Is GPT-3 Overhyped (Or Will It Help SEOs?)

Eric Siu explores whether GPT-3 lives up to the hype and what it means for SEOs. He demos the AI generating cover letters, tweets, ads, and full articles, then warns it could become "article spinning 2.0" by flooding the web with cheap, high-quality content at scale.
Chapters
All right. So today we're going to talk about if GPT-3 is overhyped. So first let's talk about what GPT-3 actually is so you can get a little acquainted with it. So let's go ahead to my healthy screen over here and a healthy screen. Let's go look at my screen over here and look at it. Just look at Wikipedia right here. Here's what it says. Generative pre-trained transformer. So GPT-3 is an autoregressive language model that uses deep learning to produce human-like text, all right? So it is the 3rd generation language prediction model in GPT, the GPT-N series created by OpenAI, which is a San Francisco-based artificial intelligence research laboratory. So, you know, what's interesting is what I like to do, OpenAI Crunchbase, this is one thing I like to do, I like to look at how much the company actually raised in funding, and I bet it's a lot. So they raised $1 billion. And if you look at the current team, I believe it's Sam Altman that is currently— yep, here he is. Sam Altman that's currently running it. And he used to— just scrolling down over here, Sam Altman. He used to run Y Combinator and he's a successful startup founder and great investor as well. And he stopped what he was doing at Y Combinator. And by the way, Y Combinator is an accelerator that actually raised or helped a lot of companies like Airbnb. They funded Airbnb, Dropbox, Twitch, all these different companies. So they've got a really extensive pedigree and network. And Sam Altman, so talented that Paul Graham, the founder of Y Combinator, you know, told him to run it. And then, you know, he sees OpenAI as, you know, his responsibility and as one of the big next steps for humanity. So that's why he's helping with it, right? So one of the big things that's going on Right now, especially a lot of people in the— not just the AI community, but you have a lot of marketers talking about this as well because this is very much something that, you know, marketers have attempted in the past, right? Especially content marketers and SEOs. They, you know, there's this article spinning software that would manipulate the Google algorithm. And basically what would happen is, you know, if you can create content at high volumes, it didn't even have to be high quality content. But it would just naturally rank very highly, right? Because the algorithms, again, they weren't that sophisticated at the time. And so, you know, what would happen is you just see a lot of junk, more and more junk appearing on Google. And this is how companies like eHow, which I don't think they're around anymore, but this is how these companies went public, because they just really were arbitraging the gap between just really how capable the Google algorithms were, right? And then, you know, algorithms got better and now Google is actually investing a lot into AI. And but, you know, now, now you have companies like OpenAI talking about GPT-3. And what GPT-3 does, by the way, is it allows you to, to create content at scale. And what I mean by that is now, you know, when GPT-3 actually creates an article, I actually can't tell the difference between a human being versus a computer, right? Back in the day, you can actually tell when an article was spun. So, you know, my kind of thing around GPT-3 is like, is this just article spinning 2.0? And that's my concern right now. And I'll kind of walk through some of these community examples that OpenAI is showing. So I have access to GPT-3 right now, and more and more people are getting access to it. And you can— after that, you have to basically pay a fee based on usage so you can use their API. And their API will allow you to basically, you know, create your own little apps, right? So here's some cool things you can do. You can actually create a cover letter. So cover letter generator, and then you would input the company, the job description, the resume right here. So boom, all of a sudden here's what happens. This cover letter will be created. And you can see right here, I'm a creative technologist and founder of Sharing Friction, a creative agency. That helps clients tell stories with technology. I'm also a writer and a musician, okay? So let's take a look at this right now. Oh, look at this. It pulled directly from here, okay? And then I'm also a big fan of Spotify. I've been a subscriber for years and I've been listening to Spotify Unwrapped since it launched, okay? Okay, look at this. It's literally pulling from here and It's hard to tell. I can't tell at all, right? And so that's kind of scary. It's like, oh my god, now we're gonna be able to really create content at scale. Not just content at scale, but quality content at scale. So SEOs are probably like licking their chops right now. Oh my god, I'm gonna be able to create a ton of content. I'm gonna be ranked— be able to rank really well. And so, you know, let's go ahead and take a look at down here. You can actually generate tweets as well. So what you can do here is You can generate tweets in the style of a certain person, okay? And then you can pick kind of the top 25% over here. So, example high-quality tweets. These were generated, these were created by generating 1,000 tweets, scrolling down to tweet number 100 or so, and then selecting the best. These probably reflect about a 1 in 3 to 1 in 5 of the generated tweet candidates, alright? So, programming is hard, people are harder. So I think that's the very first one, this is the original set over here. And you can see this is the guy that tweeted it. Programming is hard and people are harder. So literally you can pick the people you love the most. Like for example, Naval Ravikant tweets a lot of wisdom. And so they basically will create all these permutations over here. Software development is not a job, it's an identity. The first step to solving a problem is admitting you have one. I have problems. So I mean, it kind of seems useful, right? And there are actually a lot of companies right now doing headline generation because, you know, marketers need to create copy, they need to create headlines. So, okay, let's take a look at this one over here. Here's the topic: Britain coronavirus beaches. Videos show crowded beaches in Britain. Apple Big Sur software. Apple promises faster software update installation with macOS Big Sur. I'm not even sure I'm pronouncing that correctly. And so this will give you some ideas to work off of. And again, you can actually produce full-on papers, right? So if I wanted to write— imagine, I remember, you know, my— one of the classes I had in college was history. And I would just go to Wikipedia the day before and I would basically just reproduce the article and I'd probably get at least a B+ on it. I'm pretty sure I'd get an A using this, right? Now, here's where I see the problem. If I have this API access right now, and I'm assuming they're gonna give a lot more API access to people, everyone's— the bar is gonna be raised for sure, but everyone's gonna move up to that, right? So, the quality, everyone's gonna be creating quality content at scale. So, now it's gonna be even harder to create, you know, really good written content, right? So, who are the people that are gonna win? It's going to be the people that have a strong website with a strong domain authority, and they can just crank out content like this. I don't know how well Google is going to be able to tell, you know, what was created for OpenAI or not, because it's very difficult to tell right now. And so, you know, and you also have to assume they're pulling from, you know, tons and tons of data where one article that they produce for you is not necessarily going to be the same thing as somewhere else, right? So you're gonna have all these different permutations across in that. I think it's gonna make it a lot harder for Google's, the search engine algorithms. But again, my struggle with this is that if everyone has access to this, everyone's gonna be producing the same quality of content. Which is why right now for our software company, ClickFlow, you know, our developers are bringing it up where it's like, oh, we can help people create content. And I'm just like, everyone else is doing that right now. Like, what can we do that is actually unique, right? Because some people have showed me some stuff, around, oh, Eric, we have this copy generator. Oh, Eric, we can create ad copy for you at scale or headlines for you at scale. And now that I'm actually playing around with this right now and I'm looking at these community examples, which you can Google, by the way, I'm just like, yeah, I mean, it looks novel and we'll probably play around with it. We'll see what we can do here. But, you know, I'm cautiously optimistic right now. That's kind of where I'm at. So, I just hope it doesn't become Article Spinning 2.0, because, you know, I have bad memories from that, we'll just put it that way. So, that is it for today. Before we go, go to levelingup.com and get a free chapter of the book, all right? And don't forget to subscribe and hit the like button if you're coming from the videos, or just, you know, just subscribe in general if you're coming from the podcast. So, that's it for today, and we will see you tomorrow. Bye.










