WEBVTT 00:00:00.000 --> 00:00:13.480 11 Warnings About Using AI in Content Creation. Yes, that includes podcasting. 00:00:13.480 --> 00:00:17.880 Thank you for joining me for The Audacity to Podcast. I'm Daniel J. Lewis. 00:00:17.880 --> 00:00:23.160 Artificial intelligence or AI has made huge leaps in abilities within a very short time. 00:00:23.160 --> 00:00:28.020 It was only a few years ago that I felt on the cutting edge teaching how to use AI tools 00:00:28.020 --> 00:00:34.880 like Jasper, originally called Conversion AI, and even Jarvis. Even before ChatGPT 00:00:34.880 --> 00:00:40.160 was released, I was playing with AI tools. Now AI has become so prominent that it's 00:00:40.160 --> 00:00:46.800 almost surprising if a software company of any size is not offering some kind of AI-based 00:00:46.800 --> 00:00:52.280 solution. While inflation has skyrocketed the prices of almost everything, the cost 00:00:52.280 --> 00:00:58.500 for accessing AI has significantly dropped. When I first started using AI, a good plan 00:00:58.500 --> 00:01:05.720 with access to only one central AI system and very limited abilities cost $99 a month. 00:01:05.720 --> 00:01:11.600 But now you can use a tool like Magai to use a whole bunch of different language and image-based 00:01:11.600 --> 00:01:17.960 AI tools starting at only $19 a month. Now, full disclosure, I am an affiliate for Magai, 00:01:17.960 --> 00:01:22.960 but I'm also a customer. In fact, my yearly subscription just renewed. But if you purchase 00:01:22.960 --> 00:01:27.400 through my link on my website, that is an affiliate link, I do earn a commission from 00:01:27.400 --> 00:01:31.180 that. But I recommend things I truly believe in, regardless of earnings, and I'm a customer 00:01:31.180 --> 00:01:36.760 myself. All this AI potential means we need to quote the line from Spider-Man, "With 00:01:36.760 --> 00:01:42.280 great power comes great responsibility." And thus why I want to share these warnings 00:01:42.280 --> 00:01:48.840 with you to advocate for a responsible use of generative AI, large language models, machine 00:01:48.840 --> 00:01:53.720 learning, or whatever you want to call it. And these terms are often used interchangeably. 00:01:53.720 --> 00:02:00.560 These warnings apply to all kinds of content creation, not only podcasting. It could be 00:02:00.560 --> 00:02:05.000 blogging, it could be posting on social networks, it could be making videos for YouTube, which 00:02:05.000 --> 00:02:09.240 is not podcasting. I'm still going to do an episode about that soon. And in case you're 00:02:09.240 --> 00:02:15.320 wondering, I did not use AI to create any of this content. But I might be using some 00:02:15.320 --> 00:02:22.720 AI to transcribe or help me market this content. In fact, I did ask an AI a question related 00:02:22.720 --> 00:02:28.120 to this content just out of curiosity to see what the AI gave me. And I looked at the results 00:02:28.120 --> 00:02:34.760 and thought, okay, that's content, yes, but that doesn't communicate what I want to communicate. 00:02:34.760 --> 00:02:39.400 So I completely ignored everything it gave me and made my own content, which I was doing 00:02:39.400 --> 00:02:45.540 anyway. And so no AI used in the making of this content. And a little aside too, most 00:02:45.540 --> 00:02:51.480 of these warnings I'm going to share with you apply to generative AI, but not what we 00:02:51.480 --> 00:02:59.320 could call repurposing or enhancement AI. I want this to be clear. Because if you are 00:02:59.320 --> 00:03:04.040 using an AI to enhance what you do, that's different from this. I'm talking about using 00:03:04.040 --> 00:03:12.600 AI to essentially create something from nothing. I still think AI can be a great assistant 00:03:12.600 --> 00:03:20.560 on your content. And the emphasis there is on and your content. Not creating it for you, 00:03:20.560 --> 00:03:26.320 but doing something with the content you already created. For example, processing audio or 00:03:26.320 --> 00:03:31.640 video, clipping excerpts, suggesting marketing approaches, improving how things communicate, 00:03:31.640 --> 00:03:37.920 repurposing, and more. All of these things start with your intelligence. And then the 00:03:37.920 --> 00:03:41.760 AI works from that. But I see most of these warnings that I'm going to share with you 00:03:41.760 --> 00:03:48.960 as applying solely to generative AI, or when you start with nothing but a prompt. So now 00:03:48.960 --> 00:03:52.160 onto the warnings. If you'd like to follow along with the notes, they're a simple tap 00:03:52.160 --> 00:04:00.480 or swipe away inside your podcast app or go to theaudacitytopodcast.com/aiwarnings. 00:04:00.480 --> 00:04:06.320 Warning #1. Undisclosed use of generative AI can get you in trouble. YouTube, social 00:04:06.320 --> 00:04:11.080 networks, and lots of other websites and platforms are starting to require you to disclose whenever 00:04:11.080 --> 00:04:16.400 you're putting out content generated by AI. And I think this is a good thing to do as 00:04:16.400 --> 00:04:21.960 it helps the potential audience know what kind of quality to expect. Even for things 00:04:21.960 --> 00:04:27.380 like podcast transcripts, it's good to disclose whether AI was used to transcribe the audio. 00:04:27.380 --> 00:04:31.400 As I mentioned in my previous episode about using podcast transcripts, and the link to 00:04:31.400 --> 00:04:36.360 that is in the notes for this episode, someone on your podcast might say "I love to lil 00:04:36.360 --> 00:04:42.280 puppies" but the AI might transcribe it as "I love to kill puppies". Sometimes 00:04:42.280 --> 00:04:48.380 even omitting a single word can drastically alter the meaning. For example, imagine accidentally 00:04:48.380 --> 00:04:54.840 omitting the word "not" in a sentence like "I'm not guilty". This doesn't 00:04:54.840 --> 00:05:01.680 necessarily mean you must disclose every time you use AI in any capacity. Like you should 00:05:01.680 --> 00:05:05.660 be disclosing whenever you're compensated for anything you talk about. I've done 00:05:05.660 --> 00:05:10.520 a whole episode previously about disclosing when you're compensated. Go listen to that 00:05:10.520 --> 00:05:14.840 episode, it's really important for you. Link in the notes. But you should be aware 00:05:14.840 --> 00:05:20.740 of the requirements of the platforms and seek to always be above reproach. Don't seek 00:05:20.740 --> 00:05:26.920 to see what you can get away with, but always try to be above that, and far above it so 00:05:26.920 --> 00:05:31.400 you never have to worry about if the rules change. And if you're concerned about how 00:05:31.400 --> 00:05:36.760 it might affect your reputation if you disclose every time you use AI, then here's a radical 00:05:36.760 --> 00:05:42.240 thought – maybe don't use AI. And more on that in point 11. 00:05:42.240 --> 00:05:50.160 Warning number 2. AI often hallucinates facts and citations. ChatGPT, Claude, Grok, Gemini, 00:05:50.160 --> 00:05:56.760 and all the text-based AIs we know are also called Large Language Models or LLMs, which 00:05:56.760 --> 00:06:01.480 is what I'll be using to refer to these types of AIs for most of this episode. And 00:06:01.480 --> 00:06:06.360 I think that's a much better term anyway, because they're not actually intelligent. 00:06:06.360 --> 00:06:12.720 They're simply good with language. This is why you'll often see LLMs write something 00:06:12.720 --> 00:06:21.080 that grammatically makes sense, but is conceptually nonsense. In other words, LLMs know how to 00:06:21.080 --> 00:06:26.640 write sentences. For example, I sometimes like to ask AI "Who is Daniel J. Lewis?" 00:06:26.640 --> 00:06:31.240 I don't do that because of any kind of ego complex, but because I'm an interesting 00:06:31.240 --> 00:06:38.520 test subject for LLMs since I am a partially public figure, but I also have a name very 00:06:38.520 --> 00:06:45.840 close to a celebrity, Daniel J. Lewis. Thus, the responses LLMs give me often conflate 00:06:45.840 --> 00:06:50.680 the two of us (a mistake I wish my bank would make) and I've seen responses that both 00:06:50.680 --> 00:06:56.920 describe me as a podcasting industry expert and highlight my roles in There Will Be Blood 00:06:56.920 --> 00:07:03.320 and The Last of the Mohicans. I'm sure I'm not helping any LLMs scraping my content by 00:07:03.320 --> 00:07:09.200 writing that information together in close proximity like that in my notes for this episode, 00:07:09.200 --> 00:07:11.880 but that's where people have to look over this. 00:07:11.880 --> 00:07:17.160 So for anything an AI or an LLM writes for you or creates for you or whatever it is, 00:07:17.160 --> 00:07:23.440 I urge you to fact check it. I've even seen some responses completely make up citations 00:07:23.440 --> 00:07:29.720 that don't exist. So even if it says "According to CNN or according to such and such study 00:07:29.720 --> 00:07:36.400 by such and such doctor, this is the thing" look that up and if you can't find it, don't 00:07:36.400 --> 00:07:42.080 use it. Make sure you fact check. Make sure you check any citations if it gives you one, 00:07:42.080 --> 00:07:46.960 even if it looks like a valid URL, visit it. If it's a broken link and you can't find 00:07:46.960 --> 00:07:50.000 the replacement, do not use it. 00:07:50.000 --> 00:07:56.920 Warning #3. AI lacks humanity. From the moment of conception, you have been a unique being 00:07:56.920 --> 00:08:03.440 of tremendous value and potential with unique DNA, unique experiences, unique thoughts, 00:08:03.440 --> 00:08:10.360 unique emotions, and more. Like a snowflake, there will never be someone or something exactly 00:08:10.360 --> 00:08:17.040 like you. Not even an AI trained on all of your content. Like a comedian recently had 00:08:17.040 --> 00:08:23.800 done to him after he had died, it still wasn't him and it wasn't as funny as he was. That's 00:08:23.800 --> 00:08:27.600 what a lot of people were saying. That even led to a whole lawsuit. 00:08:27.600 --> 00:08:36.120 AI is not an actual intelligence and I believe it never will be. And AI will never be human. 00:08:36.120 --> 00:08:43.960 But you are. You can feel, express, and empathize through emotion. You can question, explore, 00:08:43.960 --> 00:08:48.180 change your mind, or change other people's minds. You can create things of great beauty 00:08:48.180 --> 00:08:55.400 and originality with no outside prompting. And it's because of this that I think AI 00:08:55.400 --> 00:09:01.260 can never replace you. While it might have better skills than you in some areas, it will 00:09:01.260 --> 00:09:08.000 never beat the quality and personableness (which is an actual word) that you can offer. 00:09:08.000 --> 00:09:15.480 #4. AI-created images can be humiliating. AI image models have produced some hilarious 00:09:15.480 --> 00:09:22.240 or nightmarish results and lots of things that are physically impossible. Like how AI 00:09:22.240 --> 00:09:28.200 can hallucinate facts and citations, it can also make images that look real until you 00:09:28.200 --> 00:09:32.180 actually pay attention to the details. And sometimes you don't even have to look very 00:09:32.180 --> 00:09:37.440 closely to realize how ridiculous it looks even though it looks like a real photograph. 00:09:37.440 --> 00:09:42.040 Like the classic thing of a hand shaking a hand that's shaking another hand. Where 00:09:42.040 --> 00:09:46.160 you see just this arm with two hands coming out of the arm in the middle. It looks like 00:09:46.160 --> 00:09:50.900 an actual photograph because of the quality of it, the lighting, the shading, all of that. 00:09:50.900 --> 00:09:57.040 But you know looking at it, that's impossible. That's not real. That's you using your 00:09:57.040 --> 00:10:03.360 intelligence that is more intelligent than the "artificial intelligence". Emphasis 00:10:03.360 --> 00:10:07.600 on the artificial part there. And here are some things, lest you think this is like old 00:10:07.600 --> 00:10:12.240 models and I have seen some newer models that do things a lot better, certainly. So the 00:10:12.240 --> 00:10:16.440 problems that you see and that gets made fun of. And I love the minions in a trailer for 00:10:16.440 --> 00:10:21.120 Despicable Me that's coming out that makes fun of this whole AI thing. It makes total 00:10:21.120 --> 00:10:26.220 sense. And also Babylon Bee has done a funny thing about what's behind AI. But on current 00:10:26.220 --> 00:10:32.480 image models, as of April 2024, I've still seen some examples of problems I've seen 00:10:32.480 --> 00:10:37.520 from images that either I've generated or I've seen other people generate with current 00:10:37.520 --> 00:10:44.360 generation AI models. Backwards hands, limbs that seamlessly merge into the surroundings, 00:10:44.360 --> 00:10:49.480 misspelled text that you might not notice unless you try to actually read it or you 00:10:49.480 --> 00:10:54.900 pay attention to it, device parts that disappear into nowhere, placements that are physically 00:10:54.900 --> 00:11:01.360 impossible, broken, slanted, or curvy lines that absolutely should be straight and connected, 00:11:01.360 --> 00:11:07.320 incorrect size ratios between people or other objects, and more stuff like that, both in 00:11:07.320 --> 00:11:12.440 picture form and especially in video form. Watch out for these things and everything 00:11:12.440 --> 00:11:16.200 else too. And it's not really like I can make a list of here are the 10 things you 00:11:16.200 --> 00:11:22.880 should check every image for because you just have to use your intelligence to check the 00:11:22.880 --> 00:11:27.560 artificial intelligence. So for any image you generate or that someone else gives you 00:11:27.560 --> 00:11:33.440 that they might have generated with AI, look at it very carefully to ensure everything 00:11:33.440 --> 00:11:40.240 about it makes sense and isn't simply a pretty, but embarrassing, combination of pixels. 00:11:40.240 --> 00:11:45.040 For this reason, you might actually want your image AI to make artwork that is obviously 00:11:45.040 --> 00:11:51.960 not photorealistic because then you can get away with some of these things that aren't 00:11:51.960 --> 00:11:56.640 correct. It's like the uncanny valley thing that some people talk about. Like the first 00:11:56.640 --> 00:12:02.960 time that I watched the movie that stars Tom Hanks and Tom Hanks and Tom Hanks and Tom 00:12:02.960 --> 00:12:08.160 Hanks and Tom Hanks and Tom Hanks is basically everyone in the movie, Polar Express. I remember 00:12:08.160 --> 00:12:13.920 I was super into 3D animation at that time and doing some of my own 3D animation work. 00:12:13.920 --> 00:12:19.720 And so this movie was heralded as this is the best 3D animated movie yet. And it is 00:12:19.720 --> 00:12:26.000 so realistic. I watched it and I was freaked out because the facial animations just weren't 00:12:26.000 --> 00:12:33.000 correct and the eye movements were really weird and certain stuff that even as I, a 00:12:33.000 --> 00:12:39.240 very, very junior, inexperienced animator could tell this just doesn't look right. 00:12:39.240 --> 00:12:44.680 It's because they tried to make it look realistic that it looked really bad. If they 00:12:44.680 --> 00:12:51.160 instead went like the Pixar route, which has always looked cartoonish in their original 00:12:51.160 --> 00:12:54.880 movies, their newer movies are so much better and technology's gotten a whole lot better. 00:12:54.880 --> 00:13:00.400 But when you embrace something like that and you don't try to make it look realistic, 00:13:00.400 --> 00:13:06.180 then you get a little bit more artistic license. Like it's totally okay for heads to be disproportionately 00:13:06.180 --> 00:13:12.140 larger than they should be for adults or children or eyes or smiles or expressions or anything 00:13:12.140 --> 00:13:15.680 like that. Animators know this kind of thing that they can get away with and that's totally 00:13:15.680 --> 00:13:21.800 okay. You can do that same kind of thing with AI generated images when you are not trying 00:13:21.800 --> 00:13:27.280 to make it look photorealistic. But anyway, even if you do make some kind of abstract 00:13:27.280 --> 00:13:32.960 or cartoonish or illustrated or 3D looking or whatever it is kind of image, make sure 00:13:32.960 --> 00:13:38.160 you check it. Check all the limbs, check all the text, check the eyeballs, all of that 00:13:38.160 --> 00:13:44.080 stuff. Look very closely at it because it can be really embarrassing if you're putting 00:13:44.080 --> 00:13:50.240 this bad image out there along with content that you're trying to get accepted as professional 00:13:50.240 --> 00:13:55.440 content. And I see this happen quite frequently with businesses, very large businesses that 00:13:55.440 --> 00:13:59.920 I think should know better and yet they make these mistakes. Sometimes want to make fun 00:13:59.920 --> 00:14:04.280 of them and maybe I'll make a website specifically for that or maybe someone else will. Maybe 00:14:04.280 --> 00:14:09.640 you should make a video podcast where you make fun of the bad AI stuff out there. That 00:14:09.640 --> 00:14:15.480 could be really fun. But moving on to warning number five, AI is biased because it was fed 00:14:15.480 --> 00:14:22.580 biased content and programmed by biased people. Now big disclaimer here, the following is 00:14:22.580 --> 00:14:30.560 not intended to push a political or moral direction but just to expose some facts. So 00:14:30.560 --> 00:14:35.760 that said, let me get into this. Please keep that in mind. I'm not trying to push a particular 00:14:35.760 --> 00:14:42.040 political or moral perspective here. Most LLMs lean a particular political and moral 00:14:42.040 --> 00:14:48.400 direction because they were trained with content that leaned that direction. Thus, even if 00:14:48.400 --> 00:14:53.940 not intentional, the outputs will often have the same political and moral leaning. Imagine 00:14:53.940 --> 00:14:57.880 it this way, if the majority of content on the internet, especially the most popular 00:14:57.880 --> 00:15:06.080 sites said that 2+2=5, then LLMs trained from internet content would also propagate that 00:15:06.080 --> 00:15:12.600 fallacy because of how much there is out there, how often that lie is spread, and how many 00:15:12.600 --> 00:15:18.640 places are citing that lie, the LLMs would just also spread that. Furthermore, many of 00:15:18.640 --> 00:15:24.600 the companies behind these AIs or LLMs also lean the same political and moral direction 00:15:24.600 --> 00:15:30.520 as the majority of the internet and so they will favor content from that same echo chamber 00:15:30.520 --> 00:15:36.580 and sometimes even intentionally train the AI to push that agenda. For a great example 00:15:36.580 --> 00:15:41.600 of this, or really a shameful example of this, look at the bad images that Google Gemini 00:15:41.600 --> 00:15:48.600 was originally generating, they later fixed it or tried to, even going so far as to render 00:15:48.600 --> 00:15:57.240 Nazis as Asians or Blacks instead of Whites because of diversity, inclusion, and equity. 00:15:57.240 --> 00:16:02.920 And that's why there's a market for LLMs that lean the opposite direction. Even taking 00:16:02.920 --> 00:16:08.080 out the political and moral leanings, I see LLMs regularly put out "myth information" 00:16:08.080 --> 00:16:12.720 (I just love saying that) even in the podcasting space, like saying that podcast ratings and 00:16:12.720 --> 00:16:19.600 reviews affect your rankings in Apple Podcasts. That is not true. But it's been said so 00:16:19.600 --> 00:16:26.760 many times on the internet that LLMs think it's true. It's the 2+2=5 and so many 00:16:26.760 --> 00:16:31.640 people on the internet have said that, the LLMs think that too, and that's the kind 00:16:31.640 --> 00:16:37.040 of information they will give you. And because of that, and sometimes it's just the nature 00:16:37.040 --> 00:16:43.760 of what content is out there which doesn't necessarily reflect the thinking of the actual 00:16:43.760 --> 00:16:50.460 majority or the actual population of people. It's just within that particular echo chamber. 00:16:50.460 --> 00:16:54.880 Keep that in mind and that's why it happens. Sometimes it's intentional, sometimes it's 00:16:54.880 --> 00:17:00.720 just the nature of the beast. And again, I say this not to push a political or moral 00:17:00.720 --> 00:17:06.520 perspective but just to point out that that does exist and that is a fact of how some 00:17:06.520 --> 00:17:08.360 of these LLMs operate. 00:17:08.360 --> 00:17:14.840 Moving on to warning number 6. Content from AI always needs editing. It's because of 00:17:14.840 --> 00:17:21.120 these previous warnings that I come to this one. Edit, edit, edit. I'd love to hear 00:17:21.120 --> 00:17:26.720 your opinion on this too, but I'm starting to think it reflects worse on someone when 00:17:26.720 --> 00:17:32.740 they put out bad AI-created content than if they put out authentic content with typos 00:17:32.740 --> 00:17:37.040 or small mistakes. What do you think? Do you agree? It's kind of the direction I'm leaning. 00:17:37.040 --> 00:17:41.960 For example, you might accidentally write about George Wishington instead of George 00:17:41.960 --> 00:17:47.880 Washington, but an AI might think that George Washington fought in World War II. Which, 00:17:47.880 --> 00:17:52.680 in case you didn't know, George Washington didn't fight in World War II. In this case, 00:17:52.680 --> 00:17:57.480 your typo is a human error and your meaning could probably still be understood by your 00:17:57.480 --> 00:18:02.980 context. But if you put out something that an AI hallucinated, then people have to wonder 00:18:02.980 --> 00:18:08.880 if you're actually that misinformed, also known as stupid. 00:18:08.880 --> 00:18:14.160 Number 7. AI-generated content raises copyright concerns. In the United States of America 00:18:14.160 --> 00:18:18.800 and some other countries, anything you create is considered immediately and automatically 00:18:18.800 --> 00:18:25.400 protected by copyright. And thus, you reserve all rights to it. That's right, it's really 00:18:25.400 --> 00:18:31.180 not necessary anymore to write "all rights reserved" (at least most of the time). But, 00:18:31.180 --> 00:18:36.760 you also share or forfeit some of your rights when you consent to using some tools or publishing 00:18:36.760 --> 00:18:41.240 through some platforms. For example, most places have clauses in their terms of service 00:18:41.240 --> 00:18:46.600 that allow them to use the content you provide, in whatever form that is, in their own marketing 00:18:46.600 --> 00:18:52.700 materials. This could be as simple as your podcast cover art, visible, with 999 others 00:18:52.700 --> 00:18:57.160 on a grid image for the app's homepage. You have to grant them permission for that. 00:18:57.160 --> 00:19:00.760 They're not going to do that automatically without your permission. You're granting 00:19:00.760 --> 00:19:04.680 them that permission in the terms of service. That's super innocent. Some people might 00:19:04.680 --> 00:19:08.960 read something that allows that in the terms of service and think it's malicious, but 00:19:08.960 --> 00:19:13.600 it's totally innocent. Or it could mean you granted the platform a license to clip 00:19:13.600 --> 00:19:18.400 your content and use that in an advertisement for their platform. Maybe they're saying, 00:19:18.400 --> 00:19:21.960 "Here are all the greatest podcasts!" and then they have a little clip of your podcast 00:19:21.960 --> 00:19:25.280 and someone else's podcast and then someone else's podcast and then they promote their 00:19:25.280 --> 00:19:31.280 own platform. Again, that's innocent. That could even be good for you because you got 00:19:31.280 --> 00:19:37.160 featured! While most of these terms of service have been safe, despite some fear-mongering 00:19:37.160 --> 00:19:42.400 and misinterpretations here and there, some places are starting to update their terms 00:19:42.400 --> 00:19:47.200 of service, which require your consent in order to continue existing on that platform 00:19:47.200 --> 00:19:54.680 or using that platform. And in so doing, you are giving them a license to use your content 00:19:54.680 --> 00:20:01.160 to train their AI tools. Even if your content has a registered copyright, you are still 00:20:01.160 --> 00:20:07.640 granting other places licenses to use your copyrighted content. And if you, the copyright 00:20:07.640 --> 00:20:14.120 holder, share that license or grant that license to someone else, you can't just say, "It's 00:20:14.120 --> 00:20:19.440 copyrighted! You can't do that!" Well, no, you actually said you agreed and gave 00:20:19.440 --> 00:20:25.600 them the right to do that. However, it's been uncovered that many LLMs were trained 00:20:25.600 --> 00:20:31.160 on copyrighted material without any license or permission of any sort from the copyright 00:20:31.160 --> 00:20:37.800 holders. And if you use an LLM to generate new content from nothing, you might potentially 00:20:37.800 --> 00:20:43.600 be infringing on someone else's intellectual property rights. And you would be held liable 00:20:43.600 --> 00:20:49.080 for that. Just like if you hire a cheap "so-called designer" to make your podcast cover art 00:20:49.080 --> 00:20:54.600 for you and they steal images from a Google image search, you would be liable for that 00:20:54.600 --> 00:21:00.960 theft. And you would be liable for any kind of theft if you use an LLM to generate basically 00:21:00.960 --> 00:21:06.480 stolen content. Some might argue that this isn't very different from going out and 00:21:06.480 --> 00:21:11.360 reading all the content yourself and writing your own conglomeration of your newfound knowledge. 00:21:11.360 --> 00:21:17.720 But even then, you can be guilty of plagiarism by putting forth something as your idea when 00:21:17.720 --> 00:21:22.160 it was actually someone else's. And the more niche the subject is, the less information 00:21:22.160 --> 00:21:27.960 there was to train the AI, or that there is now to train whatever AI models are learning 00:21:27.960 --> 00:21:34.960 content currently, and thus the higher chance of it outright copying other information or 00:21:34.960 --> 00:21:39.920 making up something factually incorrect. Go back to number two for that. Like, when I 00:21:39.920 --> 00:21:46.920 first started using AI, none of the AI models knew anything about COVID-19. So if you asked 00:21:46.920 --> 00:21:51.960 it to write about COVID-19, it would make something up, whatever it thought COVID-19 00:21:51.960 --> 00:21:56.280 meant. You had to kind of tell it, "This is what COVID-19 is, now write an article 00:21:56.280 --> 00:22:00.320 about this thing I just told you about." Now, though, the models are more up-to-date 00:22:00.320 --> 00:22:04.960 on their information. But there's going to come a point where you might ask the AI, 00:22:04.960 --> 00:22:09.800 "Who is the president?" immediately after the new president is sworn in, and the AI 00:22:09.800 --> 00:22:14.680 won't know it because the AI is not a source of information, it just knows how to write 00:22:14.680 --> 00:22:20.520 sentences. Or you might ask it certain questions, and it could be quoting from someone else's 00:22:20.520 --> 00:22:25.000 article, it might have done just a little bit of word spinning. You really can't know 00:22:25.000 --> 00:22:31.320 sometimes, because it could be very close, but not quite close enough to be detectable. 00:22:31.320 --> 00:22:38.680 However, this is probably never a problem when you're using AI on your already created 00:22:38.680 --> 00:22:43.320 content. Warning #8. AI might already be stealing 00:22:43.320 --> 00:22:47.480 your intellectual property. And we could put quotation marks around "stealing" there, 00:22:47.480 --> 00:22:52.120 if it makes you feel better. I've had my own original content and images plagiarized 00:22:52.120 --> 00:22:58.960 or directly stolen before. But AI is making it only easier for that to happen, and harder 00:22:58.960 --> 00:23:05.080 for me to catch it. For example, I often talk about my Podcasting P.R.O.F.I.T. Paradigm. P-R-O-F-I-T: 00:23:05.080 --> 00:23:10.120 Popularity, relationships, opportunities, fun, income, or tangibles. I did an episode 00:23:10.120 --> 00:23:14.280 about that just recently. If someone used an AI to talk about podcasting profit and 00:23:14.280 --> 00:23:19.480 replaced only one of those words of mine, it's still theft. But it wouldn't be as 00:23:19.480 --> 00:23:25.000 easy to spot because I wouldn't know which word they took out or what word they put in 00:23:25.000 --> 00:23:31.160 instead. And because LLMs have been trained on a large percentage of the internet, it's 00:23:31.160 --> 00:23:36.680 very possible your own content has already been scraped and used in the training, but 00:23:36.680 --> 00:23:41.600 you might never know. That could be either your written content or your recorded content. 00:23:41.600 --> 00:23:46.480 Many places are proposing legislation that would require AI companies to disclose their 00:23:46.480 --> 00:23:51.940 sources, allow people to have their content removed or exempted, or only use properly 00:23:51.940 --> 00:23:57.480 licensed content for training the AI models. What a novel thought, right? This is why some 00:23:57.480 --> 00:24:02.400 AI companies have taken an interest in purchasing publishing companies that own the rights to 00:24:02.400 --> 00:24:06.800 large amounts of content, then they can use that content to train their model instead 00:24:06.800 --> 00:24:11.920 of having to scrape content off the internet. And I think you should have this protection 00:24:11.920 --> 00:24:17.880 over your own content even without having to do the technical process of blocking the 00:24:17.880 --> 00:24:22.840 AI user agents from scraping your website or transcribing your audio or video. Did you 00:24:22.840 --> 00:24:28.920 even know that you could do that, by the way? You could find out what are the user agents, 00:24:28.920 --> 00:24:33.080 what are the bots that are being used from some of these models, and you can put in your 00:24:33.080 --> 00:24:38.880 robots.txt file on your website to block those bots from accessing your website and hope 00:24:38.880 --> 00:24:44.760 that they respect that. Or you could put in a firewall restriction that prevents any user 00:24:44.760 --> 00:24:49.800 agent that matches that name from scraping your content, but that only works if that 00:24:49.800 --> 00:24:55.480 particular bot is using an identifiable user agent. And I just now got technical for a 00:24:55.480 --> 00:25:01.160 very specific reason. You should not have to understand any of that. You should be able 00:25:01.160 --> 00:25:07.520 to just put your content out there and know no one is going to steal it. So I support 00:25:07.520 --> 00:25:13.940 the idea of saying, companies, you need to stop scraping other people's content unless 00:25:13.940 --> 00:25:19.320 you have their permission to do so. And we as content creators need to pay very close 00:25:19.320 --> 00:25:25.240 attention to the terms of service we say that we're agreeing to. Because many of these social 00:25:25.240 --> 00:25:31.020 media sites are already doing this same kind of thing, or they make some kind of non-obvious 00:25:31.020 --> 00:25:36.200 change to their terms of service. And what do most of us do? We just click that checkbox 00:25:36.200 --> 00:25:41.000 that says, "Yes, I have read and agreed to the terms of service," without actually reading 00:25:41.000 --> 00:25:47.100 what we're agreeing to. And we're signing away our rights, and we don't even realize 00:25:47.100 --> 00:25:54.520 it sometimes. You might be granting permission to these companies to copy your work, and 00:25:54.520 --> 00:25:59.640 then if someone else used one of these LLMs to copy your work, you might not have a good 00:25:59.640 --> 00:26:06.920 case against them because you granted the company to use your work, and that company, 00:26:06.920 --> 00:26:12.240 in their terms of service, say something basically like, "You're also granting anyone who 00:26:12.240 --> 00:26:17.740 uses their product to also use your content." That can get you in trouble, or it can prevent 00:26:17.740 --> 00:26:22.160 you from getting other people, guilty people, in trouble for stealing your content. And 00:26:22.160 --> 00:26:27.240 this is happening. For example, Reddit, Zoom, and XTwitter have used or continue to use 00:26:27.240 --> 00:26:33.480 content on their platforms to train their own AI models, and we've probably given them 00:26:33.480 --> 00:26:37.940 the right to do so. And more and more companies are going to start doing this, and your only 00:26:37.940 --> 00:26:43.640 option is to say, "No, I don't agree to this," and probably delete your account, 00:26:43.640 --> 00:26:48.140 and you miss out on the benefits of using that platform, or that tool, or whatever it 00:26:48.140 --> 00:26:52.920 is. Also watch out for terms of service that allow the AI to train itself from whatever 00:26:52.920 --> 00:27:01.160 you input into the AI. That's the case for ChatGPT, but supposedly not for the API that 00:27:01.160 --> 00:27:07.460 accesses the same OpenAI GPT model, and that's what powers ChatGPT. The ChatGPT is just a 00:27:07.460 --> 00:27:14.120 particular front face for the backend model, and many third-party websites, like Magai, 00:27:14.120 --> 00:27:19.600 can access the exact same model through an API, and API is a way that programs communicate 00:27:19.600 --> 00:27:25.000 with each other. And supposedly, feeding content to the AI and getting content back through 00:27:25.000 --> 00:27:31.940 the API is protected from training that AI on anything. But, you need to read the terms 00:27:31.940 --> 00:27:39.640 of service to know, so don't give it any kinds of special trade secrets or anything 00:27:39.640 --> 00:27:44.760 like that. You might be disclosing something to the AI that should be kept secret. Watch 00:27:44.760 --> 00:27:45.760 out. 00:27:45.760 --> 00:27:51.640 Warning number nine. Claiming "fair use" might require a higher standard. I think anyone 00:27:51.640 --> 00:27:57.240 using AI might face a more difficult time trying to use the "fair use" claim as 00:27:57.240 --> 00:28:02.960 a legal defense if you're taken to court. Especially if you haven't properly disclosed 00:28:02.960 --> 00:28:07.940 your use of AI, like I talked about in warning number one. One of my favorite things to test 00:28:07.940 --> 00:28:13.440 on an image-generating AI is giving it the prompt "Harrison Ford as a pirate". That's 00:28:13.440 --> 00:28:18.520 simple and innocent fun, and I'll probably never publish that for the public. It's 00:28:18.520 --> 00:28:24.640 just something silly I do just to see what kind of result I get. And sometimes, the AI 00:28:24.640 --> 00:28:30.520 models won't generate an image of a celebrity. I think that's sometimes a good idea, because 00:28:30.520 --> 00:28:35.360 then that can prevent some malicious stuff from happening. Sometimes the AI systems know 00:28:35.360 --> 00:28:40.000 how to recognize malicious content and can prevent you from generating that, but they'll 00:28:40.000 --> 00:28:46.440 let you generate something fun, like Harrison Ford as a pirate. But imagine if I used AI 00:28:46.440 --> 00:28:52.560 to make a realistic photo of Harrison Ford using or endorsing my products. That's illegal. 00:28:52.560 --> 00:28:57.400 Using a real photo of them, or even just an artificial photo showing them using my products. 00:28:57.400 --> 00:29:02.520 That would be illegal. Or maybe using a voice or video AI to make Harrison Ford say something 00:29:02.520 --> 00:29:09.640 he didn't say. Indeed, I've seen some intentionally hilarious results with AI. And those kinds 00:29:09.640 --> 00:29:16.620 of things are often excused when they don't cause harm and are obviously parodies. Now, 00:29:16.620 --> 00:29:21.920 this is not legal advice. I'm not saying you can do this, only that this is an observation 00:29:21.920 --> 00:29:27.920 that I see this happening a lot and it's obvious most of the time that these things are parodies 00:29:27.920 --> 00:29:32.160 and I think that's why they're excused as fair use. That doesn't mean you can do it 00:29:32.160 --> 00:29:37.280 too. It's just something that I see happening and it can be really funny to see some of 00:29:37.280 --> 00:29:43.080 this stuff. There's a particular Twitter account that I follow of someone who has for months 00:29:43.080 --> 00:29:51.060 made hilarious, obviously AI generated images of a particular person and it's been hilarious 00:29:51.060 --> 00:29:59.640 to see. Just this person as a pirate, as an astronaut, as whatever, as a shoe, as a microphone, 00:29:59.640 --> 00:30:05.160 as whatever. Things like that. And it is really funny to see. And people for years have been 00:30:05.160 --> 00:30:10.080 doing that kind of thing with Photoshop or their own artistic skills with drawing or 00:30:10.080 --> 00:30:16.160 paint or anything like that. So it's not like AI has enabled something to happen that hasn't 00:30:16.160 --> 00:30:22.400 happened before. It's the accessibility of it and how quickly someone can do it and sometimes 00:30:22.400 --> 00:30:28.080 how realistically it can look. So it's no longer just cutting and pasting someone's 00:30:28.080 --> 00:30:34.360 head onto someone else's body doing something wrong, but it's now it can look exactly like 00:30:34.360 --> 00:30:39.480 that person doing something wrong or doing something that they didn't give their consent 00:30:39.480 --> 00:30:46.120 or license for you to make it look like they're doing. But AI lets things get far more complicated. 00:30:46.120 --> 00:30:53.240 And far too easily. Copying or making a derivative have some clear limitations even within fair 00:30:53.240 --> 00:30:58.960 use stuff. There are clear limitations of "no, this is not fair use". But generating 00:30:58.960 --> 00:31:04.880 something that seems real and uses someone else's likeness or intellectual property 00:31:04.880 --> 00:31:11.680 might be in a whole different category and we might see new laws about that kind of thing. 00:31:11.680 --> 00:31:18.200 Thus, while I cannot give you legal advice as to what you're allowed to do, I can urge 00:31:18.200 --> 00:31:24.680 you to NOT do anything that might get you in trouble. Maybe even just for the mental 00:31:24.680 --> 00:31:29.440 exercise of this, pretend there's not even such a thing as fair use when it comes to 00:31:29.440 --> 00:31:36.680 how you use AI to create stuff for you. And then that might help you to be above reproach. 00:31:36.680 --> 00:31:42.880 Concluding number 10. Affiliate marketing with AI might get you in trouble. AIs, LLMs, 00:31:42.880 --> 00:31:48.920 or whatever you want to call them, are very good at creating a lot of content very quickly. 00:31:48.920 --> 00:31:56.600 And that is very alluring to people who want content only for the purpose of promoting 00:31:56.600 --> 00:32:00.800 their affiliate links. I even saw that years ago when I was among the early users of what's 00:32:00.800 --> 00:32:06.100 now called Jasper. I would see people frequently talk and ask about what kinds of inputs could 00:32:06.100 --> 00:32:12.480 be used to get an AI to write a full "review" (massive quotation marks around that) of an 00:32:12.480 --> 00:32:17.160 affiliate product. But remember that thing about how large language models are good at 00:32:17.160 --> 00:32:24.080 writing sentences? But newsflash, they're not actually good at testing products, or 00:32:24.080 --> 00:32:31.000 sharing experiences, or offering opinions. They're good at writing sentences. Thus, 00:32:31.000 --> 00:32:41.600 using AI to write a "review" could lead to misleading information about the product, 00:32:41.600 --> 00:32:46.520 about its uses, about who knows what. Is anyone actually reading the stuff that's created 00:32:46.520 --> 00:32:50.640 by the AI? Doesn't seem like whoever's publishing it sometimes. For probably this 00:32:50.640 --> 00:32:57.000 reason and more, some places will probably start to forbid using AI to create content 00:32:57.000 --> 00:33:00.760 for promoting their products through affiliate links. For example, and I haven't heard 00:33:00.760 --> 00:33:06.440 anyone talking about this, the Amazon Associates Operating Agreement was updated effective 00:33:06.440 --> 00:33:11.940 March 1st, 2024, and in their "What's Changed" note, which I link to, might be 00:33:11.940 --> 00:33:15.580 replaced at some point in the future, but I link to it in the notes for this episode, 00:33:15.580 --> 00:33:21.600 it has this point as "This is what's changed." It says this, and I quote, "Revised the 00:33:21.600 --> 00:33:27.560 language in Section 5 of the Participation Requirements to clarify that program content 00:33:27.560 --> 00:33:32.080 and special links (special links are your affiliate links, really) should not be used 00:33:32.080 --> 00:33:37.840 in connection with generative AI." Now at first glance, that sounds like they're 00:33:37.840 --> 00:33:43.200 saying that they are completely forbidding the use of generative AI to promote your Amazon 00:33:43.200 --> 00:33:49.760 affiliate links. However, the actual points in the operating agreement seem to restrict 00:33:49.760 --> 00:33:58.960 using AI on the Amazon site content, and especially training the AI, maybe not necessarily on 00:33:58.960 --> 00:34:04.720 creating content that promotes the Amazon link. It's a kind of grey line, and this 00:34:04.720 --> 00:34:08.880 is again where I talk about "you need to be above reproach." So here, this is going 00:34:08.880 --> 00:34:14.440 to be really dry, but I will quote to you the two sections that I thought seemed relevant 00:34:14.440 --> 00:34:20.760 to this idea of generative AI or machine learning as it relates to the Amazon affiliate program 00:34:20.760 --> 00:34:25.560 called Amazon Associates. From Section 2.E, and I have the links to this in the notes 00:34:25.560 --> 00:34:34.080 by the way, at theaudacitytopodcast.com/aiwarnings, it says, "You will not, without our express 00:34:34.080 --> 00:34:43.900 prior written approval, access or use product advertising API or data feeds for the purpose 00:34:43.900 --> 00:34:50.440 of aggregating, analyzing, extracting, or repurposing any product advertising content 00:34:50.440 --> 00:34:55.060 or in connection with any software or other application intended for use by persons or 00:34:55.060 --> 00:35:01.040 entities that offer products on an Amazon site or in the direct training or fine-tuning 00:35:01.040 --> 00:35:07.360 of a machine learning model." Wasn't that a wonderful long run-on sentence. That's 00:35:07.360 --> 00:35:14.320 very technical and it is specifically talking about the API and data feeds, not simply a 00:35:14.320 --> 00:35:21.600 link. However, moving on to Section 5.D, "Distribution of special links through software and devices," 00:35:21.600 --> 00:35:26.720 it says, and I quote, "You will not use any program content or special link or otherwise 00:35:26.720 --> 00:35:33.240 link to an Amazon site on or in connection with any client-side software application 00:35:33.240 --> 00:35:39.180 e.g. a browser plugin, helper object, toolbar, extension, component, or any other application 00:35:39.180 --> 00:35:44.360 executable or installable by an end user." You know what, let's just skip to the interesting 00:35:44.360 --> 00:35:51.360 part. "You will not, with or express written approval," yes, that's actually what it 00:35:51.360 --> 00:35:57.720 says, they have a typo in there. I think it should be "without our express written approval, 00:35:57.720 --> 00:36:04.040 use or allow any third party to use any special links," again, special links are your affiliate 00:36:04.040 --> 00:36:12.400 links, "or program content to develop machine learning models or related technology." These 00:36:12.400 --> 00:36:18.640 are the Amazon program policies that were instituted on March 1st, 2024 and as I extracted 00:36:18.640 --> 00:36:26.360 them on April 14th, 2024. And the emphasis in the notes is my own emphasis added. Now, 00:36:26.360 --> 00:36:31.760 after that super dry section, you can look at the notes to read that again and also get 00:36:31.760 --> 00:36:38.360 the link to the actual policy. The first part is clearly forbidding using the Amazon API 00:36:38.360 --> 00:36:45.000 with an AI model to programmatically create content for you. For example, you can't set 00:36:45.000 --> 00:36:50.160 it up that whenever a new product comes out that matches this particular keyword to automatically 00:36:50.160 --> 00:36:56.080 generate a blog post for you from that Amazon link. That would be forbidden. However, it 00:36:56.080 --> 00:37:06.880 seems to still allow you to use AI to create your content about the product itself and 00:37:06.880 --> 00:37:14.480 even use your Amazon links in that content, but you should not give the links to an AI 00:37:14.480 --> 00:37:19.740 system that reads the content on the website and generates content for you from the content 00:37:19.740 --> 00:37:26.120 on Amazon's website. So, kind of technical there, but even if you're just creating your 00:37:26.120 --> 00:37:32.520 own original content with an AI and only linking to Amazon's page for the affiliate links, 00:37:32.520 --> 00:37:39.080 I still think you shouldn't use AI to do this. I, for one, would love to see a stop to all 00:37:39.080 --> 00:37:44.600 the AI-generated worthless "reviews" – again, massive quotation marks around that – on 00:37:44.600 --> 00:37:50.320 YouTube and other places. For example, I've got one in the notes for this episode. It's 00:37:50.320 --> 00:37:55.400 from the YouTube channel The Smart Kitchen. You can look them up if you want to, but it's 00:37:55.400 --> 00:38:02.880 just AI-generated garbage. It is a computer voice, low-quality recording at that, and 00:38:02.880 --> 00:38:07.680 it's stock footage or footage from the commercials of these different products. And all that 00:38:07.680 --> 00:38:12.840 they've really done, probably, is just looked up what are the top 5 search results for this 00:38:12.840 --> 00:38:18.000 or what are the most rated products on Amazon for this, and then they run it through an 00:38:18.000 --> 00:38:23.520 AI, maybe even violating the Amazon Associates Terms of Service. And then you end up with 00:38:23.520 --> 00:38:29.880 content that repeats itself that's ridiculous. I see this often. I hate it when I run into 00:38:29.880 --> 00:38:35.160 a video where it looks like it answers what I'm looking for, and it's comparing these 00:38:35.160 --> 00:38:41.200 products when it's just this AI garbage. And, for example, if it's a knife, one section 00:38:41.200 --> 00:38:46.120 will say "Are you looking to cut things?" Then look no further than the such-and-such 00:38:46.120 --> 00:38:49.840 knife. It's great at cutting things. Now, it's not that stupid with how it creates 00:38:49.840 --> 00:38:53.040 the content, but it's that kind of thing. Where it's like saying "Are you looking 00:38:53.040 --> 00:38:58.320 to cut vegetables and fruits?" Well, yes. Why would I want a knife? What do you think 00:38:58.320 --> 00:39:04.760 I'm going to do with the knife if not cut food with it? And then they have the sections 00:39:04.760 --> 00:39:09.080 for the different products, and then they present them in different ways with different 00:39:09.080 --> 00:39:14.360 lead-in introductions like that, as if one product is designed for cutting vegetables 00:39:14.360 --> 00:39:19.600 and the other one isn't. They're all knives. They're all designed for cutting vegetables 00:39:19.600 --> 00:39:25.080 and fruit and meat and whatever else. It's not like one is designed for vegetables and 00:39:25.080 --> 00:39:29.040 one isn't. They're knives. They're all the same style of knife. They're just different 00:39:29.040 --> 00:39:36.040 brands. Can you tell? This whole episode, I know I've gone on several soapboxes and 00:39:36.040 --> 00:39:41.520 had several examples of bad things, but I do have a good example coming up in just a 00:39:41.520 --> 00:39:45.880 moment because as an owner of a couple of affiliate programs myself – did you know, 00:39:45.880 --> 00:39:49.480 by the way, that Podgagement has an affiliate program? So if you're a member of Podgagement, 00:39:49.480 --> 00:39:54.520 I would love for you to join the Podgagement affiliate program if you think that you might 00:39:54.520 --> 00:39:58.520 reach an audience of other podcasters. But because I have multiple affiliate programs 00:39:58.520 --> 00:40:03.760 myself that other people can promote my products and earn from any sales that come through 00:40:03.760 --> 00:40:09.520 those links, I know that I would not want anyone promoting my products with AI-generated 00:40:09.520 --> 00:40:16.800 content. In fact, I think I'm going to update my affiliate terms to explicitly forbid that 00:40:16.800 --> 00:40:24.160 kind of garbage content. I want real people with real experiences promoting my products. 00:40:24.160 --> 00:40:29.600 For a good example of this, among all of these bad examples I've given and this big soapbox 00:40:29.600 --> 00:40:35.040 of an episode and soapbox against this AI-generated trash that I see all over the place, here's 00:40:35.040 --> 00:40:42.800 a good example. Danny Brown did a very nice thing in his recent episode where he authentically 00:40:42.800 --> 00:40:49.360 promoted my Podgagement service in his episode of One Minute Podcast Tips about two ways 00:40:49.360 --> 00:40:54.280 for podcasters to get feedback from their audiences. He promoted Podgagement and he 00:40:54.280 --> 00:41:01.440 promoted one of my competitors. That was awesome. And he shared genuine thoughts, human-inspired, 00:41:01.440 --> 00:41:06.540 human-created thoughts. That was a great example of it. And I link to it in the notes for this 00:41:06.540 --> 00:41:14.320 episode too, where he is authentically recommending it and he is a customer of both products so 00:41:14.320 --> 00:41:24.000 he can speak from his experience. An AI cannot do that. An AI cannot tell you why this feature 00:41:24.000 --> 00:41:29.320 might be better for some people than others. Yes, it can write sentences, but it can't 00:41:29.320 --> 00:41:36.400 actually communicate what people want to know or most likely can't. And even if it could, 00:41:36.400 --> 00:41:41.180 I don't think you should use it for that. I think when it comes to some of this stuff, 00:41:41.180 --> 00:41:47.720 you should be creating the content yourself. That's why I consider this a warning. Because 00:41:47.720 --> 00:41:53.320 originally when I first read that, and I read only that brief note from Amazon of the short 00:41:53.320 --> 00:41:57.760 "What's Changed?" when I read that thing that said "Program content and special links should 00:41:57.760 --> 00:42:03.920 not be used in connection with generative AI." I actually thought that meant you cannot create 00:42:03.920 --> 00:42:10.040 a blog post with AI that links to an Amazon product. That's the way I read that. I don't 00:42:10.040 --> 00:42:16.000 think that's actually the correct reading of that based on actually reading the actual 00:42:16.000 --> 00:42:20.440 terms of service, operation agreement, and all of that. But they could change that. And 00:42:20.440 --> 00:42:24.200 I don't think you should create content in that way anyway. 00:42:24.200 --> 00:42:29.040 Moving on to number 11. I spent a lot of time on that one. Can you tell? I am passionate 00:42:29.040 --> 00:42:35.000 about this and I want you to succeed. And that's why I come to this. Last but certainly 00:42:35.000 --> 00:42:42.440 not least, relying on AI can cost your authority and influence. I urge you to consider the 00:42:42.440 --> 00:42:48.720 intangible cost of relying on any kind of AI as you create your podcast episodes or 00:42:48.720 --> 00:42:54.160 any other kind of content. I've said for many years that what I love about podcasting 00:42:54.160 --> 00:43:00.800 is that it allows you to communicate with your voice. So people can hear your authentic 00:43:00.800 --> 00:43:06.620 emotions and they can hear how well you communicate your thoughts. Even if you do some editing, 00:43:06.620 --> 00:43:15.160 they can still hear you. Imagine if you used an AI to create and communicate all your, 00:43:15.160 --> 00:43:20.520 quotation marks around that, all your content. And then you're put on a stage in front 00:43:20.520 --> 00:43:26.040 of a live audience and you have done no preparation, aside from any stage fright that you'll 00:43:26.040 --> 00:43:32.280 get in that situation. Could you actually communicate your message authentically, understandably, 00:43:32.280 --> 00:43:37.380 and memorably if you don't have any notes, you haven't prepared, just speaking from 00:43:37.380 --> 00:43:42.920 what you know and your expertise and your passion. Practical example of that. 00:43:42.920 --> 00:43:47.080 Several years ago, I was invited to speak in person about podcasting to a Cincinnati 00:43:47.080 --> 00:43:53.960 business group. And for the first time ever in my life, I completely forgot about it. 00:43:53.960 --> 00:44:01.760 I remembered only because about an hour before I was supposed to speak, the organizer sent 00:44:01.760 --> 00:44:08.080 me a kind message just to say how excited she was to have me and I think to give me 00:44:08.080 --> 00:44:12.760 a heads up about parking or what door to walk through or something like that. It certainly 00:44:12.760 --> 00:44:17.760 wasn't she was reminding me that I was supposed to be there. The event was about 45 minutes 00:44:17.760 --> 00:44:23.480 away so I had only enough time to throw some stuff in my car and think about my presentation 00:44:23.480 --> 00:44:30.400 on the drive up. Now imagine if AI was my crutch and most of my content had been created, 00:44:30.400 --> 00:44:36.280 organized, or even optimized by AI. What kind of predicament would I have gotten myself 00:44:36.280 --> 00:44:43.080 into? Instead, I was able to speak for about half an hour and confidently and thoroughly 00:44:43.080 --> 00:44:49.760 answer 15 minutes of questions, all with no notes except a 5 word outline in my head and 00:44:49.760 --> 00:44:56.120 I think I nailed it. I could do that because I know my stuff, inside and out. I've created 00:44:56.120 --> 00:45:01.480 how many thousands of hours of content and pages of content about podcasting and still 00:45:01.480 --> 00:45:06.640 so much more that I want to share and teach you about podcasting. And I don't share this 00:45:06.640 --> 00:45:11.960 to brag about me or try to make you think I'm some amazing thing or anything like that. 00:45:11.960 --> 00:45:19.040 Please understand that and hear my heart in this. I share this to point out what a catastrophe 00:45:19.040 --> 00:45:26.520 that could have been if I was merely a fraud using ChatGPT. And if I got there to the event, 00:45:26.520 --> 00:45:30.720 ChatGPT and even some of these AI models didn't even exist back then. But if I got there to 00:45:30.720 --> 00:45:35.400 the event and I was glued to my phone while I was speaking because I had typed something 00:45:35.400 --> 00:45:41.880 into ChatGPT and I was trying to speak from what ChatGPT told me to say, imagine what 00:45:41.880 --> 00:45:49.720 would have happened then. So don't let AI cost your authority and influence. Certainly, 00:45:49.720 --> 00:45:56.000 artificial intelligence can be a really powerful tool to help you do many things or save lots 00:45:56.000 --> 00:46:04.520 of time. But don't trade your value for artificial intelligence. That's what this really all 00:46:04.520 --> 00:46:10.520 comes down to. These 11 warnings. Undisclosed use of generative AI can get you in trouble. 00:46:10.520 --> 00:46:17.920 2. AI often hallucinates facts and citations. 3. AI lacks humanity. 4. AI created images 00:46:17.920 --> 00:46:23.520 can be humiliating. 5. AI is biased because it was fed biased content and programmed by 00:46:23.520 --> 00:46:30.800 biased people. 6. Content from AI always needs editing. 7. AI generated content raises copyright 00:46:30.800 --> 00:46:37.480 concerns. 8. AI might already be stealing your intellectual property. 9. Claiming "fair 00:46:37.480 --> 00:46:42.520 use" might require a higher standard. 10. Affiliate marketing with AI might get you 00:46:42.520 --> 00:46:50.180 in trouble. 11. Relying on AI can cost your authority and influence. It really comes down 00:46:50.180 --> 00:46:58.680 to that. I don't want you to pay that price. The $19 or whatever dollars or yen or whatever 00:46:58.680 --> 00:47:05.680 you pay with per month to use AI and whatever time that saves you to create content for 00:47:05.680 --> 00:47:12.180 you. It's not worth it. Put yourself authentically out there. And that is so much better. Going 00:47:12.180 --> 00:47:17.720 back to that soapbox example of those super frustrating videos that I find on YouTube 00:47:17.720 --> 00:47:22.840 about products and stuff. You know what that does for me? Well, I downvote those videos 00:47:22.840 --> 00:47:29.480 obviously. I dislike them. I never look at videos again from that channel. And instead, 00:47:29.480 --> 00:47:35.800 because of that darkness that AI has created, instead that makes the people who share genuine 00:47:35.800 --> 00:47:42.300 content despite their mistakes or as Mike Dell likes to quote "warts and all" when 00:47:42.300 --> 00:47:48.800 they make content that is authentic and sharing their opinions, even if it's not thorough, 00:47:48.800 --> 00:47:53.340 even if it's not comparing every option out there, even if they didn't test every feature 00:47:53.340 --> 00:48:01.560 of the product, but because it was authentic, it was them, it was their original perspective, 00:48:01.560 --> 00:48:08.080 their experiences, their opinions, their expertise, even if they're not experts in it, because 00:48:08.080 --> 00:48:15.320 it was authentically them, those kinds of pieces of content, videos, episodes, articles, 00:48:15.320 --> 00:48:21.880 anything like that, those become shining lights to me in this darkness of all this cheap AI-generated 00:48:21.880 --> 00:48:30.280 content or garbage that's out there. The authentic people then shine. And I've purchased products 00:48:30.280 --> 00:48:36.240 because of that. Not because of what some stupid AI video recommended or what some badly 00:48:36.240 --> 00:48:42.480 written AI article that starts off telling me about why Amazon is a popular retailer 00:48:42.480 --> 00:48:47.080 and offers many details. It's like, just get to the actual content and tell me about the 00:48:47.080 --> 00:48:51.880 product I want to know about. Instead of that kind of garbage, seeing someone authentically 00:48:51.880 --> 00:49:00.080 communicate why they like that product, that bears so much more weight. So in this world 00:49:00.080 --> 00:49:06.880 where people are relying more and more on AI to shortcut stuff and create their content 00:49:06.880 --> 00:49:14.920 for them, if you continue to put in the work and create quality content, well the bar has 00:49:14.920 --> 00:49:23.080 been lowered so much by people using AI, your quality content can stand out. Then if you're 00:49:23.080 --> 00:49:28.360 stuck in front of an audience and you don't have notes, you can speak from your authentic 00:49:28.360 --> 00:49:37.200 knowledge, your passions, and become an authority and an influencer in your space, or make change, 00:49:37.200 --> 00:49:43.520 or inspire laughter, or share stories, whatever it is that you are doing, whatever that profit 00:49:43.520 --> 00:49:47.880 is that you're pursuing for yourself and for your audience through your podcast, I will 00:49:47.880 --> 00:49:55.740 tell you this, you are far more likely to achieve that by doing things authentically. 00:49:55.740 --> 00:49:59.640 Even if there are some mistakes here and there, I think you're far more likely to achieve 00:49:59.640 --> 00:50:07.640 that authentically than with AI. Use your authentic intelligence, not an artificial 00:50:07.640 --> 00:50:12.000 intelligence to create your content. I want to be clear about that. I'm talking about 00:50:12.000 --> 00:50:21.840 creating your content for you because you know your topic and you have passion and AI 00:50:21.840 --> 00:50:26.200 does not. For the sake of time, because this episode has already gone on so long, I'm going 00:50:26.200 --> 00:50:29.880 to skip the community corner for now. That will be in an upcoming episode, but certainly 00:50:29.880 --> 00:50:35.380 please if you found this content valuable, inspirational, if this has given you energy 00:50:35.380 --> 00:50:39.920 that you needed, because I'm certainly enthused right now, put a number on that. Consider 00:50:39.920 --> 00:50:44.040 giving it back to the podcast and I'd really appreciate that. If you're in a modern podcast 00:50:44.040 --> 00:50:51.080 app, you can send satoshis, maybe even a donation, or go to theaudacitytopodcast.com/giveback. 00:50:51.080 --> 00:50:55.160 Or if nothing else, share this episode with someone else. Get someone else listening to 00:50:55.160 --> 00:50:59.440 this podcast and I would be very grateful for that. The notes and the links that I mentioned 00:50:59.440 --> 00:51:08.120 for this episode are at theaudacitytopodcast.com/aiwarnings. Now that I've given you some of the guts and 00:51:08.120 --> 00:51:14.520 taught you some of the tools, it's time for you to authentically go start and grow your 00:51:14.520 --> 00:51:21.040 own podcast for passion and profit. I'm Daniel J. Lewis from theaudacitytopodcast.com. 00:51:21.040 --> 00:51:22.200 Thanks for listening. 00:51:22.200 --> 00:51:38.200 [MUSIC]