Thank you. So we've had a lot of conversations, like really high level, important conversations about AI this week, about ethical considerations, about digital identity and, and the big picture and how AI can help us. And this talk isn't one of those. So if that's what you want, that's not what we're gonna talk about so much. We're actually gonna talk about kind of very specifically my experience in creating an AI clone, why I did it, how I did it, things I learned about doing it that you might find useful if this is something that you wanna play with yourself. So I'm not wearing any hats.
I'm not wearing my ID pro hat, nor am I wearing my research and education federation hat. Nor am I wearing my standards hat with the W three C, nor am I wearing my identi diverse hat. 'cause I do have a lot of hats. And so this is, this is again, just my, my own experience.
So what I did on a Friday night was I actually created an AI clone.
And, you know, these AI clones can actually be really pretty transformative things. But creating one is, you know, you do have considerations about what, what are you using, what material are you using to train it? And being responsible for that in practice turns out to be a lot more difficult than you might think it should be. So we're gonna talk about that. We're gonna talk about what AI cloning is in my experience, what it can do. And maybe if there's time, we'll have a little bit of fun with one.
Okay, so what is AI cloning? How many people think they know what AI cloning is? Wow. Okay.
So it's, it refers to the, the digital replica of a human's voice, their behavior, or even their entire personality using artificial intelligence. And you can, you can create your own if you've got, you know, leap coding skills, which I don't, or you could use a third party service that will let you submit training materials and then let you configure the tone and behavior of the ai, which is what I ended up doing.
And it's been really fun chatting with other people using the service because many of them are business coaches who use it to often between coaching sessions, they've given their clients homework. Well this gives their clients a way to sort of go, wait, I have questions. I I need a, I need a prompt, I need a thing, I need some help.
So it's a nice little value add to their service.
Similarly, financial managers have been using it to help their, their clients if they've got, like they're working on their, their business models at 1230 in the morning on a Saturday. Because as one does all of these things, just having that resource to ask questions of immediately has proved to be a very nice value add. And given the attention span of humans these days, which is about this big, it's, it's that having that immediate feedback is proving to be quite valuable to them in differentiating a consultant service from others.
So that's what they're doing.
But what I'm doing, why did I create one? Because I could ultimately, that's what it boiled down to. I did it because I could, because I personally have years and years of blog posts that I've written. I've got transcripts from all the podcasts that I've been on. I've got papers that I've written and then material like the actual published standards process of how do you create an IETF standard or how do you create a W three C standard? And I fed all of that information. I thought about, you know, hey, if I do that, could I like submit Id pro body of knowledge?
Well there's some legal uncertainties about you. What the creative commons, non-commercial, no derivatives when it comes to training data. So knowing that I said, you know what, I'll, I won't do that part. If you're at all worried about the copyright and license details or on the material that you might use to train your own ai, just don't do it. It's like when you're, when you're posting online, if there's really a question in your mind about I wonder if this is gonna be okay, don't, don't post it, don't use it.
So I'm gonna step back a moment and talk about, you know, some of the, the real world experiences where I've seen this kind of ai, this kind of cloning being used. 'cause it is kind of fun to watch. So Jet BI is a Salesforce support company and what they did was they created AI clones of all their employees so that their clients, when they wanted to talk with their representative, was able to talk to their representative in their representative's voice with their representative's, information about them, you know, all of these things. And I was like, oh, well that's really kind of cool.
But what you don't find out, and what I don't know the answer to is, okay, so what happens when that that consultant leaves the company? Where does their AI go? Huh?
That's, that's, that's a problem. Okay, so, or there's Air Canada.
A lot of us heard about the Air Canada debacle where using the, the chatbot, someone managed to get a really incredible discount, which Air Canada didn't wanna honor. And Air Canada's argument was that the chatbot was actually a separate legal entity.
It's a, it's a whole nother person. So Air Canada couldn't possibly be held accountable for that material, which the courts didn't quite agree with that. But I thought it was a fascinating argument that someone would make, you know. And then there's, you know, smaller AI clones like the one created and controlled by me, which is trained on my material. And someone said that it was eerily similar to talking to me on Slack. 'cause that's how I set the tones for it.
Now, if I were queen of the universe, which I'm not, but if I were, this is the kind of tool that I would suggest to really start helping with onboarding new employees. 'cause I think the enhancing your internal team dynamics could be really cool.
So your organization probably has documentation, it probably has HR handbooks, blog posts, maybe as there's more AI tools being used for meetings. You've got transcripts from the various meetings going on. You've got summaries of those meetings.
And by the way, if you're using AI to summarize your meetings, please check what they say because it comes up with some really crazy shit. But I digress, there's gonna be a lot of material to work with. And if you happen to be a company that contracts with a government, well then a lot of that government rules and regulations and law is probably quite freely available for you to use to train the ais that you have to chat with your internal employees, your new folks who may not feel entirely comfortable exposing just how ignorant they feel as they start in a day.
So think, think of that, think of that as having this sort of interesting on the spot mentor to help someone get through the onboarding process. I think it could be a really powerful tool in that respect. So that's great. Now we're gonna get into what did I personally do with one?
So the way it started, there we go. Before I even got to what data I was gonna train it with, I got to make choices. I got to say, what kind of tone did I want it to have? Did I want it to be an empathetic, they're there dear? Or did I want it to be okay?
No, you're being dumb. You know, what, what, what tone did I wanna set for my ai? Did I want it to be, you know, a really formal type thing to which I laughed hysterically because I'm not a very formal person? Or did I want it to be, you know, conversational? Did I want it to be funny or serious? And this wasn't a binary, there was a range of choices for all of these things.
And perhaps most importantly, for people in this group who, who are quite correctly concerned about these things, there is also the choice of do I want to solely rely on the data that I feed it or do I give it the wherewithal to, okay, rely heavily on my data, but if you, if my, if the answer isn't in my stuff, go look wherever else you've been trained to look for other things.
Now the way it's set up right now, as I said, no, just use my data and if you don't know, say you don't know.
Now all these choices were, you know, based on my own personal, what do I think are, are ethical and legal choices that I can make about my stuff. But as it turned out, that wasn't as easy as I had hoped it would be. And this is where using a third party service, which certainly simplifies your life a great deal, you have to start reading the terms and conditions.
So, so hosted services, by all measures, you know, they can save you a whole lot of time, but there are, there are problems. So there's some pretty standard clauses in terms like this. They will set cookies, right, that will collect data that when someone submits their email and accepts the terms of service, which is where this text comes from. And of course everybody reads the terms and conditions, right?
They, they know this, right, right, wrong, you know, I can see what email address that they submitted. Now personally I'd rather not have that. I'd rather that be an option that just don't collect that. 'cause I don't want to know.
So I, I submitted a feature to the, the vendor and said, could you turn that off please? And in the meantime, if you go to the, to the actual landing page for my bot, I basically say, if you're at all concerned, lie, 'cause I don't want your data, but I can't not collect it, I'm apparently a minority for not wanting that data. Go figure. So even for an internally facing bot, it is really critical that the terms and conditions are clear. I know people won't necessarily read it, but not only should the terms and conditions be clear, the bot should know what the terms and conditions are.
So this was an actual conversation I had with my bot just to make sure, if someone asks you, is this stuff legally binding? Say no, the answer is no. Right? That is in the terms and conditions.
But make sure your bot also knows when to set someone very, very straight about what's okay and what isn't. Let's see how much time, all right, we don't have a whole lot of time, so I won't actually do the live demo, but you are more than welcome to go play with it and lie about your email address. 'cause I don't want it.
I do see, I do have the opportunity to read the transcripts of what conversations happen. My life is too short. I don't have time for that. But the option is there for me to do so. So just be aware.
Okay, so one of the questions I ask for for, you know, since I wasn't sure if a demo would actually happen was, you know, I would like some advice on how an introvert can develop a personal LinkedIn brand. It gave me pages of material starting with absolutely, I would love to help you navigate that, which was kind of fun. And this leads to a story about will I keep using this chat bot?
I mean, is it actually been helpful to me as a one person LLC? Yes. And I'll tell you why. Because my mother called me and my mother said, I've got questions about how to use LinkedIn. And my mother's 75 years old, right?
And she's, she's still running a business. Do you ever wonder where plastic aerosol caps come from? I didn't know that was a business, but that's the business that she runs and now she's trying to build a presence online in LinkedIn.
I'm like, yay. Go mom. I don't wanna do mom, why don't you go talk to my chat bot and it can give you some ideas. So she did and she said, this is great. Can I talk to it instead of you?
So 100% a win. It may help you with family matters. They don't, they don't say that in their selling point, but they totally should.
So there's other questions that you can ask it and you can even like seed it with when, when the screen pops up and someone's like, well I just wanna play with it and I don't know what to ask, ask about, you know, well what what kind of work am I I work in standards, so ask about standards. You know, I'm really interested in non-human identities. So ask about non-human identities. 'cause all that stuff was in my blog. It's in my transcript for my my YouTube channel, right?
It's, it's in some of the podcasts that I do. So the material is there if you're interested in what's happening in the browser space and how they are making changes that will impact federated identity, ask about that. Right? Do you have a list of what identity conferences are out there? Because I happen to maintain a list.
It should, if I trained it correctly, give you that information. All sorts of just weird advice and things. So
You can't ha you can't actually do one of these talks without having that sort of closing moment. And I think having an an AI bot and a clone of me is a lot, it's fun. It's a value added resource. If I ever grow enough that I actually want to bring people on to help me with what I do now I've got something that will make training them a lot easier and give them an opportunity to ask questions and me to have weekends where I don't have to be on call for that.
And I think it's, it's going to be an a different way to think about your own personal digital identity, what it is, and getting used to it now so that you can own your identity and you can train what your identity wants to be is probably a good idea, but it's gonna be chaos for a while. I mean this is so not a settled space. So taking the opportunity to just play and get familiar with it is probably a good idea because then you personally can help that chaos settle in a direction that's gonna be sensible. And with that, we are at end of time. Thank you very much.
I hope you've had a great conference and I'll see you next year.
