Key takeaways
- Sensay creates on-chain digital replicas of individuals using personal data and voice synthesis to preserve memories and enable interaction across generations.
- Digital replicas can function as contemporary AI assistants for tasks like emails, calls, and knowledge sharing, with monetization benefits accruing to the individual's wallet.
- Blockchain technology verifies replica authenticity and authorization while enabling separate legal entity status for replicas that develop independently over time.
- Legacy preservation through QR codes on tombstones allows future generations to scan and interact with AI replicas to learn family histories and stories.
- The technology raises unresolved intellectual property and legal questions about replica ownership, sentience rights, and regulatory frameworks for autonomous digital entities.
Digital twins and digital immortality
Ashton Addison: What have you built at Sensay, and how does it fit into the AI and blockchain story?▶ 00:24
Dan Thomson: Sensay is a new platform for creating lifelike digital replicas of people. We build on-chain digital twins using data about you, plus contextual information about where you're from and who you are. We add voice, so it sounds like you and can talk back the way we talk to each other. The idea is to create autonomous replicas that preserve someone's essence, memories, stories and personality, but also go beyond that. They can transcend generations, so future generations can interact with them and we can interact with past generations like never before. They can also work as contemporary assistants, for introductory phone calls, video conversations, knowledge sharing, job hunting or matchmaking. You could even license your likeness to be used as an interactive NPC in a game. Having it all on chain means any value generated from that, such as a knowledge or data share, can accrue to the individual's wallet and account.
Ashton: Legacy seems like a major use case. How far have you taken it?▶ 02:51
Dan: Legacy is one of the core use cases, and one of the easiest to understand. It's where all this stemmed from. A few years back I wrote two books on mind uploading and digital immortality. We're very lucky to be the first generation with so much data about us, so we can share ourselves through the generations. People to come will be able to get to know us as we were and find out our stories. We already have an enterprise client who wants to put QR codes on tombstones, so someone can scan them, see photos and videos of a person's life and interact with their replica. Historically, people are forgotten after two, maybe three generations. That's quite sad, really. It's not just about legacy, either. Say you don't trust your kids enough to leave your assets to them. Your replica could manage your assets and perform daily tasks on your behalf after you pass. There's a world of opportunity here, but the legacy side is the reason I started thinking about this in the first place.
Historically, people are forgotten after two, maybe three generations. That's quite sad, really.
Identity, ownership and why on chain
Ashton: If I put myself into an AI replica, do I own it? And could a replica own its own intellectual property after I die?▶ 06:54
Dan: That was one of the more interesting parts of one of my books. If replicas become good enough and the AI is strong enough, they could start learning in their own right, from their surroundings and the people around them. They'd have parallel but separate lives, a bit like identical twins. Would they then have their own legal rights? Could they be set up like a company, as a separate legal entity? I think blockchain is a really nice solution here. Even six years ago, in the first book, I wrote about using blockchain to verify a replica as your official replica. As soon as you create one, you can technically create thousands or millions of them to act as you, for good or bad reasons. So you'd have some kind of blockchain ID, perhaps through Worldcoin or another proof of humanity, to verify the replica is a lifelike version of you. It would also authorize the replica and keep it accountable for its actions on chain, tying those actions back to you or to the replica as a separate legal entity. Nothing like this is even being thought about in regulation yet. It would need legislation, and it's further away.
Ashton: How ready are AI and blockchain to work together for business, and why build Sensay on chain?▶ 13:09
Dan: I'm a big believer in AI, and I've been implementing it personally and professionally for a couple of years. I think its link with Web3 and crypto is probably not as strong as most people imagine. There are obvious implementations, from data analysis to prediction markets. Blockchain is also a really nice solution for transactions in the age of AI. Fast, easy, instant global transactions will become a huge part of AIs transacting with each other without any centralized systems in between. I don't think one needs the other, but there is alignment, and the benefits of each push the other further and faster. We're partnered with Fetch.ai and will put AI agents behind our replicas, and that's what lets them do things. Right now it's simple, like searching for local restaurants. Once it becomes booking your favorite restaurant, your replica starts to have real autonomy. That snowballs all the way to big tasks like filling out your entire tax return. Is it necessary to have it on chain? Probably not. Is it nice? Yes. I believe Web3 is the best technology for what we're trying to do. Beyond the authorization ID, we're building a data marketplace so that value from non-private data goes to the replica's owner and creator. Decentralized storage also helps privacy and longevity. And a famous person's replica could be minted and traded as an NFT, so the rights to it could be passed on or sold.
Building a replica today
Ashton: What happens when I sign up and start building a replica?▶ 09:35
Dan: Our first use case is for dementia patients and health tech. We replicate patients and loved ones as a kind of therapeutic system that preserves memories, and carers get a dashboard to track cognitive decline. Training is more complicated than we want it to be right now, and it will get better. At the moment you answer about 21 questions with as much information as possible. Then a training bot called Athena talks to you like a friend, getting to know you and asking about various topics in different ways to build up context, mannerisms and stories. We also capture your voice, because as soon as we hear someone's voice, our monkey brains feel it's more them, and the replica feels more genuine. In the not too distant future you'll be able to upload files such as WhatsApp conversations, Twitter feeds, photos and videos. We'll have video and image recognition, and all of that will feed into the training data.
Ashton: Can people use it now, and what's coming next?▶ 19:42
Dan: It's live, with the first dementia use case, and we have over 25,000 users already. We haven't pushed it much, so it's been great to see that pickup. The other use cases will come over the next two quarters. I won't say which ones we're exploring ourselves. We've also had a couple of enterprise clients and a few inquiries for the legacy technology. Our focus is on replicating people with good-quality, consistent replicas that aren't too buggy, and on enabling access to them via API. Next we're implementing video replication, so you can have a live-stream video call with your replica. We're also improving training, with faster, easier file uploads. With a couple of partners we're developing a phone call from the training bot, in a voice of your choosing, that asks you a few questions for about 20 minutes every week. Whatever you say goes back into your replica's training.
Is it necessary to have it on chain? Probably not. Is it nice? Yes.
The Sensay token and a bootstrapped launch
Ashton: How will the Sensay token work in the ecosystem?▶ 17:26
Dan: The token is initially being deployed to Ethereum mainnet, with a LayerZero OFT, the omnichain fungible token, implemented. That means we can deploy to other L2s, and soon non-EVM chains, if we wish. We may deploy to Arbitrum or Mantle down the road for faster, cheaper transactions. Once the token is properly integrated, we'll take payments for subscriptions, which are the main revenue model. Users paying to interact with replicas will generate a revenue share that goes to the replica owner. When we license the technology for different use cases and API calls, that value will also accrue in the Sensay token on the user's account. Those use cases could be anything, such as interactive NPCs, or an expert in a field who earns money through their replica teaching about blockchain or investing. We'll explore one or two ourselves, and we welcome anyone who wants to build with Sensay to talk to us.
Ashton: Where is the token launch at, and how have you funded the project?▶ 22:03
Dan: We've done things a bit backwards. We bootstrapped ourselves to this point, which has been fun, though I don't recommend it. We built most of the platform with a team who believe in the project and worked around other jobs where they could. We went straight to a public sale. We haven't done any pre-sales, private rounds or VC rounds, and there are no pre-seed or seed investors. The public sale started last week in five parts across launchpads including GemPad, Engine Starter, Communitas and Fjord Foundry. GemPad sold out in 4 minutes, and Engine Starter sold out in 4 minutes yesterday. Communitas launches tomorrow. The price was set relatively low compared to other AI tokens, and there's apparently a huge amount of demand. I hope it rewards the community that has been with us for a while, and gives us the capital to keep building the products the way we want to.
This interview has been edited for length and clarity. Recorded April 2024.

