Beginner’s Guide to AI Clones: Everything You Need to Know

Let me just start by asking you this, like, what is an AI clone?

It's a great question. So there's a lot of, there's a lot of ways a clone can be defined or it could be even. It's like, what's the word? Like, is it a clone or is it a digital twin? Is it a, you know, a digital mind or a digital representation of me? I would say yes to all of them. Don't know what the term is. Maybe if you, you put in the comments or something like this or write right to us and let us know what you think is the best terminology. Because I think it's actually interesting because we're in this phase right now where it's being called a lot of things. You see AI influencers too. You know, it's like. And they're all made differently. So I mean it's like some. You can literally mimic anyone or make a fictitious character. I think this is where I want to start actually with Delphi specifically because that's not what Delphi is. But there's certainly platforms and things that will spin up and have already. There's definitely a lot of them that have not gone deep into. Well, I mean, yeah, I think there's ethics behind all of it. So it's like, okay, who are you cloning more or less or representing? With Delphi, it's very clear. It always has to be you. It's a person, it's your name. It's not some made up name. I mean, there's literally, you know, identity verifications and all that stuff that happens and a whole bunch of other checks and there's a ton of security. So when it comes to like what is an AI clone? Or like. So there's a whole lot of variations and I guess we won't go through all of those, but we could chip them off, you know, later on. But the idea is that you're taking your body of knowledge, your ip, your content, your thoughts. It could be public content could be things that you just share with your team or within your business. It could be your personal, like diaries and thoughts. It could be literally anything that is in written text. There's audio format or video format and of course you can decipher things like images and create text out of that. There's. But the point is you can actually take all that content and put it into a secure place within Delphi. And you know, security is definitely a big question. There's a whole bunch of documentation on this for Delphi and we can talk about this too. But the fact is you control your data. You can. It's not getting trained, it's not training ChatGPT or OpenAI or Anthropic and all these, you know. Whereas if you give, I think it was even Sam Altman the other day or sometime recently said, you know, I wouldn't recommend putting all of your, you know, your uni, your personal information into chat GPT, financial information, all this stuff, because we don't know where it's going to go in the future. It's like, well, you don't have that problem with Delphi.

Yeah. So if I can just step in there. One of the issues, because that's a really important issue that you bring up, is there really isn't any case law around data privacy yet in the large language models. And I get asked this question all of the time. And for example, there's a lawsuit right now against ChatGPT where if you're using ChatGPT and you're familiar with this, they effectively have what's like an incognito mode like you have in your browser. So theoretically anything in that incognito mode gets wiped out, erased. It doesn't show up in your chat history and things like that.

But because of this lawsuit, they actually have to keep all of that information.

Huh. Interesting.

There has never been a lawsuit. So for example, people put all their medical information into ChatGPT.

Right.

We talked about this earlier, Joe. They put their medical information, but there hasn't been a lawsuit yet or case law that's been created where like there's HIPAA laws and things like that where you can't just get access to another person's health records. It's not clear yet on ChatGPT and the large language models whether or not you'll be able to. So I mean, to your point about security, the thing, one of the things that I love the most is that the way Delphi is built, it's not just built on the back end of a large language model. This is real new, cutting edge AI that they're building and your content is completely walled off from the large language models. And so there's no leakage and there's no way that it's going to be able to seep through. Which, you know, if you are, if you're a content creator and you own ip is absolutely critical.

Yep.

And if I may, you know, when I think about these different types of platforms of clones, there's a couple things that kind of stick out, stick out to me. So I think that there's kind of like four big buckets of clones right now. Right. There's so many like scenarios and so many ways that people use it. But I think the way we're all talking about clones right now, like a digital representation of you, the way I break it down is first you have what's called a scripted clone. So a scripted clone is like what's being used on hey Gen. And I know a lot of people use hey Gen. It's a really cool service. I love their product. It's built really for sales and for marketers. And what you do is you write a script and you train the clone on that particular script and that's what it delivers. And I think they do a really good job doing that. Then you have what's called a non scripted clone. And a non scripted clone would be something like Delphi. So where you basically give it all of your data and all of your training, you give it everything and it trains on all of your data and it can answer whatever question that you ask it in real time. And so you've got scripted versus non scripted. And one of the biggest questions people ask me is what's the difference between like a hey Gen and a Delphi? And that's it. It's scripted versus non scripted.

Yeah.

Then what you do is you find the clones out there and this is like a whole different like bucket area. You find the clones where they're not even necessarily human, but what they are is people are making characters and things like that and they're calling them clones. So we're going to just kind of like put that as a little bit because it doesn't relate as much. And then I think the fourth is sometimes you hear people call clones. They say it's like an audio clone or a video clone, voice clone or there's or void. Right. But with, with something like Delphi, it does everything. It does the video, it does the voice, it does the text, it does like all of that stuff. Yeah. But again, if you're, if you're new to the whole cloning space, think of scripted and non scripted. And that's probably at this stage the easiest way to differentiate in your mind.

I like that. Yeah. And to go a little deeper on the, on why, you know, scripted clones, like you said. Hey Jen, there's also synthesia, there's, there's all these other ones that will modulate your content and create something like it'll modulate what you give it, that is your script or outline and then maybe provide. Yeah. Video Audio, even a presentation, like I would say Gamma app is almost like a. You're not going to see a clone of you on there, but it'll clone your thoughts and present it in a different way. It's almost like these are all different ways to scale yourself. Right?

Yeah.

Whereas Delphi, back to this whole secure place that you could put your ip because that is everything and literally on. There's a whole help doc on this where they're your knowledge, your control, you've. You maintain full ownership of all content. You, you can export this. Your security is their priority as Delphi. So it's like that is because now people can interact with that body of knowledge in real time in text, audio, video. And it's going to act differently than these other scripted clones. Because scripted clones, I mean, are like, they're instant, they're generated because they're scripted. Whereas this, there is no script. Like literally it's being made on the fly. There's this time in between a request, like something sent from a user of text, voice or audio or video, and there's, you know, it's going to a whole bunch of different places to gather that knowledge or that response. It's basically taking like 10 to 20 different data points within your content arsenal, your digital mind, and forming this very concise response that you can see. And you could turn this off if you wish, but it's really cool to see. Like, even a very simple question, seemingly simple might be referenced 10, 15 different times. And like, oh, some came from a podcast interview here. This is a blog post. This was an email that was uploaded or coaching call or a YouTube video and all that is mashed up to come up with this thing. So there's like this like inference time in between that sometimes. That's why just like with ChatGPT, it's just, they're getting faster all the time because of data centers and energy and all that models. But it's like that's why there's that little lag there because it's drumming up all this information and basically smashing it together in a way that is going to speak like you, it's going to sound like you or look like you.

Yeah.

And all this stuff's happening in like seconds or. And it's getting tighter all the time. But it's like. That is a really good point. It's like you got the scripted ones. There's a reason why they're looking.

You can't, you can't have a conversation with anything that came out of like a, hey, gen as an example.

Absolutely not.

You can have a full conversation with Delphi and you can run that conversation until you run out of breath. And it will continue. It's just going to continue to answer that, that, that person. And so it's again just for like creating one to one training one to one connections, getting to a user like where they really are.