Episode Transcript
Kyle James (00:00.952)
Hey, welcome to the AI Chronicles podcast. I'm your host, Kyle James. Today we'll be talking about how a software development company called originality.ai is using AI inside of their own business to help publishers publish with integrity through fact checking. And we'll share the exact steps that you can take in order to implement AI for yourself. Now, before I dive into that, listen closely. Are you looking to implement AI inside of your own company or maybe just struggling to get your AI to stop hallucinating? Speak to GPT trainer.
GPT trainer literally builds out and manages your AI for you, eliminating hallucinations for good. Go to gpt-trainer.com. I promise you it'll be the biggest time saving decision that you've made all year. Trying to set up AI on your own is like trying to build a house from scratch. Sure, you could do it, but the time of frustration that the AI could cause may not be worth it. Instead, it's a thousand times faster and safer to hire professionals. Scheduled consultation today.
Once again, that's gpt-trainer.com. Say hi with me, John Gillum, who is the founder of originality.ai in November of 2022. And he launched it back when ChatGBT first started. After selling his previous content marketing agency, having recognized early the significant impact generative AI would have on SEO and content creation, originality.ai provides advanced AI generated content detection.
and a modern user friendly plagiarism checker with features such as scan history, detection scores, shareable results and team collaboration. Really excited for this conversation today. Welcome to the show, John. How are you doing, my friend?
Jon (Originality.ai) (01:41.996)
Yeah, I do well. Thanks for me, Kyle.
Kyle James (01:44.052)
Awesome. So tell us what it, what is originality.ai? Give us some back like background. Like how did you find the company and like, what does it do exactly?
Jon (Originality.ai) (01:52.334)
Yeah, so what it does is it takes text from someone. So if you get provided an article or a written assignment from a student or a writer, you can enter it into originality and then originality will tell you a bunch of information about that text, whether or not it was likely generated by AI, whether or not it was plagiarized, the factual correctness of it, grammar, spellcheck, a bunch of other checks. So kind of a complete quality control tool for written word.
It started back in November 22 is when we launched. We had previously built and sold a content marketing agency, had been a really heavy user of generative AI that predated CHET GPT using GPT-3 and a tool called Jasper. And then we had sort of seen this wave of generative AI coming and thought that there was going to be a need for this tool. We built it and launched it and we actually ended up launching it on the Friday.
and Chachapi Tea launched on the Monday. So kind of a crazy, crazy timing around that.
Kyle James (02:53.666)
Wow. What was it? What was that? Like, was it, I would say it wasn't planned, but like, were you, what was going through your head when you saw that, chat to me, T just launched, like right when you launched yours, like what did that change things at all when you did that or like, well, walk me through that.
Jon (Originality.ai) (03:08.578)
Yeah, I mean, it was kind of like busy time in my like in I enjoy ski coaching and so do that a lot. And so I was kind of like in peak peak of that time where like kind of the first the majority of the first 30 days I was I was off doing doing that, is which is enjoyable. But I think in it's certainly the time and in hindsight, I was like, crap, we could have like we this we could have been in the market for six months. We just kind of like took our time. This was sort of like
not fully side project, but it wasn't my main focus at the time when it was being built. had sort of a developer and one researcher that were working away on it, but we could have pushed the pace on that project so much harder and been in the market for a good length of time before ChatGPD launch would have been great in hindsight. But yeah, it was exciting. It was exciting and I think it was pretty clear that even though the functionality wasn't that drastically different than what was available from sort
Kyle James (03:55.138)
Yeah.
Jon (Originality.ai) (04:04.876)
other tools that were wrappers around GPT-3, it was clear that just sort of the capability had just gotten high enough and the accessibility of it had just gotten high enough that it was going to be something significant. No idea that was going to be sort of this big a wave.
Kyle James (04:22.496)
Yeah. Yeah. it's definitely a, definitely a big wave. I would say there's so much like content out there. And like, I like even when I'm reading content now, it's just like, I think I've got a good idea. Like, I think this is AI generated content, but I think it's getting so good. It's like, well, is this like, maybe it is like, it's so, so to have something. So, but like, tell me a little bit. So you've been, mean, AI has been in the mix since like day one, right. with originality, but like what specific challenge were you trying to solve? Like what was it just for publishers? Was it for businesses? Like.
Walk me through like, kind of like what exactly was it you're doing for a lot of your client base now.
Jon (Originality.ai) (04:56.942)
Yeah, so a lot of our clients, they care about fairness and they care about if they've paid somebody to write content for them, they want to know that it was hueing created and not copied and pasted out of chat.cpt in five seconds. That's sort of a big chunk of our users. That's their core problem that we're solving for them is that if they are, if they're okay with AI generated content and all the risks that come with that hallucinations,
risk within Google, they're okay with that, they want to be paying less than they're paying if it's truly human created and all the benefits that come from that with sort of known editing checks in place and truly unique thoughts. And so if they're open to having AI content on their site, they don't want to pay full price for it. And so that's sort of the core problem that we're solving is that sort of fairness for people that have
writers that they're paying.
Kyle James (05:57.272)
Yeah. Yeah. So in other words, like if they're, if they are right, like if businesses are out there, you're going, okay, we're paying these copywriters, for example, to create content. Like essentially originality is going, okay, we're going to actually going to check your work to make sure that you are creating the content yourself. And if you're not like some companies, they might be okay with it, but I mean, if they're paying the high dollar premium for a, you know, really well thought out, really well crafted blog posts as an example, then it sure as heck better be not AI generated.
So, cause that's, I mean, that's a big problem. And so like walk me through a little bit, like how does the AI work step by step? Like when someone goes in and creates an account, like what exactly does the platform do and like give us some background on
Jon (Originality.ai) (06:39.192)
Yeah. So there's, there's a bunch of different products within the platform. So there's the AI detection, plagiarism checking, fact checking, readability, grammar, spell checking. Some of those are sort of pretty kind of like rule-based formula sort of approaches. Some of them sort of rely on some databases. And then there's a couple of them that are sort of like more sort of like AI native tools. So the AI detector itself, it's, it's a classifier that has been trained up on millions of records of content, AI content, human content.
some blended content to try and sort of understand where lightly AI edited content should land and shouldn't land. And so that's what the AI Detector is, is sort of this classifier that says whether it's likely AI or likely human and the confidence of that judgment. And so that's the way that the AI Detector works. The fact checker has some sort of, it's got a bit of an AI wrapper with a data pipeline and some rag built into it.
that helps ensure that identifies facts in the content and then goes out and provides judgment on whether or not that statement is factually accurate or potential incorrect so that it can be an aid for an editor to go in and make sure that each fact in that document is correct.
Kyle James (07:57.186)
So like, are for some of like, say like, either businesses using it or publishers, maybe they're just making sure like, is there a chance to like, when they're, when they create, like to say I'm a, I'm a publisher and I go out and make a blog post, five blog posts. But then I know I created the content on my own, but then I saw that, Hey, it not, it said I, I scored like maybe a 90 or like a 95 or like a 75 or even like something lower. Like, is that, is that like,
How does it do with that? is it 100 % accurate? Like walk me through what, like what does that look like?
Jon (Originality.ai) (08:27.97)
Yeah, they're all like all AI systems. They can be highly accurate, but will never be perfect. And so that's the same case with AI detectors. AI detectors are highly accurate, but not perfect. So in all our testing and in much third party testing, false positive rate is below 1 % for truly human created content. Where it can get complex is where's the line between human, cyborg created and AI created content. So if you've created it,
yourself, but then it got a polish by AI, should that be called AI or human? And that's there's a gray zone there that is, that is challenging. Um, it's pretty clean, like with 99 % accurate, 1 % false positive rate, um, where we call human AI on, on sort of pure human content, and then very, very accurate sort of 99 % plus on calling AI AI. Um, then there can be that gray zone. uh, I'd say that's, that's sort of the, the,
Kyle James (09:05.303)
Yeah.
Jon (Originality.ai) (09:27.278)
the kind of answer your question sort of that that that's the challenging zone that can happen but yeah highly accurate but not perfect.
Kyle James (09:34.36)
Yeah, yeah, definitely. And because I would imagine too, like we're just, you know, my wife, she took a copywriting course like way back when, and like when AI hit the market, she was like, I don't know how, like, if I should keep pursuing this, because like, if AI is going to generalize content. And obviously that was like, you know, a couple of years back and, um, but now like we've noticed is like, even when I'm reading content, it's like, Oh, I could tell this or, Oh, I see this, but if there's still, it's like, I still want the content that's going to be.
Like a human actually, like nobody wants to just read from a robot. mean, if you, unless you're just looking for a quick answer, then like, okay, sure. But most people are looking for that content they can't naturally find. like, I'm sure a lot of like, know, copywriters and publishers like, thank God that there's something like this out in the market because it's keeping the value that they bring to the table versus like having AI just blow over. And then on the flip side, even some of these businesses. So what
Talk to me about like some of the maybe key metrics or results you've been seeing since, you know, really since you've been, you founded the company back when really around TractGBT when it first came out, like walk me through what you've seen that's maybe worth noting.
Jon (Originality.ai) (10:40.302)
Yeah, so we've seen some of the interesting things that we've seen sort of since the launch of sort of Chat GPT is Google has penalized, has been on this sort of like tricky navigation of saying this being this AI forward company while making sure that search results are not totally overrun by AI because why would somebody go to Google if their results were nothing but AI instead of going to the AI agent or the AI chat that would have more context about themselves.
And so we've seen sort of Google on this war path against AI generated spam. And we've seen websites coming to us sort of saying, Hey, we didn't think our site just got deindexed. We weren't using AI help. And then we've run their site through our tool and we're like, actually your writers were using AI and your website has just been deindexed. And so we've seen, we've seen sort of that's been sort of an interesting, interesting sort of area of this market to watch is sort of how Google.
what Google says versus how they act and tracking that. We've done a lot of studies on looking at the amount of AI that is showing up on different platforms. so LinkedIn long form posts are hitting 50 % plus of those posts are AI generated. And then it's sort of
Yeah, it opens up to your point on like, probably don't want to read AI for everything, but there's some stuff that you define reading AI for. And so it opens up this societal question around where do we want AI to exist and where do we not want? I don't want to read a baby formula review that was generated by AI, having never experienced it. I don't want to read a travel experience review that was AI generated. I wanted to hear it from the person, but there's other situations that, yeah, an AI summary would be perfect for that.
Kyle James (12:34.614)
Yeah. So I'm going, I'm going to ask an elementary question here, but so obviously when you say DNX like that, that rang, that's a bad thing. What is that exactly? Like especially for those listening, like what do mean by DNX? Like always sounds like something like shuts down or websites not working.
Jon (Originality.ai) (12:45.998)
Yeah, shut down. It means totally removed from Google. that Google will no longer, that website that used to be getting a bunch of traffic from Google has now been de-indexed by Google, which basically means it's been removed from the Google database. The Google will no longer return results for that website. And so we'll send zero traffic.
Kyle James (13:10.68)
So, so that's crazy. that would be almost like catastrophic for some companies, but, but a lot of times, like I would, I would imagine a lot of these bigger companies, like they're, mean, if they're not using AI, it seems like they're at a disadvantage, right? So then if they're using AI for content creation, almost blog posts, as example, if what you're saying is, Hey, if there's not a checker for like, there's too much AI content, then Google could potentially shut down that website and
I mean, can't imagine what that can do for revenue overall if that happens to some companies. Have you seen a lot of that? Or is it like, it's that risk factor that's out there?
Jon (Originality.ai) (13:49.262)
Yeah, so we see there was significant manual action. So we looked at 75,000 websites and about two and a half percent of them got de-indexed by Google. And these were content publisher sites. not sort of like small businesses have less risk of this unless they, like there was this famous tweet that went a little bit viral and like a of people from some of these biggest AI companies retweeted it.
But it was about somebody that kind of did this, that looked at, did this sort of mass AI generated content spam. And they got totally de-indexed by Google where they went to the competitors, said, what are all the keywords that my competitor is ranking for? And then said, great, found all those keywords and said, AI make a piece of content for basically each of these. And then just published thousands of articles. Traffic spiked up and then came crashing down.
So yeah, it is definitely, it is a risk that I think the big thing that I like to say is that whoever the risk owner is should be the one that is making the decision on whether or not AI content should be published and at what rate because there is a risk associated with it. And so if you're the business owner, you don't want your sort of marketing intern to be spinning up an API and an agent to be publishing, you know, 50 blog posts a day. Traffic might look great for the first couple of months.
but it's a huge risk that's being placed on your business by somebody that doesn't ultimately own that risk.
Kyle James (15:20.406)
Yeah. like a little picture, like a stock market, like that you have the bull shooting up and then all of sudden big old bear and it's going straight down. We don't like anybody, anyone wants that for their business. so Matt, that the extra sharing that's that's I, I never knew that and it's pretty crazy. believe it though. So, so obviously a lot of change happening. Where do you see kind of maybe originalities or where do see the maybe kind of upcoming AI initiatives that you're maybe
Jon (Originality.ai) (15:27.074)
Yeah.
Jon (Originality.ai) (15:32.738)
Yeah.
Kyle James (15:49.376)
look for seeing in the future and like, do see it playing maybe some of biggest role in your operations next, John.
Jon (Originality.ai) (15:55.426)
Yeah, so for us, I we use it certainly heavily, like within our core tools. And so the more tooling that we, right now we're trying to provide as much tooling as possible to anybody that's acting as a copy editor. Anybody that receives a piece of text, we're trying to provide tooling. Any kind of AI enabled tooling is great. So that's like our fact checking aid, grammar, spell checking, sort of all of these, all this tooling to help.
we'd love to sort of build a more sort of agentic workflow where that function of a copy editor, sometimes people want sort of the functionality of sort of a fact checking, of a copy editor aid, but could we produce the sort of agentic workflow of everything that that copy editor needs to do? There could be sort of a full solution for them. So that's kind of what we're continuing to work on and sort of just take that
take that process and just try and sort of make it 10x more efficient right now, but then can we get to kind of 100x more efficient by totally replacing that function for some types of content. So that's what we're working on right now.
Kyle James (17:03.16)
Hmm.
Kyle James (17:14.328)
Yeah, I appreciate that. And that always helpful to have some automation set up too, with, some of the tools and where can people learn maybe a little bit you'd recommend learn a little bit more about you than maybe a little bit more about originality.ai that you'd recommend.
Jon (Originality.ai) (17:28.814)
Yeah, so you can find me on LinkedIn, John Gill on LinkedIn, and then you can email me, Johnjohn at originality.ai. And you can check out originality.ai with a bunch of studies that look at sort of the detection efficacy and how to think about it, how to kind of think about this whole world. And so we try and provide as much transparency on the accuracy, how the tool works. And if you're sort of into learning about sort of both AI
how AI works, AI content and detection and sort of the two sides of both ethical, you know, there's ethical development and ethical use of AI and sort of trying to be as transparent as possible on the ethical use side.
Kyle James (18:11.448)
Thanks so much, John. appreciate you having on the show today. And thanks for everybody else who's listening in. Again, if you're looking to implement AI into your business today, don't try and do it yourself. The time of strut that the AI could cause, it may not be worth it. Schedule a call with GPT trainer and let them build out and manage your AI for you. Once again, that's gpt-trainer.com. Signing off for now, have a great rest of your day everybody. Thanks for listening in and looking forward to seeing everyone on the next episode of AI Chronicles.
