(Transcribed with A.I. Some errors may appear.)

Okay, we need to just spend some time talking about AI slop because we've been making these videos about the industry and sort of the landscape we're all navigating. And if we don't talk about AI generated content, we aren't talking about reality anymore, which is ironic. So let's talk about it. And some of you guys know I'm very pro AI. I've been doing it for a long time. But there's pros and cons that we need to be aware of. So. So let's get deep into that and really pull some threads that I think some of you guys haven't thought of that matter before we get into that. The only commercial in these videos is hypnothoughts Live is the largest convention for hypnotists in the world. To come together, learn, share, support each other, find mentorship, find friendship, build relationships. It's just awesome.


00:44

We've been doing it for over a dozen years and the price goes up for this year's in just a couple days through March 31st. The price is lower. And no matter what, we still have a three month payment plan so you can get on that. We also have the pre and post conference classes that are really good. You want a one, two, three day deep dive into a topic with a presenter. That's how you do that in this world. So go check out htlive.net before the prices go up and you'll see why it's awesome and why everybody looks forward to that conference. Enough of that. Back to AI slop, which is the topic of the hour. So I've been thinking about AI slop for kind of a long time relative.


01:21

I mean, it's only been around for three years, but I've been thinking of it for us for a little bit and I've been having a hard time figuring out how to articulate it because it's kind of insulting to like a term to throw around like your stuff is slop and my stuff isn't. Like it sucks. But I heard a YouTube science guy named Hank Green talk about it over the weekend and he gave me the missing piece that I was looking for that gave me the perspective. So this video is in a little bit response to him. So if you're a, a Hank Green fan, you'll hear some stuff that rhymes with that in this. Starting with, I think a useful functional definition to use in this for what slop is it's AI generated content that exists primarily because it was easy to generate.


02:06

So it exists just because of laziness, basically. Like, for example, about 2021, I was using A. I was using ChatGPT to write essays, like, LinkedIn article kinds of stuff. One on AI is like, and how it ties in with Greek tragedy. One, how it tied in with wall E. One how it tied in with Jurassic Park. Things like that kind of, like, make it easy to understand. And I wrote those with ChatGPT. So what I would do is I came up with an outline. I would say, hey, help me flesh out this outline. Give me some deeper points, Get a couple research things that support these. And then we did that. We went back and forth, we changed some stuff, and then we wrote an essay in my Voice based on YouTube clips that had my V in it. And great, I did that.


02:50

That is not slop. A couple days later, it was April Fool's Day, and ChatGPT had an image generating tool that was good enough that I could take a picture of my car, tell it, hey, edit this picture so it looks like I was in a car or a little fender bender. And I sent it to my wife on April Fool's Day. I was like, hey, someone backed into my car in front of the house. Did you see anything when you left? And she was like, oh, my gosh, what is this? And I was like, april Fool's Day. Amazing, clever joke. I would not have opened Photoshop to put in a dent in my car. I would not have done real work to make that happen. I only did it because it was easy. That's fine. That's slop. That's just how we do it.


03:32

Here's why this matters. Oh, the first thing isn't slop because I tried hard. And the second thing is slop because I didn't. Right? So here's two things. One is I just watched the Olympics, like many of us, and would just watch that figure skating girl be all happy. And, like, she was so happy. It was really great. And she did a good job and she got a medal and she was all really happy for her. And I don't know anything about figure skating because I'm not from a figure skating place and I'm not interested in it. But I was proud of her and I was invested. And I know she's doing hard stuff because they explain it so well.


04:12

You see, they show, like, her dance is, you know, her little routine is three minutes, but they show, like, 19 minutes of, like, coaches, parents talking about it. Video footage of her training at Salt Lake City. Like, the injury that set her back three months, two Years ago. They do the whole story. And you see it so that when you watch her three minute routine and she's smiling, you're not just watching the routine, you're watching the skill, the talent, the effort, the struggle, the team, the commitment, the rallying of everyone, the community that did this. It is the culmination of all of those things and it is greater than the sum of its parts. And that's why we cry when somebody the song plays and we watch the Olympics because we appreciate all of that.


05:05

And the point of slop is that it is the antithesis of that. I think that's a good way of thinking of it. AI slop is the opposite of the Olympics. We love the Olympics because it's hard and we create slop because it's easy. And if when that girl was done with her routine, like whatever her routine was, if they had shown me an AI generated version of it that looked exactly like her, like the Will Smith eating spaghetti, but it's exactly like her and it's twice as good as her. So when she did like a double thing, it does a quadruple thing and when she does a triple thing, it does a six of that thing and she's doing super lutzes. I wouldn't care. I wouldn't be impressed. Like the novelty of, oh, wow, that's good, that's convincing.


05:52

But that wears off and that's not neat. So an interesting thing about generative AI, generative, it makes it easier to generate. However, Supply and demand says if there's a lot of generated stuff, it goes down in value. And if we don't have effort going in, it goes down in value. So the irony of generative AI is it reduces the value of the generated things twice. So there's this slop that's out there and that's fine, but it's going to change shapes. And you're going to see why this matters in our conversation, because it's going to keep changing shapes and then we're going to come back to the good of it. So it started off with words, right? Chat GPT, it was words and, you know, whatever. Three and a half years ago, you could generate emails and tweets without having to think about it.


06:51

Thank God, we can finally generate more thoughtless tweets and emails. Finally, Santa got my letter. And so now we have even more bullshit to consume than ever before. Because functionally infinite wasn't enough. More than I could get through in my lifetime wasn't enough. Now I've got to have the shit that you didn't even care enough to finish writing yourself. And I'm supposed to consume that, right? And now people are just using AI to consume each other's AI generated shit. So you write like, a summary. The AI writes a long one. The other person's AI reads the long one and then reads the summary. Like, what are we doing here? Right? We're just digging holes and filling them anyway, but let's kind of go. Sorry, I distracted myself with that metaphor. Let's stay on this.


07:45

The shape of slop is going to change and has been changing. So it was just words and then it was pictures. Like, I did my little slop thing with the car accident thing. That's just, you know, AI generated and then it became video and it was Will Smith eating spaghetti, right? And it started off and it sucked, and then it got a little bit better and a little bit better, and then it got really good. Then it was. The sound was in it and, like, he could talk and with his mouth full. And you're like, wow, that looks just like Will Smith eating spaghetti. Thank God we have that. And it's a neat technical demo. I get it. But you wouldn't call Will Smith's agent and say, hey, I really want to film him shooting some spaghetti.


08:19

And then you wouldn't go into, like, after effects and, like, color correct it and like, make sure it looks real good and then edit it together so it's like, dynamic. You wouldn't put in the work and you wouldn't go into, like, a 3D rendering program to, like, make a video of Will Smith eating spaghetti. It's not worth it. But if you can just type in a sentence, it is. Okay, so it went words and then pictures and then videos with sound. Then you can make the thing change and you can use, like, 11 labs or any of these technologies, or you can do all that stuff. But here's what's neat, guys. The next layer of slop is about to be something different that's going to change stuff because AI slop is about to be apps and businesses. Slop. Apps, app slop is coming.


09:02

And here's why this really makes things weird, because I have mostly just thought of it as slop. If it was artistic, creative, and, like, when I need to, like, transcribe something, I don't think of it as being lazy slop because it's just utility. It's just functional. Like, hey, transcribe this for me. Hey, you know, get the background crap out of that thing so that I can use it for something else. Like just these easy, like, little tasks. Okay, Task this, schedule an appointment, you know, put this on my calendar, make a list of this thing. I'm not critical of that as slop when it's lazy utility when it's utility.


09:42

But we're about to reach this interesting spot because of this thing called Koby and Koby code and Codex and all these fancy new things that have all just come out in the last little while that are great. Because now the generative in generative AI isn't just words, pictures and videos. Now it's apps and programs and businesses. So you're about to see a bunch of slop that sounds something like this. I can put together a custom business for you that runs your entire hypnosis office. I can create a daily planner that gets you more productive. Want to find the best script for your client? We can use this thing, too. And some of those are going to be great, and some of those are going to be absolutely trash. And this is why it's tricky.


10:27

Because the people who are making those are not magic special geniuses. They're people who are about a couple weeks or a couple months ahead of you on the same generative AI path. And because they're a little bit ahead of you're still at generating pictures, but now they're generating this other thing. So it looks really impressive, but they're just a couple months ahead of you. It's not impressive. Now, here's what's neat. I'm a couple weeks and months ahead of that person, which is equally unimpressive. It's just not. I just started a little bit before that dude and found a better YouTube video that educated me than he did. And I read the document a week before him. Like, not special.


11:07

But it's useful in this conversation because I'm the person who can tell you that there are slop apps coming, there are slop business offerings coming. And we need to measure these things based on how we just measured the other stuff based on effort that goes in. And there is nothing wrong with making a funny picture. But I also want to talk about, here's an interesting thing, why they look similar. Have you noticed this? Because it's part of it and it's part of why we don't appreciate them. Because we're good at recognizing patterns, right? Humans are so good at that. And we started seeing, oh, there's only like five flavors of generic AI pictures, right? There's the one that looks like the studio Ghibli one.


11:49

And then there's the one that looks sort of like a caricature of you where it doesn't really look like you anymore, but it looks sort of like this waxy caricature. And what's that? Uncanny valley. It's in the uncanny valley. It's not realistic of you yet. Then there are some that are on the other side of the uncanny valley. They look photorealistic and they look good, but they had this like, tealish color and gold kind of hue all over them. There's a few other simple examples. And the reason that exists is because of a thing called reinforcement learning with human feedback. RLHF and this is how they train AI systems. Before you get them, they do a whole bunch of hard math. Skip that. It's hard. But then they do this part that's easy, where they just start showing two versions to a person.


12:36

Say, which one do you prefer? This one's a thumbs up. This one's a thumbs down. Thumbs up, thumbs down, thumbs up, thumbs down. Up, down, up, down, up, down. Yes, yes, no, yes, no. With two at a time. Two at a time. Two responses, two responses. Two outputs. Two outputs. Which is better? Which is better? And they ask a bunch of people, and all of a sudden you have, oh, these seven just kept being the most popular on average. And so they're the middle of the bell curve. And so that averaging is what AI thinks good is. And so like generically appreciated averaged is AI slop. It's not thoughtful, it's not art. It's not a position. It's not something interesting. It's just slop that's coming for all the things that I just mentioned.


13:24

And also the words mean we can already do it with scripts, obviously. And by the way, script gnosis that I started working on with some people in 2019 was that. And it wasn't amazing technology when we did it in 2019. It was mad libs back then and it's mad libs now. And the idea was script labeling processing for reinforcement learning tools. But we didn't explain that to you guys yet because it was 2019 and none of you had heard of it. But that's what were doing there. And your guys are going to see the slop versions of that coming soon. Cool. But also you're going to see real content and real work happening. So I don't want to throw everything out and I don't want to throw the baby out with the bath water. And also.


14:05

So let's get to the good side when Windows came out, it was a graphic user interface. And we had not used a mouse and keyboard, we had just used the keyboard in dos, right? And so now we had a Windows, we had the mouse. So we had to get Minesweeper and Solitaire because we had to learn. Minesweeper taught us point and click, this square, that square, this square, that square, Drag and drop for Solitaire. Drag this one over here, this one over there. And that taught us the basic functionality of how to use this new operating system. And I would put to you that is what the slop is. It's Minesweeper. So the pictures you're posting of your family that don't look like your family, it's Minesweeper. The script ideas that you've mass generated bullet points of, it's probably Solitaire.


14:58

And that's not a bad thing. Keep using those, keep learning Minesweeper and Solitaire. Get confident with the tools. Learn this stuff, get better. It's gonna be important. And also we're all gonna get better at judging what is slop and what isn't, because the tools are gonna keep getting better, like I said. So it's gonna be newer things, but it's also gonna get easier and easier for everyone to make slop. So right now, if I wanted to make a 30 second video that was like, looked very real, I could make that with AI. And some of you can go, oh yeah, I know how to do that. And some of you are like, I can't. But it doesn't make me like a miracle worker. And in six months, those of you who said I can't, you'll know how by then, okay?


15:48

And so now all of a sudden, the amount of people who can generate slop is increasing and the amount of systems generating slop is increasing. And so we had an infinite amount of content, and now we have an infinite amount of content with infinite generation on top of it. And so I believe humans are going to get increasingly good at judging content's intent and effort. And I think real is going to get incredibly valuable over the next few years. And I think hypnothoughts live is going to become more valuable for all of us because of that. And so slop isn't the enemy. I actually think slop's going to help us appreciate real more. I think slop's going to help us get better at using all of these new tools and operating systems more.


16:41

And I think it's going to be a fun way to do it. Just like Solitaire, a word of warning, though, is if you're using slop to promote your business, it's going to age very badly for you. And in an industry where reputation matters so much and corner cutting looks so bad, using AI generated content lazily is going to paint you with a brush you don't want to be painted with, and it's just something to consider. So let's get good at using these tools. Of course, let's be aware that you are going to be offered things that are definitely slop, and this is all going to be really fun. So that's what I got. Htlive.net for the conference. This was a long video, but I hope you're liking these, and I hope they're helping you understand the world we're all trying to help people in.


17:42

So that's what I got. Thanks. See you next time.

​