are being generated by AI that people are publishing. And then the other aspect of concerns here too is privacy concerns around access to our information. I'm kind of moving on from the kind of creation side, but how AI, once you start getting into these deep AI agents and once you're really kind of having it do everything, it really needs to get access to a lot of deep information about you, not the least of which is your ability or its ability to generate your voice. So you're going to have to give authorization to it for it to use your voice. At least that's what I hope happens here. But there's the potential on the other side of this that this technology could be used to replicate your voice and used in other things in other ways that may be potentially fraudulent. So the technology has a good use as well as a dangerous use at the same time. So that's a little bit of the balance here on the AI technology. And the economic impact, I think, is another big factor of AI in podcasting or new media creation, is the potential of job displacement., right? Uh, work, work displacement. Because if we do have podcast publishing creation platforms that are all AI-driven, all automated, then it basically is going to put out of work all the podcast editors, all of the artists, all of the producers, all of the audio engineers that maybe have jobs in the podcasting space today. And I think in some ways that's the last thing we need in the podcast industry is more people getting put out of work. So, or being displaced by technology. And I think we're already seeing the early stages of that. And it goes beyond that too. It's script writing, there's marketing people, there's PR people that are involved in podcasting now. They're I mean, I've never seen so many careers. There's booking agencies that are out there that help you book guests for your show. So there's lots of people being employed in the podcasting space that potentially AI could basically eliminate entirely. And unfortunately, I see that as a very real possibility at this point. I don't know what the timeframe on that is. I think what we have to decide as consumers of content is that maybe we have to decide as humans what we want to allow. You know, at the end of the day, you know, are we going to vote with our ears and our eyeballs for content that's produced via these methods, or are we going to vote with our eyes and our ears to only consume content that is not using all this technology. So maybe it's okay to use some of it, but maybe it's not okay to use all of it for everything. And so I'm kind of pulling back on that. And then the other element is cost barriers. The, you know, another element of the negative impacts on using AI in podcasting is some of these AI tools can be expensive. They're not you know, easy to build. They do take engineering ability to create. Now granted, even that's getting easier and easier to do too. So, you know, and that could create a gap between, you know, new creators and existing creators. It could be that the existing creators, the ones that are already successful, will increasingly use the really advanced AI tools because they can afford it. And the new podcasters or the ones that are just getting started have an unfair disadvantage here around the type of tools that they can be to be most efficient. Maybe using everything with AI to do pre-production and post-production and marketing. And distribution using AI, but the actual content creation part remains with humans might be the real answer. And then I think the kind of last big one here that I wanted to cover, and we can, we can kind of move the show towards the end here, is all of us losing our skills if we have too much technical dependency on AI, then the skills of podcasting kind of start to wither away. Now granted, AI could become very, very talented at this and really learn to optimize every part of the podcasting process. So if it can be used at the beginning and the end, of the process and the humans remain as the creators of the content for at least most of the content. Doesn't have to be 100% of it. I know that there's a lot of pressure right now around people using AI to dynamically create like ads that could be used with programmatic platforms. So, I think that's a possibility that that could evolve into being something that people can accept. I know I had a guest on the show, Glenn Rubenstein, who works for Adopter Media. He was kind of maybe like a wait-and-see kind of attitude around AI-generated ads. You know, can they be that good? To, were they not distinguishable from a human-created host read or something like that, I think is the big question. So I hope I didn't depress anybody around AI tonight. And I see that there hasn't been a lot of comments. I would love to see some more questions and comments. About the topics that I've talked about tonight. They are not easy ones to absorb. And Dave, thank you for the comment, um, uh, thank you for the comment about looking good and hope all is well. Yes, everything is all, all good. This is a pretty serious topic tonight and just wanted to make sure that everybody kind of heard maybe a different thought on this. I'm not sure that this topic has been talked about at this depth in any podcast that I've heard, um, short of that Conan O'Brien one, which I think was a really good example. And I think that episode has been seen quite a few times, so there are people that are aware of this. But we have seen increasing interest in the podcast medium of adopting these these platforms. And AI is coming at us really fast, and we all have to learn from it, adapt to it. I know I am. I'm doing my very best to learn as much about these AI platforms as I can and find out what they're good for and find out what they're not good for and experiment and things like that, which I I've been doing for the last, you know, almost probably about a year now. I've been playing around with these things and seeing how they work, but I, I still don't think that we've seen the, the full spectrum of what they're capable of doing, I think, as we look to the future. And so it's, it's going to be very, very interesting to see how all this plays out. And, and I want to just tell you as I wrap this show up, um, I appreciate you being here tonight. It looks like we've had some, some viewers. I'm not sure how long people have stayed listening, um, to the show tonight, but we've had a lot of people pop in here and take an interest in this topic, which I do think it's a, it's, it's a very important topic. And, and if you wanted to get more of these episodes from me Um, this is a good website to go to. It's just podcasttips.blueberry.net. If you wanted to see the past 35 episodes that I've done of this series, it's a great place to go to actually see that. And if you wanted to go to my, my website, robgreenlee.com, you can get access to this show on that website too. And then all my other podcasts that I do, I'd also do another show called the New Media Show., and it's, um, actually live on Wednesday nights at, um, 3 PM Eastern time. Uh, you can see that show there. And then here's my contact information, rob.greenlee@gmail.com, if you wanted to reach out with an email. I would love to hear from you. I'm also on X at Rob Greenlee as well. I'm also on YouTube if you happen to be watching this on, on YouTube. There, there's my address. I have a bunch of other shows up there too that I would love to have you check out and, and subscribe to my channel. And, and Podcast Tips is up there. Actually, I did the first like 30-some episodes of this series on these StreamYard channels. I'm, I'm no longer on the StreamYard channels now, so I'm doing this entirely independent on my own now. So I can, I, I can feel a little bit more open to getting into topic areas that maybe could be considered to be a little controversial. I would say that this, this one is, but I've tried to treat it as fairly as I can. If you wanted to check out my, my New Media Show, this is the website address for that. It's just at newmediashow.com. That's like I said, live every Wednesday at 3 PM Eastern. Noon Pacific, and, and we do that every week talking about the industry of podcasting too. So some of these topics are talked about on that show as well, but I don't believe that we've done a show quite as extensive on the topic of AI and podcasting as I did tonight. So I appreciate you being here. Thank you so much, and I hope you have a wonderful evening, and I hope you got something from this. I would love to see comments after the fact about this episode and what your thoughts are. Do you think that what I've shared tonight is a good projection into the future of the implications of AI in our content creator lives, which I'm sure a lot of people that are watching this are content creators, and as a consumer too, you know, there's a lot of issues that we as an industry, as a podcast industry, need to talk about and to, to realize that there are things happening that we need to be aware of and give feedback on and, and, and consider. Do we want those in the future? Because they could be coming whether or not we want them or not, unless we're aware of them. So So anyway, I hope you have a wonderful evening. Thank you so much for joining me tonight on Podcast Tips with Rob Greenlee, and I'll hopefully be back with a new episode next Thursday at 7 PM Eastern, 4 PM Pacific. So thank you so much and have a wonderful evening. Thank you so much. Bye-bye.