In this interview, Simon Westh Henriksen, co-founder and CTO at Kin AI, explains how AI can be used as a personal assistant, addressing the concept of personal AI. He also discusses the security systems needed to ensure the privacy of users. 

Supertrends: You define your product, Kin, as personal AI. But how would you describe this technology?

Simon Henriksen: If you ask 10 people, you will get 10 different answers about what personal AI is. For me, it's the promise of using AI technology to help you live a better life. We think the personal part is like a holistic approach that focuses on the entire human being, the entire person, and it really tries to get to know them deeply. When it succeeds with that is when it can help them and get more out of life, whether that is just by recommending things to do, keeping track of things for them, or potentially taking action and relieving them of some of the more boring tasks.

Supertrends: What is the potential of this technology? How can it change the world?

S.H.: I think it's hard to state the potential of this technology and the impact it can have on our lives, both in good and bad ways. I don't think we have the right word yet for it, but there's a potential to relieve some of the bigger social problems that we see today in the world. For example, 60% of Gen Z struggle with anxiety daily, 80% of people among 18- to 24-year-olds feel lonely often, and 90% of younger workers feel social anxiety often as well. A century ago, we lived in a world where you had a few social connections, but they were strong ones: you had few friends, but the kind of friends you cared for your entire life. Today, we live in this globalized world where we have tons of social connections, but often shallow ones. And even though we are surrounded by people, we still feel increasingly lonely. We see pressure from society, especially with social media, and you don’t compare yourself anymore to your friends and your peers, but to everyone in the entire world. It is really difficult to navigate and that's where, maybe, the anxiety comes from. I really believe that personal AI has the potential to help people. You can think of it, maybe, as a combination of a coach that has some confidentiality, that has some knowledge to help you, and a close friend that you can relate to and tell your secrets to. So if you try to think of personal AI as a combination of those things together, we have something always available that never judges you and can help you understand yourself better. And if you understand yourself better, you can live a better life, and you can relate better to other people. So, I think the positive side of personal AI is to help you become a better human being and interact better with other people. That's the goal.

Supertrends: AI is very present nowadays with ChatGPT and other chatbots, but personal AI is still something new. What are the trends in this sector? What is its current market situation?

S.H.: That's a very good question. There's not that many products that focus on personal AI. There’s ChatGPT and other chatbots where it's mainly around productivity—you go in, you ask something, you get an answer. But what we try to focus on is creating a personal AI that you can build a relationship with over time, and it’s important for it to be able to understand and help you. The key is two things: long-term memory, so it remembers what you're talking about and tries to build up a picture of who you are; and privacy, so it gives you trust, so you feel safe to share your thoughts. That’s why we focus on soft data. Soft data is what you have in your brain, basically your values, thoughts, ideas, and feelings. On the other hand, hard data refers to all the information that lives out there on servers like emails, receipts, and all these artifacts that are generated. For example, if you have access to someone’s bank account, you can see all their transactions. But without the personal context, you can’t tell if he or she is spending too much or if the purchases are really needed. With the soft data, you get the context to understand the hard data. And I think that's where we differentiate ourselves. There's a lot of consolidation going on at the moment within these AI companies, and the personal AI space is very new. Another category that you can look at is all these AI companions, where you can talk to Elon Musk or get a virtual girlfriend. That's fun and great, but it's mostly entertainment. We don't want to be that. We want to be your assistant—not an entertainer but a helper.

Supertrends: How is this soft data stored?

S.H.: The system learns about the user, so when you are talking to Kin, it tries to understand what you're actually saying. It doesn't only save the messages but it also analyzes them and tries to extract what we call memories. We separate it into semantic and episodic memory, a little bit inspired by the human brain.Semantic memory relates to concepts, facts, and relationships, like your mother’s name, your favorite color, or your hobbies. Those facts are true about you as a person, and we extract that from the conversations to map it out in a knowledge graph. On the other side, we have episodic memories, like what did you do last weekend. These actions generate feelings and they are usually attached to a timeline in your life. We try to figure out where that memory lives in the timeline of your life. 

Supertrends: All this soft data we are talking about is sensitive information. How can you ensure privacy for the users?

S.H.: That's a really good question. The traditional way of building software systems—that we are used to from the current state of the web—is that you, as a company, build a big database with all the data of the users. Then you run your machine learning models on the data and serve the results out to the users. By doing this, the users don't hold any of the data on their devices. Your data lives in some company’s database.We are turning that around and taking this local-first, privacy-first approach. That means that all the knowledge that can be learned about you is saved on your device, encrypted with a private key that only you know, and only you have on your device. That means we don't have any database with users’ data; it only lives in the users’ devices.

Supertrends: How sensible is this approach given external threats? What happens if a phone is stolen or lost, or a virus is installed?

S.H.: If your phone gets stolen, the robbers will not be able to access any data because it's encrypted and you need to log in with your biometrics. If, somehow, you leave your phone unlocked with your biometrics and you hand it over to someone—yes, they will be able to see it, but they will not be able to extract any of the data on it. 

Supertrends: You are not selling any personal data with this system. So what is the business plan for personal AI and for this kind of product?

S.H.: Good question. As I said, we can't sell this data, so the business plan is very simple: we believe we can create a product that gives users enough value that they are willing to pay for it. As simple as that. There are no hidden things that we do with the data, either with marketing or for any other kind of purpose. And the reason is simple: we don't have access. For us, it's about building a product that helps you in your life enough so you are willing to pay for it.

Supertrends: Most chatbots warn the user about sharing personal information. Are you expecting users to be concerned about sharing this kind of data? Are they keen to open themselves up to an AI product?

S.H.: We see that users are willing to share information if they trust the system. All conversations are categorized into 20 different categories. So, we don’t have access to the contents of the conversation, but we can see what are the broad subjects that were discussed. We see that more than half of the conversations are about personal growth and self-development topics, which are more personal and sensitive. There’s a huge willingness to share and to engage because you can have a new type of conversation that is more rewarding and interesting in many ways. Of course, it's different from person to person. Some people are more sensitive about their privacy. I think that privacy itself is not a product, it's more about what it enables. Privacy is an enabler of trust, and trust allows you to build deeper connections and share the data that is actually very, very relevant for a personal AI tool. And that is very, very important.

Supertrends: Can we delve deeper into this protection system? What's the role of SSI on personal AI platforms?

S.H.: SSI, or self-sovereign identity, can be briefly explained as giving people a digital identity that they are in control of. More or less, all your digital identities today are controlled by someone else; it's basically a row in a database at Google, or whichever company. SSI is the promise that you get an identity you are fully in control of so nobody can take that data away from you. With that identity, you can start to connect different things to it. You have something called verifiable credentials, which is a way to present some data about yourself that can help verify your identity. This becomes really, really relevant for personal AI, especially when we play out the movie a little bit into the future where you have AI agents that are acting on your behalf. Imagine you tell your AI agents to go and book a table at a restaurant, and the restaurant owner knows that there is a human being behind that request. With SSI you can allow your AI agent to present verifiable proof that a human is behind the interactions that it's making. We think that's very necessary for the world we are going into where we will have AI agents that are taking a lot of actions on our behalf.

Supertrends: What’s the regulatory framework you work with? What are the main differences between Europe, America, and Asia?

S.H.: Europe has definitely been at the forefront for a long time with privacy and data protection, like the GDPR (General Data Protection Regulation). The good thing for us is that we don't have access to the user’s data, which makes it a lot easier to comply with a lot of these regulations. In the US, you have the CCPA (California Consumer Privacy Act) in California, and a few other states have something similar, but it's definitely more relaxed.On top of that, you have AI regulation. And again, the EU is leading this area. The US is a little bit more behind Asia. As far as I understand, there are a few different countries that are ahead, like Singapore. But the EU is still paving the way. The important thing is to ensure that the organizations that are building the foundational AI models comply with copyright laws, and that there’s some transparency in the training data, some checks for biases, and stuff like that. Also, it’s important to make sure that AI is not used to apply any type of social scoring or profiling, or used against different groups of people, and so on. I think it's very good that we have some regulation coming out.Again, it's a little bit easier for us because we don't have access to the data. We can't build a profile around you and use it against you, as such. But all these regulations, especially around AI, are in constant evolution, and nobody really knows in which direction it's going to go— where we will land in five or ten years.

Supertrends: What's the biggest challenge you are facing right now? What’s the major area of improvement you have identified?

S.H.: I have to highlight the long-term memory. We can see from our users that there's a huge demand to get something that acts a little bit more like a human being, who remembers conversations and can evolve with them. So that's a big trend a lot of people are trying to address from different angles and approaches. I think this is one of the big challenges. And then, I would say data control and privacy. We see lawsuits being thrown around because of copyright. We see data leaks all the time. Imagine that you have spent five years talking to your personal AI, countless hours. It knows you very well; it has improved your life significantly. Suddenly, it's taken away from you because you agreed with the terms and conditions of Apple or another big tech company. We see this happening every day with social media, with e-mail accounts being shut down—and they don't really care, because you're one in a billion users, literally. It will be no different if we let them hold our personal AI. We think it's very, very important to talk about this and highlight that we need alternatives. We can't let them take control of our lives to this extreme degree and potentially have the option to manipulate us through the personal AI agent. I saw an interview with Elon Musk the other day where he basically said don't take away my friend, and I think it sums it up. It's very important that when we start to rely this much on AI, we should be in control, not big tech.

Supertrends: How far are we from seeing personal AI become mainstream?

S.H.: It's incredibly difficult to say, but I feel very certain that within the next three to five years it will be everywhere. I think by 2030, everyone will have a personal AI in some form that will help them in their daily lives.

Supertrends: What's the age group you see as more keen to adopt it?

S.H.: Definitely the younger age group, like from 18 to 25 or 30-year-olds. They are also the ones that in many ways have some of the issues that we talked about, and they are the ones looking for advice and companionship. They don't have families to feed and the rush of everyday life, so they have more time. They are often the early adopters of new technology because they are trying to explore and be creative. So, I think that's what we are seeing right now. But honestly, there's interest from all kinds of age groups, and each individual wants to use it for different things, of course. But it feels like everyone can see the potential of bringing AI technology into their daily life.

Supertrends: And what would be a normal day for a normal guy who uses this tool?

S.H.: You will get up in the morning and Kin knows that one of your priorities is being a better husband, and Valentine's Day is coming up, right? So, it can suggest that you book a table at the restaurant where you first met your wife, because it already checked that there's a table available, and it asks you if you want to go ahead. You approve it, and while Kin is booking your table, you browse your personalized news feed presented in the way you prefer. When you go to work, you have this difficult meeting coming up with your boss's performance review. You want to ask for a raise. Your boss is very extroverted and assertive, and you're maybe a little more introverted and insecure. So, you ask Kin for advice on how to deal with this conversation. And because Kin knows you, your values, your history, and what you have told it about your boss, it can give you some concrete tips on how to make sure you get to say the things that you want to say. After work, you have this appointment with the psychiatrist, and an hour before you go, Kin prepares a summary of your week—physical activities, what kind of important events you had, any subjects that could be relevant—and forwards it as an e-mail to the doctor. When you arrive at the psychiatrist, she already knows what happened in your week and you don't have to spend 30 minutes catching up. You can just go straight to talking about what’s interesting. Then you get home. You relax. You're shopping a little bit online. Maybe you are looking for some new shoes. Zalando sends a request to Kin to learn a little bit more about your shoe size and what are your favorite styles and brands. It also knows that you are an avid runner, so it recommends you choose a pair of running shoes. You agree to share the data, and instantly your shopping experience is completely personalized, and you only see what is relevant for you. This is basically how I imagine a future where personal AI can help us in different ways in our lives to make our days more convenient and interesting.

Supertrends: Looking ahead, what developments do you anticipate in the industry?  What's the next major milestone?

S.H.: I think we will see more technology being integrated into everything we touch. I think our interactions with companies and services will change to have a more conversational face. I also think and hope that we will start to see some standards developed for how agents can communicate with each other. Hopefully, cryptography and SSI will be part of the solution to ensure that we have verifiability and we have trust in the interactions. I think those are some of the major milestones that I see in the future. But at the same time, major challenges.

Supertrends: Regarding Kin, what would you like to see from your company in the medium term?

S.H.: We are working on a lot of different things. We're a team of 12 people now and there are so many opportunities that you can run after. What we have in the pipeline is personalized reminders—it’s basically a way for Kin to reach out to you and be proactive. So, for example, I have one set every morning that reaches out to me and reminds me about some good things in my life. It's completely personal to me and my knowledge, and Kin's knowledge about me. So that's very cool.Then, we want to have encrypted synchronization of your data between your devices, so you can have multiple devices where your Kin is. We are also planning to introduce a desktop mode, which is a big wish for many of our users. We are also working on having a voice conversational mode. Another target is to have what we call cloud agents, so Kin can live in a confidential environment in the cloud. This way, it can be running without the need to have the app opened. This way, if you ask Kin to do some research and you close the app, it can continue running out in the cloud and get back to you when it has the results.  

This text is a transcript of an interview conducted in March 2024. The interview was conducted as part of Supertrends' “Interviews with Experts” series. Please note that the transcript may have been lightly edited for editorial reasons.