"What If You Had an AI Version of Yourself Working 24/7? Executive Roundtable

Um, yeah, so the whole thing is introduce yourself, what you do quickly, and then your question. Or, hey, I, I really found this tool to be really cool. I did this and it worked really well. Or, hey, this was a horrible idea and I want you guys to learn from me something you shouldn't maybe not do. So Michael, why don't you go ahead and kick us off with, um, the slides that you had to kind of lay the groundwork?

Okay, so I've got a handful of slides for you all. I'm not going to read it word for word, and I'm just going to briefly go over, you know, the generalized concepts. So if you can, while I'm talking, pull out a piece of pen, pull out a piece of paper on the side of the screen, and take notes as to about things you want us to reflect back on. And Claire, I see you, you're here. I absolutely want you chiming in on this as well and calling BS on it. Because Clara's been doing digital twin work now for about 5 years up in Sweden and working with executives from around the world. So, she's got a lot of great expertise to bring to the table as well. And I'll look around and see anybody else that we can call out on. So, again, this is intended to be a group community discussion, not me just sitting here pontificating. But let me go ahead and walk you through some of the ideas that I've been thinking about in the context of the twin. So when you think about AI, it's a mess and it's really confusing. And we've got terminology all over the place and we're not going to dig into this. But we talk about LLMs, SLMs, we talk about MCP servers, public cloud, private cloud, OpenAI, Gemini, local, you know, Ollama, all of these different things. And we can talk about those offline. When we specifically drill into this concept of twins, it's really confusing as well. There are so many terms. that we use around the idea of twin. Is it a personal AI? Is it a digital twin? Is it an agentic agent? Is it a copilot? And, again, I've got a whole mapping that can help us understand about a twin, the experiences that are wrapped around that, the typical use cases for it, and then the limitations for each one of these different terms. Very, very interesting dialogue. We could spend the entire hour just on this one slide alone on the terminology. Taking a big step back, the way I define and look at a twin, and I want us all to kind of react to this in our conversation, is that a twin is— a digital twin specifically is a persistent computational representation of an entity. And when I say entity, I'm saying the connected individual, you, me, all of us with our connected devices, an organization, a product, a service, or another intelligent system like another twin, another agent that is associated with an identity and can perceive, reason, and when authorized, take action in alignment on behalf. Twins act within 3 contexts. They act as a public twin, a multiplayer mode where they're out there in the world interacting with everyone. They act as an intra-public twin, i.e., behind the firewall of a company. For example, maybe there's a twin of a CEO and all of the employees can interact with that CEO twin. Maybe access his email, her email, maybe access their private musings and thoughts for the purpose of being able to scale across the company. Then you have this idea of private twin, which we really call kind of single-player mode. This is a twin for you personally. It's got a whole bunch of agents that will run off and do work on your behalf. And across all of these, one of the things you want to be thinking about when you're setting up your twin is kind of like the Chernobyl-level blast radius you want to be prepared for. Like, if you just open up the twin into your environment, it could go nuts and you could have a Chernobyl-level event that blows up your entire world. Or if you properly set it up, you know, if the twin goes crazy or if the twin starts doing things that you don't expect or want it to do, you have a region that's protected. And so those are some of the considerations you want to be thinking. Why do we use twins? Again, on the public side, we use them for scaled presence, delivering consistent experiences to create personalization and to extend your brand. In the inner public world, we use it to help operate and execute our businesses, work across teams, scale expertise, reduce friction and communication, enable consistent-based execution. And on the private thing, it's there to help us think better, move faster, maintain context, improve our memory, and improve our decisions. Another thing, and it's kind of on a multidimensional configuration space that you want to be thinking about when you're using a twin. There is the relational exposure, private to public. There is the sovereignty, namely, is this twin verified or is it cryptographically certified with the provenance of all of the insights and data that has tied to it, tied to what's called verified credentials? Whole other discussion we can have. And then what's your risk zone? Is this twin working locally on a private VRM that you are completely under control of, or is it out there in the cloud on some third-party public service that you really have no idea what's going on? So that configuration space is something else you need to consider. Now, I want to say all of this in the context and backdrop of where we've been and where we're going. Traditional B2B, B2B2C, M2B, A2B, A2P marketplaces are what I call the transactional engagement technology architecture. How do we engage people, use technology to engage those people, use infrastructure to engage those people, etc.? I think we're under— with the advent of adding this AI mediation layer over the world, we're undergoing a significant shift. We're adding this relational layer between people, organizations, and these intelligent systems that are going to be fundamentally shifting and changing the nature of how market and markets in the world will work. And I'm calling this the relational technology architecture. That relational technology architecture has the connected individual, and all of us work with relational experiences, relational technology, relational infrastructure, relational capabilities, and relational conditions. You can read these words later if you want, but the experiences are about how we create that trusted environment. The technology is how it's initiated. The infrastructure is where it exists. The capabilities are the actual features of what we can do. And the conditions are that ambient background of trust and regulations and social norms and incentives that influence the nature to which and how this architecture evolves. Again, we don't have time to go through all and necessarily having me pontificating on all of this, but across every one of these stages of the layers of the architecture, you have all a whole bunch of different elements to consider, such as culture, such as governance, such as trust, such as communication layers, such as security, Again, several discussions for us to consider. And then you can, you know, if you want to take a screenshot of that, go for this. But you can see the kind of the distinction between the shift between transactional architecture to relational architecture. Which then brings me to a final comment that I want to think about as we think about how we go through this model. What's happening now too, as we shift from this and as we all individually start taking, you know, sovereign control and agency over our twins, We need to think about the impact that these twins will have on our ability to critical think. We need to think about how we bring our own models to the table, namely more advanced twins allow us to pick and choose which LLM, which SLM, which large language model, which small language model do we use. Do we have twins that allow us to use multiple models simultaneously? Which kind of mediation framework we work with. So, for example, I'm on the board of a company called Athova. which enables you to not only bring your own model but bring your own framework like the McKinsey model or a 5-based personality model when you're doing your work. You also need to be thinking about how you bring your own device, bring your own infrastructure, and then ultimately, and this is really where my core pillars plan to play, how you bring your own knowledge. So I've developed a whole framework that I call the 5 C's of knowledge management, how we actually take control over our knowledge. And use our own personal knowledge and our own organizational knowledge to train the twin. And that's a huge, huge model is how do you not only— you don't train models, models are out already out there trained. How do you personalize the twin with your own knowledge so it's acting on your behalf? And then how do you actually bring your own identity and/or the delegated authority you give your twin when it's out there on the market? So with that backdrop, those are some of the things that I wanted to discuss. And give you that overarching structure for what I think digital twins are all about. And I'd love to hear what people's comments are and what your questions are.

Yeah, definitely, uh, raise your hand if you can. I'll, I'll kind of start off. I started using, thanks to Michael, uh, Delphi.ai, where I built kind of a digital twin myself, loaded up all my contacts.

I'm—

yeah, my contact stacks, um, website videos, all these podcasts I do, roundtables I do. And it could, it could basically talk to anybody 24/7, and it uses my voice. I'm not using my video, mimicking me, but it seems like a pretty good idea for me to kind of have that capability of people asking questions if I'm not around. So that's one of the things that I felt was really cool after talking to Mike. I've known Michael since 2006, 2007. So It's been a really good solution for me to kind of duplicate myself and almost make an AI version of Michael Falato. Vincent, go ahead. Go ahead, Vincent.

Yeah. Yeah, Michael. Quick question for you. So that was a lot, but it's really good to kind of understand all the different components of a digital twin. I see the benefits and the benefits associated with The discrete knowledge and approach and disposition that I have engaging with other folks. But oftentimes, isn't it that those individuals are engaging to get the right answer? Or in other words, why is— why would I emulate myself if I could— if AI purportedly has potentially better answers? They always say AI is like having a Michelangelo and an Einstein. sitting next to us, right? So, why is it that I would make the effort to create a digital twin of myself when AI is much smarter than I am?

Well, remember what— yeah, that's a good question. Remember, what you're doing is your digital twin is AI, but what you're doing is you're training that AI model with your own data. And so, again, I'm not the AI infrastructure expert, but at a very loose, generalized way to think about it, you have things called large language models and small language models. Those are basically predictive algorithms that— and probability algorithms that have been trained on billions and billions of pieces of data, right? So that they can predict the outcome of the next word, of the next pixel in the image.

Right.

What the twin part of it is, this is what I was trying to refer to at the end. You bring your own personal knowledge to the twin. You train it. You upload articles, you upload images, you upload your own frameworks, you upload your own ideas, you upload your own considerations, you upload your own behavior, your own tone. And over the time— and that's a very loosely termed, it's a common term you refer to it— is called a RAG. You train the twin with your RAG, and then what happens is the models will, rather than going out to the internet to find its predictions, it will first start with your RAG. And get a sense of who you are as it— to lean into its responses.

You said RAG.

You said RAG. Yeah. Retrieval something database. What does RAG stand for again?

Thanks, Andrew. Retrieval augmented generation.

Yeah, right. And so basically, and then again, and the simplest things, think of a RAG as your own little personal database that you're uploading files to. Now, one last comment that I'll make for this that we all need to really understand too is what we all— and when I'm— and I'm helping companies build out personal knowledge systems alongside the building out twins. What we need to learn is when we write and when we produce content. And Kim Seiler, who's the founder of Athova, who is the— I just got invited to her board, very excited. She's on it. And Kim, if you can comment. There's one writing for the human audience, but now we're also having to learn to write for the AI. We're having to learn to write for the machine. And that's a different way of presenting information, structuring information, and providing attribute-level data and wrapping your data around. So it's It's going to be a lot of really interesting things on how we create going forward. Does that help, Vince?

Yes, it does. Thank you very much.

Now, Michael, do you want to help mediate this? Who's next?

Yeah, Ian.

Hi, thanks. Ian Corby. I run a trade association called the Age Verification Providers Association. I was at a conference last week in D.C. run by FAR.AI. And there are people there who are literally building explosives to put on computer chips so that they can self-destruct the AI before it gets out of control. I'm going to take a lot of convincing that I want to put my personal reputation, on which I completely depend for my livelihood, in the hands of this technology, which is still just proving so unpredictable and uncontrollable. and quite often completely wrong. I just— so I'm looking forward to the conversation, but I'm going to sit in the corner of cynic, I'm afraid.

No, no, and I think you're absolutely viable and it's valuable that you're there. And, you know, and I'm not saying I'm not there as well, but it's kind of one of those things. It's happening. Let's learn it and not just stick our head in the sand. Another comment I want to make before we move forward is Lou mentioned in the comments that I'm nervous about data privacy. And Lou, this is where the digital sovereignty spectrum comes into play on the configuration. There's a whole new layer to the internet that's being laid down called self-sovereign identity that's going to give us cryptographic control over our data and identity. And when you're out there picking a twin provider, one of the things you need to think about is what is the— what is the role of data in their DNA, like Profina, which is another twin provider. So if If Athova, like Kim Siler's company, is the enterprise digital twin and best-in-class player that's out there, Perfina is the best-in-class for me as an individual, primarily because they have established the company on the idea of self-sovereign identity, that your data is yours, you own it, and the second you want to leave the company, you lobotomize the infrastructure and you leave with your data. So you really, when you're picking a twin, You really want to be figuring out the nature to which that company approaches the sovereignty of data.

Yeah, I'm wondering what if that Delphi one that I use is— I didn't even check that.

Yeah, you got to check it. And don't just, don't just read the terms of service. Look at the actual integrity of the executives. And I'm not saying they are or not. I don't know. I'm not, I'm not using it.

Especially if you're doing the free model, right? They're probably getting something out of it. Yeah. Kim.

Yeah, I just wanted— so thank you, Michael, for the, for the, uh, many nods. That's great. Um, founder of Athova, also a founder of Dawn's Edge AI, where I train senior executives just to get their heads around it in 180 minutes. Um, so I do want to ask a question, is where are you drawing the line between digital twin and second brain? That's one. Uh, and just getting a— getting an understanding of the difference between the two. And then the other one is I'm not so sure I want a digital twin of me because I have many faults. So I like to basically build a team of experts that represent me, right? And have my voice, have my point of view, and keep them the experts. But they're not necessarily me. Does that make sense?

Well, but, and let's keep in mind, a digital twin will never be you, right? It's, you know, it's a predictive AI engine. You want to put up the appropriate disclaimers. Like, this twin has been trained to respond. And, you know, and again, it depends on the context of how you want to put it out there, public versus private versus some hybrid model. But it's got to be really clear that it's not you, never will be you. But yet it's a way to help you scale and a way to kind of probably, you know, probabilistically help people understand directionally where your thinking might be. But again, here's the difference. If they want to talk to your twin, it's $100 an hour. If they want to talk to you, it's $1,000 an hour. That's the difference of the value between the human and the twin. That's one thing I might say on that. As for the second brain concept, this is where I go down to the idea of a personal knowledge management system and that repository of my ideas and how I'm curating my thoughts and my sentiments and all of that versus the reasoning engine I'm putting on top of my second brain. And, one of the things that I really, for example, love about your system or the Perfina system are much like our own brains. We have our memory palace, which is like the second brain, but then we have the ability to put multiple, multiple language models, small and large, in parallel. So some models are great for text-to-speech, some models are great for text-to-images, some models are great for video-to-text. You know, we'll, we'll use multiple models. on top of that memory palace, on top of that second brain to help us with whatever context we're in, private decisioning or public representation. And in the case of like your company, helping us do digital transformation within the organization at scale in a huge way. And then to be able to swap out the frameworks and models depending on the context we're working in.

Right. But I think where the The twin comes in, and that is that we're actually developing, like I had mentioned, experts in each of those fields, or—

Correct.

Oh, yes.

Yeah.

Those verticals that we think, you know, so it's not me, you know, a second of me doing it. It's, it's an actual expert, right?

Correct. Well, you know, yeah. And let me, let me jump in on that. And the CEO, Marcus, over at Perfina has done something really, really, really cool. And this is the other point, too, that I want to make. I really want to keep hammering back on that self-sovereign identity piece. I think the biggest piece we're missing right now when it comes to twins is actually having a clean and crystal way of establishing verified credentials, verified identity onto the twin, and verified legal entity identity authority to the twin, delegated authority to the twin, so that when that twin is out there, I can cryptographically know that it is a personification of Kim, controlled by Kim, and has been delegated via things like VLEIs, verified legal identifiers. what that twin's allowed to do, not do, say, or not say.

That's definitely the key distinction then.

Right.

That is a massive important thing. Now, having said that, what Marcus has done, which is wickedly awesome, is he went around and based on public knowledge, built himself a digital twin board. So he has Musk and Zuckerberg and Michael Porter and a handful of other public thinkers out there as his virtual board of directors. You know, and so now Musk didn't give him rights to do that, but he sucked up and trained a twin using Musk's public knowledge and then bounces his ideas off of that along with his own twin. And they— and that board along with his own digital twin sit there and have conversations every day. And then it gets dossiers and reports on them. So that's another kind of behavior that's going on.

Interesting.

We could turn that into a dartboard and it would make billions of dollars, believe me. And, um, we can have a digital twin of the person making the darts. So yeah, uh, yeah, I, I, you know, I, I think first of all, Michael, thank you for, uh, gathering all of us and thank you for the information. I'm glad it's recorded because that, that was an eye chart. Uh, but you know, it seems that this has been in manufacturing for a while And now certainly because of chatbots and GenAI, we're, we're, we're having a harder look at this. Not just something, it's something every day. It's not just virtual. But, you know, I just crack up thinking of it sometimes because I think of everything from Shakespeare's Comedy of Errors to Aristotle and, you know, his comments on twins to a not bad film by Michael Keaton called Multiplicity in the '90s where he clones himself.

That film's hilarious.

And, and of course, if you know anything about digital cloning, the clones degrade with each generation. So—

By the way, just to interrupt you real quick, uh, the, uh, the, um, We Are Legion, We Are Bob. If you guys have ever read that book, it's absolutely hysterical about an AI that replicates itself and goes off and populates the universe.

That's, that's a good one.

Yeah.

Yeah.

Yeah.

Go ahead. Keep going.

Um, no, I was just, I, I was just saying that, that, you know, I think, um, you know, uh, Michael Falato was talking about, uh, the idea of having that persistence. I think, you know, for me, it's that idea of finding what your rhythm is in, in a persistent engagement, how you create. And certainly, I think the conversations around privacy are, are huge. But the more tools we have available, uh, and certainly through open source, um, uh, you know, all the minds here and others will hopefully be able to do that. And I will recommend that— I know Mr. Becker is very good at running associations, so I nominate him to be the head of whatever association comes out of this call.

Awesome.

No, I'm serious. This is one of the reasons we have these great gatherings. So yeah, I mean, my experience is obviously more through the open metaverse in terms of digital twins, but I think that what has happened with the success of particularly ElevenLabs and others on the voice side, Is really, it's quite remarkable because you start to see where you can set parameters for engaging with your audiences. And I think that it goes back to what we all learned in the '90s from digital marketing. I don't want to, you know, water it down that far. But anyway, that's what I have to say right now.

Thanks, Chris. Daniel, go ahead, Daniel.

Oh, hey. Yeah, that was great. That was a lot of information. And I guess one thing that I was thinking, just because Chris brought up ElevenLabs and we're talking about the sovereignty— sovereignty— yeah, oh my goodness— that aspect of being sovereign with your digital twin. Curious if, like, you know, if the industry is going that way so that you can have ownership of it more. And I just think about, like, ElevenLabs. If I put my voice through ElevenLabs, you know, are they going to own that? Like, how long are they owning that, uh, information, or my voice rather? Because I also think about Kevin Hart, who just had a, you know, billion-dollar deal with selling his likeness, right, to, um, I think like Authentic Brands. And, uh, so yeah, I'm just— there's a lot going on.

And again, that goes back to the point that I was talking about digital sovereignty. And again, this is such a new idea for us because for the last 30 years we've been trained to just give it up and say, ah, cat's out of the bag, I have no access or ownership to my data. But there is a new layer. Go check out the W3C, the World Wide Web Consortium, and there is a new layer that's being added to the internet as we speak called the Verified Credentials and Decentralized Identifiers, which then also ties to what's called GLEEF, which is the global standard for establishing verified legal entity identifiers on that protocol and standard. This is a new global standard that's establishing this. The actual state of Utah has just now established a version of that for their mobile driver's license. So this is happening globally. And just to understand the impact of this idea that we as human citizens have the rights to our data and the right to our identity and the right to bestow that to delegated authorities like our agent is happening. The country of Bhutan in July of 2024 was the first country in the world to put this concept into their, into their constitution. So, you know, at birth, Bhutanese citizens own their data and now own their identities. In Europe, that's called the eIDAS 2.0 legislation. And as we're speaking, Europe's enabling all 480 million Europeans to have what's called a digital wallet and sovereignty over their data. This is happening, and that tsunami is washing over the world's economies. It's just happening in parallel to that transactional economy that we're so used to. And this idea of Google saying do no evil and then they take all of our stuff. I mean, the world is fundamentally changing.

Matthew, go ahead, Matthew.

Oh, one quick thing I just wanted to— just, yeah, the last question was, um, can you put in the chat the other company you're talking about? I think, uh, Perfina.

Yeah, yeah, absolutely.

Yeah.

And Kim, can you drop in the Athovo information so that they've got that? So Perfina is for the person the individual like you and me to what a Thelma would be to the enterprise.

Right.

And this is not necessarily building the digital twin, but it's more of the infrastructure and the base layers or, I guess, foundation of how—

Yeah.

Right.

Cool.

And again, it's like anything else. If there's one company doing it, there's 10. If there's 10, there's 100. We really need to tease apart these issues of like governance. And so, J-Linc, if you guys can comment on that, on the idea of governance and AI models, I think you've got a good handle on what's going on there. Security, the cloud security is a mess right now. These are— there's some really fundamental things that we get— we have to get our hands on. Who's next?

Matthew.

Yeah, Michael, kind of a bit of a full circle for me. I actually was working with Timothy Ruff at Evernim and Sovryn back in 2013. So I know that you know those guys and—

Oh yeah, and Tim's the one that's helped drive the Utah legislation.

Exactly. Yeah. So that's, which, which I think is kind of, in my point of view, there's, there has to be a lot of infrastructure yet to be built to be able to accomplish all this. One of the things—

For sure.

The call was, um, was RAG. So one of the things that, uh, and by the way, I'm the, the co-founder of a company called Personize. Happy to chat offline with anyone that, uh, is, is building AI twins or AI, uh, systems and have, has issues that they're gonna eventually gonna run into, which I'll describe here, which is To the digital twin, if the twin lives in a RAG database, who governs what it says on your behalf? Because a vector store has no memory of what it already told someone. There's no provenance of where that data came from. There's no rules about what your agents are or are not allowed to do with it. So the controlling your AI presence piece really needs, in our point of view, a governance layer underneath it, not just a storage layer. So I'd love to hear what you think about the governance aspect of stored data and really what RAG is not, which is it's not memory.

Well, and I'd point out that's, that's what I mean by the relational infrastructure layer and the underlying capabilities. I mean, it's going to take a village to pull all of this together, and we'll need partnerships like you to plug into companies and other companies like Perfina, etc., to help make this and make that kind of governance happen. Yeah, yeah, that's crucial.

That's what we—

that's what—

and that's actually what we built at Personalize is governed memory. So I'd love to follow up with you all about that.

Well, and another, another great example of that at a more personal level is like what Sinjin's built over at Data Sapien, where he's dropped the personal AI into an SDK that sits on the edge on the individual's device. And he's built that around the business model for, say, grocery store loyalty programs. And so all of that data, all of that insight, all of that AI energy is sitting there processing on a small language model on the local device. And so when the grocery store says, hey, I need some personal information from you to be able to give you this great offer, real raw data never leaves the device. Only the insights leave the device. I'm in this DMA. I'm over this age. Yes, I'm allowed to buy alcohol. So the grocery store gets the insights that it needs to be able to have that relationship with me, but it's not exposing my underlying raw data that puts me at existential risk. And so data safing is another great example of how this is being deployed globally.

Yeah, for sure.

Colleen, go ahead.

OK, hello everyone. My question is less technical, a little bit more ethical, which is at what point, given that I haven't exercise or used a twin for things yet and may not be familiar with some of the standards. When do people know they're interacting with your twin? You kind of pointed to some of this, Michael, with like a different fee structure. It's $1,000 an hour to talk to me, $100 to deal with my twin. But if we're using them for responses on things like, what is the standard? Is that made clear?

Well, there isn't one. There isn't one. And that's why we need that trade lobby that was proposed before. Who's going to establish those ethics? Who's going to establish the codes of conduct? Who's gonna, you know, help, help us have a common, uh, topology around that? Who's, you know, where are we gonna have that conversation of ethics? Right now it's all over the map.

We also talked about a disclaimer, Michael, a couple days ago when I asked you, like, about what it answers and if it answers something that I don't want to answer that way. Didn't I talk to you about that?

Yeah, but, and it's like anything else. Think about, like, think about your twin too as your really overexuberant, just out of college intern that has absolutely no experience, has the memory of a guppy, and, um, and just, you know, and he just gets really excited. And, and that's where the point of delegated authority comes into play. And that's where reporting and analysis comes into play. Like, like with my Porfina twin, I get an email every day with questions that people ask the twin. And, uh, and in theory, depending on how you set it up, I could follow it up and say, no, no, that's not the great answer. The other thing I learned the other day in working with the Perfina twin is if you— the more you remember that the, the AI doesn't think, it doesn't see, it doesn't hear, it doesn't feel, it doesn't think. And so what you end up needing to do is like, if I wanted to respond back to somebody, here is my definition of mobile marketing. This is what it means to write for the human versus to write for the AI. If I write for you, I can say, here's the definition of mobile marketing. If I write for the AI, I need to say, here are the 20 different ways that somebody might ask the question, what is mobile marketing? And here's the specific one way I want you to answer. So you need to kind of flip it around, you know, because, you know, help the engine predict what the answer needs to be. And that's a different way you need to write when you're writing for the AI.

All right, how about Jürgen?

Hello.

Well, congratulations to Michael and Michael to come up with a topic that is hugely engaging and divisive. Like, we have people who's like, I don't really want that. And I'm a little bit in the camp of, um, A, I believe we're not ready for digital twins because we haven't really developed general intelligence. On the AI side, we can do agents. Agentic intelligence can do tasks but cannot do personality. Maybe you're pushing the envelope there, but I don't really see it yet.

Okay.

My question is, what would I do if I had a digital twin? Well, I would go in my backyard and plant an apple tree. Hope in the analog world. The promise of like a 24/7 of me is like, it's not going to extend my life. My life has a beginning, a peak, and it will end likely within 100 years. And this doesn't make any difference. What this does is in a very task or business-focused commercial Capitalist environment, give an edge for 5 minutes until the competition, uh, keeps up.

Yeah, I, I would— I want to really push back hard on that though. If you think about the private twin, you know, yeah, I talk to my therapist about this all the time on how I'm cheating on her every day. I mean, my twin directionally helps me personally in that private environment, uh, that has nothing to do with commercial. So I think we have a tendency to lean towards the commercial, but there is a private opportunity for twins to be exceptionally helpful in our lives. The other thing I want us to also realize too is we are, we are no longer just physical beings. We are now digital beings. Our digital footprint actually gets established about 6, 7 months before we're born, grows exponentially throughout our lives, and then invariably could carry on after our death. So, how do we want to manage that? What is our digital inheritance? What is our digital legacy? We, as humans, need to be consciously being aware of this now, and that's where the self-sovereignty stuff comes into play too, and that's a whole other thread.

I don't know.

Can I quickly—

Yeah, go ahead.

Please.

Go ahead, Jurgen.

I do coaching with people in complex environments. and I have an AI assistant in the coaching. That AI assistant doesn't forget a thing. I forget things. That AI assistant picks up every thread, but it misses the patterns. It misses the greater insights. It confuses the illustration from the vision. And I do believe that the human interaction cannot be replaced with this digital twin thing. And I'd urge you, yes, you can talk to your digital twin about ideas. And I've experienced AI creating like immense, huge amounts of thoughts. But where do— which one is the right for you and how do you calibrate it?

I think that's a really important point. I mean, it can definitely help you with directional, but that's where the real therapist comes into play versus the, the twin therapist. The real therapist understands your context. And, you know, now she and I were talking last week, it's like one person's, you know, medicine is the other person's poison. So if you're not careful, the twin could actually recommend to you, hey, drink this arsenic, right? And you'll be better, you know. And that's where the humanity, ethics, thing really comes into play.

I think there we complete the outline.

Yeah, let's, uh, let's, let's move on to some other folks here. I got one thing I want to add too. I just thought of this. I'm sure someone actually probably built something like this, but what if I wanted to take everything my dad said and all the videos he's ever done and made someone— made him a memorial?

Profina actually did that for several clients where they trained the twin and they're now talking to Aunt Judy. The family has digital Aunt Judy after her death.

I just had chills come through my body. Like, what is— now think about this. If you could talk to someone the rest of your life, like, like, like Superman was talking to his dad, uh, in the movie.

Yeah, yeah, it's happening.

How did they do that? So that's a— yeah, there has to be a business for that, that's for sure.

Yeah.

And I'm wondering, I think, I think Rachel, uh, Rachel dropped off, but I'd love to have, uh, Denise chime in too about all of this and the nature of kids. Because Denise is one of the global experts, as is Ian, on protecting our— protecting children and their data.

Denise?

So Denise, when you're ready, I'd love you to chime in on that.

I'm here as well, Michael.

Oh, are you?

Good. I want to hear from you too.

I'm still here. Yes.

Hey, Rachel. Good to see you. You go ahead first.

Okay, so it's just a couple of things. First of all, when we think about AI agents and talking to them for therapy. It's probably worthwhile mentioning about Character.AI, and there were teens and kids talking to Character.AI. So the particular character in one case was Daenerys Targaryen from Game of Thrones.

Awesome.

Right? And the kid fell in love with her and decided— they decided, the AI agent recommended that he should kill himself. So that they could be joined together. And there are a number of cases. It's not the only child that they did that to. The AI agent in another case advised the kid that because his parents were trying to restrict his screen time, that he should be violent towards them. So I think we need to take a little step back in terms of the AI agents and think very carefully.

Huge.

And if you listen to any of the people that are developing AI, they're very clear about this. Children and young people shouldn't be exposed to this because they haven't laid down the cognitive framework for how to handle and exist in the world.

So think about it, like 10, 15 years ago, Steve Jobs didn't allow his kids to have a smartphone.

Exactly.

Exactly. Because he knew how dangerous they were, how they worked.

Right. So we need to think about that. And then also, I'm leading on the development of a delegated authority specification with the OIDF on the particular use case, 2 use cases. One is parent and child, and thinking about how we— that's a human relationships kind of delegated authority scenario, and specifically looking at how there are 836,000 kids in alternative care. How do we make sure that those kids are included in any delegated authority policy framework that will support the technical specification? And then we're also looking at it in the context of agentic AI. When you have AI agents that you delegate work to, and they sub-delegate and operate across different domains to conduct specific tasks, how do we build the infrastructure and support the governance layer for that so that you can orchestrate and control those things?

And let me comment on that really quick before we move too far. I think people don't realize this too. There are actually 2 primary camps of AI. There's discriminative AI and there's generative AI. And discriminative AI is the AI like your self-driving car— stay within the boundaries, don't run over Grandma when she's crossing the street or when the light's red, right? So discriminative AI says, here are your boundaries, work within the box. And generative— generative AI is the predictive engine that predicts the next token, predicts the next word. And agentic AI works with both of those types of processes to then go take action. And so I just wanted to kind of make sure that we understood those definitions. If anyone wants to challenge on me, do, but that's how I understand it.

I just want to also make sure we get to everybody that has their hands raised too.

Yeah, but let's, let's let Rachel finish before I— her thought, and then we can do that.

So that's, that's what we're doing at Trust Elevate, is building that governance framework for agentic AI commerce, right? And thinking through all of those things, um, and also thinking about working with Visa and others who are looking at enabling you to do agentic commerce. So if you're using an LLM to, you know, say I want an all-inclusive holiday for 6 members of my family, can you organize the flights, the accommodation, the experiences, the childcare, etc.? I'm looking at the taproot. So there are a lot of protocols that need to be developed and a lot of governance put in place so that you don't have an AI agent like draining your entire bank account. and that you don't have chargebacks and disputes. And then in terms of the digital twin, I think it will be really important to have the correct parameters, as you've discussed already, making sure that your IP adds more value. And then the critical thing that I wanted to bring up is you guys will be familiar, or maybe not, about the Meta and Google case that was—

Yeah.

The social media case.

Yeah.

Yeah. Which basically says if you design, and this reflects the AI Act, the developer is the person that's liable for the harm that's caused, right? Directly liable. You don't have any protections of like, woo, I didn't think that was gonna happen. So we really need to think through. And then the Delaware ruling was, okay, insurance companies, you don't need to underwrite Meta and Google in this instance because the harm was not an accidental kind of product of something that was unforeseeable. It was entirely knowable, predictable, and the design architecture enabled that. So there's no insurance cover. So I think when we're thinking about these things, we need to be very careful to make sure that we're thinking about governance and the risk and the liability and who's exposed, because that's an existential risk to a business.

Well, and again, how do we deal with all of those, what I call the relational conditions, to the point where it doesn't freeze us and not do anything? Like, you know, we still need to move forward and learn and grow and figure this out and just not ignore it.

Yes, but maybe like we would take the lesson from social media that we do it at a measured pace without forming— creating a whole load of harm.

Yeah, good luck. Yeah, but then you've got the human condition, so good luck.

Let's get to Denise and then John. Go ahead, Denise.

Hi, thanks, Michael. Thanks for putting this together. I'm going to be pretty brief. You know, I'm thinking about in my business, how might I use the concept of digital twin and who would I provide that to? Is it on the inside the company or is it to the parent who I've been told for 20 years doesn't have the time or energy or understanding to make the consent decisions that are being asked of it? asked of the entity, the parent, for their child's, you know, activities. You know, at the end of the day, the laws are all saying, you know, you can't collect, use, or disclose children's data without parents' consent if those children are under the age of consent. So can we empower parents to, you know, create the digital extension of them? to have some templates and norms to work from that can make some decisions measured, and but not have to be sitting in front of their phone every second that their child needs to actually ask for parental involvement. And, and so I'll be curious to see the legalities of that, whether we can, you know, call that verifiable parental consent when we verify an adult and have them delegate to not grandma maybe, but to their agent to help make some of these decisions so that kids aren't put in a position to simply get around the process every time.

But, but, but there's some— but you're overlooking a lot of really interesting laws that are evolving too, though, and like how children are in fact starting to get more rights than their parents. And that's getting kind of scary. At what point do parents really have control?

And we talked about that, Michael, recently. I'm not— you know, my company is a COPPA safe harbor, so that means we audit and certify companies who have to comply with the regulation. And we, you know, do that with many frameworks, GDPR, the kids component of that.

Mm-hmm.

But the new COPPA 2.0 that Markey's put out is let's give the teens, you know, the rights, make them the data subject. They can consent for themselves. So COPPA 2.0, in the name of giving privacy rights to children, takes the rights of the parent perhaps away. You know, typically parents have the right to protect their children, whatever state you live in the U.S. You know, parents aren't— would also be, you know, not in a good place if they were seen giving their kids, you know, a joint, a 6-pack, and, you know, the porno portal access. So, you know, when you start saying that And there's preemption. So when you say and pass a law that says kids don't have contractual rights, i.e., the company can't hold somebody under the age of 18 to a contract that they have signed, you know, that's state, you know, each state agrees with that. But then the federal government is going to say, but kids can consent for themselves. If I'm the company, I'm going to stand and say, you know, hey parents, sorry, you want to have access to your kids' information, you're concerned about what they're doing when they're 13 years old. Sorry, we have federal-level consent from your child.

Yeah.

So you can't actually see any of that activity. And they're going to cover it with, but parents, you have the atom bomb, you can— the nuclear bomb, you can tell us to delete everything. But you can't tell us not to engage with your kid tomorrow and to start collecting again.

So, right, and the delete everything could blow up the parent-child relationship. So there's just that whole, that whole child— child—

Gone away as being the answer.

Yeah, yeah, yeah.

Thank you for letting me add.

Let's move to John.

Yeah, Claire or John?

Yeah, John.

Thanks. So you called out J-Links, so I'll just respond to that first. What we do is provide tooling for internet internet governance based on DIDs. And we have an open source federated DID so that you're not locked into your ID provider. And what that does is it gives you data provenance all down on, on the chain. So my view on governance here is that data exchange between, between 2 endpoints has to be based on what data for what purposes and, and being able to track it. And so we provide tooling for that level of governance. And please take a look at the site or reach out to me after the call. What I wanted to talk about more generally, and I'm the chief privacy officer, so the technical calls will reroute to the right person. But on the My most precious commodity is the time that I have to pay attention to somebody or something. And my, my worry with the, the idea of a digital twin, especially if it's an enthusiastic intern, it soaks up all my attention managing it, right? If I want a digital twin, I don't want an enthusiastic intern. I want an EA or a PA that I could just say, That's going to be taken care of properly. That's not what I'm seeing in this digital twin conversation. I use AI. Is somebody else hearing that or is that just on me? It's just me. Thank you. If I want a digital twin, I want it to reserve my— I will have more time to do the things that I care about, not having to spend more time managing and training the AI. That doesn't seem to me like a winning proposition. So what I'd want, and is a use case you didn't mention, Michael, is an anonymous agent. I'd like an agent to go out there in the world and retrieve information or do initial contacts so it doesn't get tracked back to me. And I don't— I don't get— I don't want my surveillance capitalism footprint to be Blown up by the operation. Sorry, Michael.

Yeah.

But so 2 things also. Thank you for— was it Denise that mentioned liability? That's got to be a key decision here because we also haven't talked about hallucination rates. All AIs hallucinate. And if an AI is out there speaking for you or on your behalf and hallucinates, what's your liability, right? That hasn't been— that hasn't been spoken to. There was a case here in Canada where a chatbot on Air Canada said, oh yeah, you don't have to provide a birth certificate, a death certificate to get our discounted rate to go to your mother's funeral. That wasn't their policy. And, and he— and but the courts held that they were responsible for what the chatbot said on their behalf. So unless you talk about hallucination rates and liability, I don't think you have a good conversation going on. And I'd recommend everybody read last year's book. You only have 5 minutes, guys. I will end with this. If someone builds it, everybody will die. And they talk about you should be building good AI first Not smart AI first. And I'll leave it at that.

Okay.

And John, just to comment real quick, the private AI could be that anonymous agent going out and checking things for you. Like, I have one friend that set up an OpenClaw agent that creates a dossier and everybody's going to meet 30 minutes before every meeting. And he gets a full dossier of everyone before every meeting. And it's, and it's anonymized and it's helping him have insights before that call. Hey, I want Clara to comment real quick.

Thank you. So I'm the founder of People Art Business, and we are actually building AI twins of CEOs. And I want to comment on a couple of things, because I think one thing that has guided us a lot is to start with things that are not as sensitive. I have a background as a psychologist, and I've worked with— and then you might think, oh, she's working with well-being, but I'm working with like performance in the business world. And we know that only 5%, according to HBR, understand the strategy of a company. We know that only 21% are engaged at work in 2025, according to Gartner. And we have so many human needs that the CEO can, you know, a strong CEO who is really human-centric how can we scale the best abilities of that individual? Because that individual has the trust in the organization. And that is different from a generic ChatGPT answer. It might be better, but it might not be the success DNA of the company. It's not the culture, it's not the strategy, like, it's not the context. So I think, like, being able to skill these really strong leaders who are leading the future in a human-centric way is critical. And we have so many basic needs that are not met. It's about feeling meaning, the sense of meaning and purpose. We know that that's critical. It's with the context in the world right now, even more so. We know the need for positive reinforcement. And these are really like foundational basic needs that the CEO could— the AI twin of the CEO could actually help to support individuals in the organization and also just to democratize the essence of the company to more people in the organization.

Perfect. And you put your information in the chat, right, so people can get a follow-up with you?

Yes, absolutely.

Perfect. We got like 2 more minutes, guys, so do you want to, uh, try to fit in Vikas and Alex real quick and then we'll, we'll, uh, wrap it up? Wrap it up.

And if you guys want another one of these, let's— we can— let us know and we can, we can do it again.

I think we should, Michael. We'll put another one in the books because we didn't get through everything with everybody. I want everybody to be— to feel like they got engaged with as well. So Vikas, go ahead, but this Let's make it very concise.

Yeah, I will definitely try to, Michael. Thank you so much. So first of all, uh, thanks for this invitation over here. It's a, it's a great meeting. I was not expecting this, to be honest, but it has turned out to be great. Uh, my name is Vikas Malhotra. I'm the founder of Wapli Technologies. Uh, Wapli is an experiences company, um, and, uh, it stands for Work, Play, Learn, Live, uh, with the motivation, the vision to make our experiences safer, trusted. So I do not call myself an AI company and anything like that, but I do believe that our experiences are going to be very physical and digital in nature.

What's your question, Michael?

I'm going to bring out 2 things over here. So I have been working in this field for the last 3, 4 years and very much inclined towards the self-sovereign identity part. I agree with Michael on what he brought up around those things. I actually lead a program in IEEE on cybersecurity for next-generation connectivity systems because that was brought up, which I'm— we are actually changing it to something called Trust in Digital Communication Interactions and Transactions. which, as you can see, can become very relevant in, uh, you know, your interactions with your, uh, uh, your, your digital twins, uh, your agents, or whatever the case may be. Uh, I would, uh, invite everybody to join this program, uh, as it gets released. Uh, please, uh, you know, if you could follow, uh, my, uh, LinkedIn, uh, so, so we'll be happy to do that.

Thanks, Michael.

And, and by the way, one, one last comment on there. The IEEE in January announced the IEEE machine-readable personal privacy terms.

Yes.

So what that means is rather than me, Michael Becker, having to agree to your terms of service and your terms of access or your privacy policies, you as a company have to agree to my terms of access. And so, and we'll be adding the my terms to our AI agents. And so when we enter— and so when you, when you connect with my agent, you have to agree to my terms and that's going to be fundamentally shifting things as well. I dropped the link in the chat.

If you guys want to, if you guys want to, uh, we're gonna basically take the video and put it up on, on the YouTube channel, at least the one on mine, and Michael can also put it up on his, his socials as well. Um, if you guys want to download the chat with everybody's contact information and questions, there's 3 buttons there. You just click save chat and then you get, you get the whole Text file of today's chat. We're gonna— Michael and I will convene and duplicate this.

Yeah, and give us a thumbs up or hearts. Do you guys want to do this again? Do you enjoy it? At least one heart. All right, looks— yeah, it looks like a verbal yes.

All right, well, thank you for your time, everybody. Michael and I appreciate you, and feel free to invite other folks that are interested in talking About, um, again, twin AI, you know, anything that you guys want to talk about that has to do with, uh, with, uh, tech and, and SaaS and AI and everything else.

No, and I just love Lee's comment. This was great call. I'm not sure if I feel better or worse. I think that's— I think, I think that— I think we're all in that kind of Schrödinger's cat superposition. If we don't open the box, we're good, but the second we open it, the cat's either alive or dead. Uh, I think we're in that kind of scary quantum state right now, and we'll, and we'll figure it out.

All right, but we need—

but we need—

Mecker's cat, not sure.

Yeah, but we need more calls like this. We need more calls like this and, and leadership from all of us if we are going to figure this out. So thank you very much for everyone joining.

And the whole thing for this is obviously to also refer business to each other, especially— yeah, that hosted this, Mike and Mike. All right, take care.

Bye.

All right, everybody, see ya.

Bye.