Join Delphi Research today and immediately get access to our full Member Portal!
Join Delphi Research today and immediately get access to our full Member Portal!

Aaron and David McDonald: Altered State Machine is Enabling AI Ownership via NFTs, Hereditary AI and Minting Brains

Oct 5, 2021 · 75 min media

By Tom Shaughnessy

The Delphi Podcast Host and GP of Delphi Ventures Tom Shaughnessy sits down with Aaron and David McDonald, founders of Altered State Machine (ASM), a world-first decentralized platform for developers and NFT owners to create, train and own intelligent NFTs. They discuss integrating artificial intelligence (AI) into NFTs and gaming, use cases for AI in DeFi, the infrastructure of ASM brains and much more!



Interview Transcript:

Tom (00:00:02):

Hey everyone, welcome back to the podcast. Today I’m thrilled to have on two savage founders who I’ve spent a lot of time with over the past couple of weeks, I have on Aaron and David from Altered State Machine. How’s it going guys?

Aaron (00:00:14):

Hey dude, thanks for having us on. Really, really glad to be here.

David (00:00:17):


Tom (00:00:20):

Hey David, hey. Guys, super excited to get started. Aaron, let’s start with quick intros. We’ll start with you and then we’ll go to David.

Aaron (00:00:28):

Yeah. Cool, man. Yeah, Aaron McDonald. I’ve been in tech for 20 something years now, in the crypto and blockchain space for about six years. Been investing since my first fund in 2016. Have a pretty broad knowledge of the space now, I think over 60 or so portfolio companies across the funds and been really deep on all of the aspects in DeFi and NFTs and layer ones and layer twos and all of the shit but Altered State Machine is a project that me and David have been working on for I think, properly for almost a year now but toying around with for a couple of years. David’s my brother and we’ve worked together in the crypto space for a little while. And this one’s just come to the surface recently as something we want to put a lot more time into because it’s such a big thing.

Tom (00:01:30):

Yeah. No, I agree. David, what about yourself?

David (00:01:33):

Hey, David McDonald here. I’ve been in tech probably about six years less than Aaron, so whatever his number minus six. I started off in software engineering and got into crypto side maybe about five years ago and have been running ventures across New Zealand, Australia and Japan. And started developing the concept or the early machine learning concept back when I was in Tokyo. And when we came back to New Zealand, me and Aaron got started to get really deep into it and here we are.

Tom (00:02:15):

That’s awesome. We’re going to cover a lot on Altered State Machine today but it’ll make sense if we start with an elevator pitch. Not sure who wants to take that one but let you guys run with it.

Aaron (00:02:24):

Yeah, I’ll give it a go. And I’ll try and keep it simple, even though it’s a complex thing. Altered State Machine is two new primitives for the metaverse and the blockchain space. The first primitive is a way to prove you own an artificial intelligence agent through an NFT. What that does, is it enables connectivity of agents to processes in the blockchain and crypto space. That could be things like the tradeability of an AI. It could be something related to connecting an AI to DAO governance. It could be a way to embed AIs in protocols or use AIs as Oracles. All these different kinds of things you can do when you can connect an agent to the same ownership mechanics as an NFT. And just like good primitives, it’s actually quite a simple thing that it does but lots of use case scalability.

Aaron (00:03:30):

I think a good parallel would be something like Chainlink, where you have this very simple idea about Oracles and the Oracle network but actually, because there are lots of different types of data that you can apply to that simple primitive and then therefore lots of different applications, it becomes quite big very quickly with such a simple, basic primitive. The second primitive is the ability to inject a standard into an NFT and have that affect the way that an AI agent learns about a problem. And so this is a way for you to, say in the case of the metaverse or gaming, inject a personality into your NFT and have that personality be a way that that model is constrained and therefore, the way that it acts in a specific environment, its skills in playing a game. If you put these two things together, you’ve got this whole world of creative possibilities across gaming, DeFi and metaverse, digital humans, that you can now enable with intelligence and connect to the same protocol governance or community governance or economics that exist in other blockchain protocols.

Tom (00:04:53):

It’s a mouthful but it’s extraordinary. No, I love it. It’s a great pitch. And to zoom way out, what are the issues that you see with NFTs today? Why can’t we make Bored Ape smart? Why can’t I embed AI into my CryptoPunks? Why can’t I do it with what we have right now?

Aaron (00:05:14):

Yeah. I think the key thing that … If you look at the emergence of these crypto assets, they’re going to go through stages. And the first stage was primarily this flex. Well, it became a flex. It probably wasn’t to start off with, CryptoPunks, but it became that. And then there’s a whole generation of flexes out there, like wearing a Rolex basically but where we’re headed, this content is actually going to be much more active and dynamic. You saw something like the fluffs come along, which is actually a dynamic multimedia NFT. It’s not just a simple profile picture. And that’s where the future of content goes in the metaverse, it’s a very immersive space, right.

Aaron (00:06:00):

The problem we have today, is that a lot of these metaverse spaces. And not to dunk on anyone but they’re quite boring and dull and they’re missing some vibrance to them that you might expect in say, a AAA title game, a big massively multiplayer game or something like that. And the reason for that is that there are no NPCs, it’s all dull humans. And so you’re lacking this vibrance, “Well, okay. Let’s put in NPCs into that space.”

Aaron (00:06:33):

Well, then that breaks the model right, because the whole idea is the content is connected to digital assets and then digital assets empower the content layer inside of those worlds. What this protocol does, is allows to make that connection, bring the notion of NPCs and intelligent actors into the metaverse spaces, in the same way that you bring an avatar into those spaces. And not only that, give your avatar some superpowers. Now, it’s not just a flex on your profile pic, it can do stuff for you in that space or be a representation of stuff for you in that space.

David (00:07:14):

And you can genuinely own it.

Aaron (00:07:16):


David (00:07:18):

… which is what’s different to overlaying web two technology on top of art assets that sit within NFTs that you ultimately don’t own at the end of the day. This adds a primitive where you can actually say, “I can own and trade this asset and increase its value. And it’s mine.”

Tom (00:07:42):

And people aren’t just … I’m just trying to understand the comparison of the traditional world. And I’m not an AI expert but if I have a program or some type of AI that can do something for me, let’s say it’s in a game or let’s say it’s a chat bot or something, there’s no real way to encapsulate that in a way to trade it or to value it, right. You guys are giving that in the form of an NFT.

Aaron (00:08:05):

That’s right, yeah. The NFT is the vehicle which makes it tradable by proving that your thing is unique. The protocol essentially, is fingerprinting the genome of your agent as it grows. And that allows you to and stores that information against the NFT, so when you go to make a proof of ownership claim, you’ve got this rich history, this genome of your AI agent that you can say, “This was mine and I own it. I created it or I trained it.”

Aaron (00:08:40):

And the NFT is like a container really, to prove that uniqueness because that’s what actually is happening. Each one of these agents ends up being a unique asset in the universe. And the way that they evolve through the learning program, means that each one can be seen as an individual intelligence, non fungible intelligence.

Tom (00:09:07):

I like that. One of the best ways … Before we go into the technicals here, one of I think, the good ways to run out this discussion to make sure everyone gets it, because it’s pretty mind blowing, I guess it would make sense to describe the game you guys are launching and how Altered State plays into that. I’m not sure where you guys want to get started here but maybe what the game is and then maybe the NFTs and the AI and the … Well, the end result NPCs that you guys have at play here.

Aaron (00:09:33):

Yeah. Maybe I’ll talk to the high level about the game genre areas that we think this plays out in the metaverse, to give a bit of a broader view and then Dave can dig into the game itself. While Altered State Machine and the protocol are designed as a primitive for anyone to pick up and use, we also think it’s important to show people how it works because it is a big idea, it is a novel thing. You see a lot of copy, paste in the protocol space but this is something genuinely new people wouldn’t have played with before. What we want to do, is create some first party experiences to show people how this works and they can take that and evolve it and build their own ship. And if we look at the three areas that are interesting in that game in metaverse space, you’ve got this idea of machines versus machines. This is the genres where you can have two AIs who compete against each other in a virtual space.

Aaron (00:10:39):

In the case of our first party game, that’s AFA, the artificial intelligence football association. This is where you can create a team of AIs. You can train that team, you can assemble them and then put them in the game and somewhat challenge the marketplace and someone can accept that challenge and come and put their team in verse you. And so these are two AIs that are basically trying to beat each other at this game. Really interesting gameplay mechanic. There’s a lot of mechanism discovery there. People watching the AIs as they learn, it’s like watching a kid learn how to walk. It’s quite an engaging piece of content in its own right but not only that, it lends itself really well to this emerging play to earn space because what you are seeing in that space, is you’ve got two classes now.

Aaron (00:11:27):

You’ve got the asset owner class and you’ve got the player class, right. This whole two tier system where people might not be able to afford access but they’re renting humans to play it for them because they don’t have the time to play the game, right. We can flip that model a little bit because AIs can play autonomously. And so you can own the asset without having to rent a human to play the game. And so this new type of play to earn mechanic can emerge out of that.

Aaron (00:12:00):

The next genre I guess, in that space, would be called a companion agent. That’s where you’ve got something that might be a quest helper or it could be in decentral land, a help bot that you go and ask questions to or it could be a greeter in your virtual space and sandbox or … You’ve got all of these different things that can be companions or helpers. It could be a intelligent pet that follows your avatar around. It could be that you have 10 avatars, 10 fluffs but you can only pilot one at a time. And the other nine are your posse that follow you around the space. There’s all these different use cases for companions. And if you build out a future metaverse game space, I think more and more games that understand where this is going, will build intelligence into the NPCs. When you’re interacting with a quest giver or a helper or something like that, it knows you. It’s remembered you from last time and it’s actually adapting as the world’s adapting or the interactions are adapting and becoming a much more immersive interactive experience.

Aaron (00:13:09):

And then the third class would be what you’d call man versus machine, let’s call it terminators or zombies or some us versus them scenario. And you can imagine that playing out in a few ways. Let’s take a card style game, like Splinterlands, for example. They already have bots in the environment. Everyone knows that [inaudible 00:13:36] and there’s this cat and mouse between game developers and bot developers, which is going to get more intense because play to earn is real money. It’s not just kudos points or clout points for winning a match now, it’s real serious dollars. And I think that what we’ll see is the emergence of this underbelly of intelligence interacting with games becoming something that’s explicit because we can democratize it now. It’s no longer unfair to have a bot playing for you because everyone can have a bot using this technology. And now we know that it’s a bot because the NFT authenticates as such, then I can go in as a human and say, “I’m going to try my luck against you and see if I can beat you.”

Aaron (00:14:19):

And so you can have all of these scenarios now where intelligence can come in and play a role with some interesting mechanics. There’s a project that you are aware of that we’re also investing in, where this notion of permanent death is a big part of the gameplay. Well, we can have permanent death apply to bots now or to NPCs in the game. And so it brings this whole new style and dynamic of gameplay available to games themselves. That’s the broad brush strokes. David can go into a little bit more detail on AFA and how that works and how we’re making it simple for people to connect and use these intelligences in a user friendly way.

David (00:15:04):

Yeah, yeah. You covered a lot there. That was really good, Aaron. The thing that fascinates me about gamified machine learning is the mechanism discovery side. That’s the stuff you get with experiences like team management games and strategy games and things like that. It’s more of a overall, zoomed out picture, rather than getting in and playing the football game match yourself, which is really fascinating. Part of that journey is the trading of the NFTs themselves. Discovering what makes a a good AI, what personality makes the training have specific learning outcomes. If you need a really strong striker for your team, you can get into the weeds of going out, looking at the personality matrices. Or sorry, we call them genome matrices now, mapping that out and finding out what kind of training you need to apply to that guy to get the end result, which is a hugely rewarding process as far as gamification goes.

David (00:16:21):

And then getting into the game and assembling a team and watching them play and tweaking it as it goes along, sending them to the gym to train, is all really fascinating mechanics to me. And we want to make the entry point as easy as possible for the first game they’re releasing, AFA, to get people familiar with it in a way that’s an easy on ramp but also, if you really want to dive deep into it and start researching how the genome matrices work and how the gyms work, you can really get some real interesting mastery out of it as well, which is one of those balances that you try and create when you’re creating a game, an easy on ramp and really difficult to master. If you look at games like Celeste, which is a platform game, they’ve got a really good balance between those two things, which makes it a fun experience.

Aaron (00:17:29):

Yeah. For an end user, they can go anything from, “Hey, I’m minting these characters and putting them into the game environment and pressing a button to say train and then go pressing another button to say play.”

Aaron (00:17:43):

And it plays out in front of them. It’s very simple and they can get the key mechanisms around owning an AI, training an AI, playing with an AI, in a very simple user experience. And then we feel like that’s a good on ramp for people to get in and understand the mechanics. And then they can start to become more specialized around team composition or finding the right traits or training in the right path and those kinds of things, to get that mastery level and become an expert at it.

Tom (00:18:14):

You guys just shared a ton of amazing info. And I guess, you covered a lot of the aspects where an NPC can be used in gaming, right. It could be your friend, you can watch them compete, it could be you versus them. I guess, for context though, for maybe people that don’t have a great understanding of what AI can offer or maybe the gaming sphere, what is it like to game against an intelligent AI? And we’ve all played Grand Theft Auto. We have these dumb people on the street, you’re killing them, you’re robbing them, whatever. It’s very unfulfilling, right. What’s the difference with what Altered State Machine will offer, say games?

David (00:18:53):

Some of these AIs, you wouldn’t want to play against. I wouldn’t want to play against the AFA all stars. [crosstalk 00:19:00].

Tom (00:19:02):

It’s too good.

David (00:19:04):

Yeah. Yeah but it comes down to the game developer, how they want to implement that within their world. There’s different scenarios where you can level the playing field a little bit. You can make sure that the agents have the same average reaction time as a human does, which helps level that playing field. If you’re designing a game around humans interacting with AI agents, then you’re definitely going to want to do something like that because you can’t have light speed, fast reaction times from an AI agent against human reaction times because that’s just ultimately unfair but even at the highest competitive level, you look at the world’s best Dota players got thrashed by a bot that was a machine learning bot, which is really interesting and fun game but I think you’ve got to look at those mechanics and constraints in the background to make sure you’ve got a fair and level playing field when humans are interacting with them.

Aaron (00:20:16):

Yeah. And I think there’s probably a line that you can draw where it becomes less about how skilled you are at mashing keys, which is what makes some of the world’s best gamers, they have better reaction times than others, and more about the composition of your team and trading the right characters and understanding how scenarios might play out to optimize team composition, player composition in a world. The idea of the game itself as competing against another human isn’t about me versus the other person, it’s about my skill in assembling the right team and trading for the right team and putting that right team against the player I’m coming against. You move the level of playing from hitting the keyboard to becoming a good team manager, in that scenario.

Aaron (00:21:19):

In the scenario of companion agents and stuff like that, I think it’ll be more … You might be fighting in a multiplayer game and instead of having four characters that are teammates, you could have four characters that understand your play style and learn to play alongside you. And so they become intelligent players. And the other guy on the other side of the table might have the same thing and those players are playing along his side, his play style but they’re also playing against you but then they’re also playing against your bots. It’s this more level playing field, where you are democratizing that intelligence and therefore everyone has access to it. And so it’s not so much about human versus machine in that context, it’s about augmenting humans with machines.

Aaron (00:22:06):

And if it was in the versus scenario, you’re already experiencing that in a lot of games. There are games out there already like Gran Turismo and stuff like that, where the cars you are racing are AIS. It’s just that they’re owned by the game property and trained by them. And they tune them in a way that makes it seem like it’s fair but in this new world, you could be the owner of the AI that’s driving and you are competing against.

Aaron (00:22:34):

And so you can expose some of that stuff that’s happening behind the game, to the surface level and make it part of the game play, the trading and owning of those assets. Think about a format like a car racing game mixed with ZRun, where there’s not only economic drivers for owning and producing really great assets, there is actual gameplay happening. ZRun has been a phenomenal success but it’s just a slot machine. Every time you breathe, there’s a randomizing process. Every time you go into a race, there’s a randomizing process. And there’s a little bit of skill there in trying to acquire the right genetics and stuff like that but really, that’s just a factor in the randomization when you pull the slot machine button, whereas if you had a ASM version of that, then it would actually be a skill game because your AI is actually competing against those other AIs. And so that brings a new dynamic to that environment and provides more opportunities for people to be creative.

Tom (00:23:47):

God, the color here is incredible guys. One other question on this, I play Call of Duty sometimes with my buddies, right. I don’t know if I’d really want to play against AI, right. Where would Altered State Machine, I guess, play in there for that user?

Aaron (00:24:00):

Yeah. You wouldn’t want to play against an AI.

Tom (00:24:05):

They’d just be too good. It’s like they’re cheating.

David (00:24:08):


Tom (00:24:09):

Yeah, that’s fair.

David (00:24:11):

I would say there’s lots of opportunities in those sorts of games for more companion type and NPC type stuff that build richness into the world. I haven’t played Call of Duty [crosstalk 00:24:27].

Aaron (00:24:28):

Yeah, [inaudible 00:24:29].

Tom (00:24:29):

Yeah. All the AIs in the campaign mode I guess, would be …

Aaron (00:24:34):


David (00:24:34):


Aaron (00:24:35):

And you got to think beyond … I think the important thing to dig into and we’ve seen this with some other projects that are talking about doing stuff in this space, what they tend to do is like, “Let’s slap AI on something that exists.”

Aaron (00:24:50):

And it’s like, “Okay. Well, that’s that’s the obvious answer. AI’s going to be a bigger thing.”

Aaron (00:24:56):

More things are going to be automated but you have to look behind that and think … It’s like when you’re building a blockchain app, right. We had in the early days, a lot of people were like, “Blockchain this.”

Aaron (00:25:07):

It’s like, “Well, what does blockchain add to that thing uniquely that can’t be done better in another way?”

Aaron (00:25:14):

And it’s the same thing here. It’s like, “What does adding an AI to a game environment do to that, that’s uniquely different to what’s available in gameplay or dynamics today?”

Aaron (00:25:27):

We talked about earlier, the Game Project, they thought about what does a blockchain mean for a game? And then they designed a game mechanic around that concept as opposed to, “Here’s a game, let’s slap blockchain on it.”

Aaron (00:25:39):

And so we’ve got to do the same thing here. It’s like, “What is the unique thing that’s different about player interactions, mechanism discovery, ownership and tradeability, unpredictability of event outcomes that we can introduce now, that intelligence can play a role in those games?”

Aaron (00:25:57):

It’s looking beyond.

Aaron (00:26:00):


Tom (00:26:01):

Yeah, no. You’re not just tacking something on …

David (00:26:04):

[inaudible 00:26:04].

Aaron (00:26:04):

Yeah, yeah.

David (00:26:05):

One of those truly unique things that blockchain and NFTs bring to this, is the concept of hereditary AI. You’ve got a hereditary line of machine learning agents that have been trained and inherited traits from their parents, all the way up to the genesis AI agents. And you can see if one is performing really, really well or a hereditary tree of AIs are performing really, really well in a certain type of game environment, you can trace that all the way back up to the top level one. If you are wanting to jump into a new game and you’re looking specifically for an agent that performs in a specific way, you can go back through this hereditary tree and actually see the stats all the way to the top, about how it performs. And you can’t do that without blockchain, right.

Tom (00:27:06):

No, that’s incredible. And I’m not sure what NFTs here will have breeding or not. I don’t want to reveal … I’ll let you guys reveal what you want to here. I’m trying not to but my question on the hereditary AI thing is you’re not just breeding based on attributes, right. You’re passing down a lot of time that went in to train these AIs with a lot of data. You’re passing on real learning, not just parameters and features. I don’t know if people understand that, that’s huge.

Aaron (00:27:40):

Yeah, no. It is.

David (00:27:41):


Aaron (00:27:42):

And we think … If you look at the overall protocol structure, you’ve got the generic primitives and then underneath that sit brain specifications. Each one of those specifications is something that the community and the DAO will vote in. We will of course launch with one that we’ve come up with, which is this genome matrix, which is optimized for this gaming and metaverse space but the community, as they get to understand it or as we come roll out more use cases down on the roadmap, can start to introduce new specifications. And inside of those specifications, you can build rules like, “Can this breed?”

Aaron (00:28:23):

Because you want to be able to control the economics of each type of specification, so that the community gets maximum value out of that supply and demand curve from the assets that they own because these are assets, right. And they become quite valuable assets, like you mentioned a few moments ago, when you’ve put time and effort and data into training them, the last thing you want is for that to flood the market and reduce the value of those assets. Those mechanics around breeding and replication or cloning, are built into the spec and can be adjusted by the community over time to match that supply and demand, maximize the value of the assets, is something we’ve been thinking quite a bit about.

Tom (00:29:07):

That’s pretty cool. Before we get into the technicals of the three different parts of what you guys are building, you guys have been spending a lot of time right now talking about gaming, right but that doesn’t encapsulate everything you’re doing, right. What else can this be used for? Will I ever have an ASM AI that let’s say, auto positions my liquidity on [inaudible 00:29:31] B3? Can we get that intense here?

Aaron (00:29:33):

Yeah, I think definitely. If we build out from that metaverse gaming sphere into the other two spheres, which are DeFi and the third being the notion of digital humans. In the DeFi space, what you have now is there are a lot of bots in that space already but they exist outside of the framework of protocol governance or outside of the framework of transparency that blockchain brings to protocols. And so there’s almost these two worlds that exist. There’s the apparent world, which everyone can see. And then there is this external world, which is murky. And what we can do with the agents now, is bring them into the framework of transparency. Now, a DAO can own the liquidation bot or you can have a quant bot that is owned by a DeFi fund on chain. And people can invest in these agents and train these agents, become good at a task. And then other people can invest through the NFT and make that process of investing in distribution an on chain thing, as opposed to something that happens outside of the blockchain environment.

Aaron (00:30:53):

You’re bringing now a whole theater of capabilities to DAOs and to protocols and to users and to traders, that they didn’t have before without this primitive. The other interesting area I think for the DeFi space is Oracles. In the same way that you need this data outside of the chain to help applications perform certain tasks because we don’t have that data inside of the chain, often that data comes from machine environments but again, it’s opaque. It’s sitting behind a server in someone else’s data center and then providing that feed through. Now we can say, “Well, Oracle processes where machines collect, analyze and process data can also be owned and controlled and governed by the protocols.”

Aaron (00:31:48):

Those two things together, I think can be quite powerful for DeFi and they’ll have a different brain spec because what you’ll want to do, is introduce parameters that align to those use cases. For example, risk tolerance could be something you program into a trait on the NFT that everyone can see in the protocol and adjust in the protocol. And then that intelligence will go and learn with that trait modification, how to be less risky or more risky about how it makes a trading decision but everyone can see transparently through that NFT what’s going on and how the settings have been calibrated and those kinds of things. I think that’s a really interesting area.

Tom (00:32:31):

No, no. You’re totally right. And just to dig in a little there, when you have something like a high risk tolerance for one of these NPCs, it’s not just a 7 out of 10 rating, like on the Sims or an 8 out of 10 rating, right. The higher the risk tolerance, does that mean that this AI has trained in a sense where it’s seen a thousand iterations where it’s lost money by being too risky or how does that work?

Aaron (00:32:55):

Yeah. There’s different kinds of learning you can apply. And the protocol’s agnostic to the learning models and learning environments. It’s designed to be quite generic in that way. AFA, the soccer game for example, those agents learn by playing. They’re experiential learners. There isn’t a lot of data you need to feed into that process, the data comes from playing the game either in real time as a match or in hyper time, sped up in the gym. Whereas something like a trading bot, you want to give it data to trade, to learn how to trade and back test and all of those kinds of things that happen today. And outside of this environment, those are not new processes but now, we can connect those controls and those parameters into governance that exists in the protocol space. And so it’s not inventing a new AI or model or anything like that, that’s not the clever thing here. The clever thing is making the process of training, tweaking, testing apparent and governable within the same space as DAOs and protocols exist.

Tom (00:34:09):

Geez. Yeah, no. It’s super interesting to think about. One thing that I’d like to go over, now that we have a lot of the cool stuff out of the way and where this can go, I’d love to dig in a bit on the different … I don’t want to say areas for each NPC but you guys have a brain, you have memories and you have form, right. The brain is the base attributes. You have memories, which is basically the outputs of the learning. And then you have form, is basically what these look like, right. Hopefully we could get one of me that looks a little better and I’ll throw it on the podcast to be the video but … I’m hopeful but when you think about … When people think about NFTs, I don’t want people to think that there’s only 10,000 brains out there and they can only play the soccer game, right.

Aaron (00:34:56):

Yeah, yeah.

Tom (00:34:57):

Can you guys describe how many brains are there? Can any game or protocol create a brain? How does this all get started for anybody to use these NPCs in their game and their bot anywhere?

Aaron (00:35:11):

Yeah. Dave, do you want to go over the structure of forms and brains and memories? And then I can talk about the brain economics.

David (00:35:20):

Yeah, yeah. One good way to look at this, is if you’re familiar with the TV show in books, Altered Carbon, which obviously we’re a fan of.

Tom (00:35:32):

Oh, yeah. Great one. Yeah. I remember the references when we were investing. Yeah, I loved it. Yeah.

David (00:35:39):

You can think of the form as a sleeve. That’s just a 3D … In the case of a game, is a 3D model that you can needle cast into with your brain and interact in the world that that form is made for. With= the case of AFA football, the form is made … The AFA all stars are made to play a game in that world and they’re scaled correctly. And they’re modeled to be able to play soccer and run around in that environment. And the brains themselves are a container that contain the memories and the memories are training for specific environments. Within that container, there’ll be a memory for AFA football. There’ll also be what’s called the genome matrix, which infers the personality onto that training. And that will have stats that are mapped into the memory, like speed and size and strength and things that are relevant to the world of playing football, perception, things like that. The brain package gets needle cast down into the AFA all stars form and they can move its legs and run around in the world and play a game of football.

Aaron (00:37:14):

Yeah, so that’s the model. And the good thing is that these forms can be anything. The protocol doesn’t necessarily care about what that is. And so there’s a huge amount of creative creativity there. For example, the fluffs sitting behind me now, they will have the ability to use ASM brains, will be one of the first collections outside of the AFA all stars but for doing a different task that we’re working on. And so brains are portable between these forms. And so you might have a dozen avatars. You might have a Bored Apes for example or a punk that you want to start to do stuff with. You can inject this brain into that form. And so all of the avatars out there, now we can supercharge with this intelligence, which is a really great thing in itself because the utility is low in the space.

Aaron (00:38:16):

And if I have a collection of 200 avatars or something like that, I’m not going to even be able to use those as my profile picture over the course of the year, let alone have them be useful for something else. And providing these ASM agents means they can autonomously exist in metaverse spaces, even though I’m not playing them, which it gives this real new layer of utility across any project. And we want to encourage other projects out there to touch base with us and see how we can work together to create some awesome ASM experience for their characters,

Aaron (00:38:54):

The economics. Essentially, the way that this initial drop will work, is if you go through and get an AFA all stars toy box, which contains a team, it will also come with a brain. The first 10,000 brains will be minted by minting the toy boxes for the AFA game. You’ll have access to that. Obviously, that brain can be used for a bunch of things but probably the first things it’ll be used for are within the AFA and fluff world space because we’ve built all the underlying technology for that to work. If you want to get more brains, because obviously 10,000 brains is not enough for a team, 10,000 teams, you’re going to have to take some ASTO, which is the ASM protocol token and stake it to mine new brains. We call this intelligence mining.

Aaron (00:39:48):

And so as you stack ASTO, you can stake that and that stake will mint new brains. And that’s how we control the supply of brains in the marketplace for this type of brain specification, because it’s one that requires the ability for supply and demand to be managed to optimize the value for players who take these things and use them with their characters. Out of the gate, those brains can be used for other use cases, for other collections. There’s nothing constraining that but the constraint will be on how many ASTO are staked and we’re going to have an effect that supply can increase initially at a more rapid place and over time taper off. The DAO can control those parameters so they can get the supply and demand balance right in the marketplace for that spec.

Aaron (00:40:41):

The second layer of new brain production will come in this variations, creating or cloning and a similar mechanic will exist there where you need to stake from the protocol to charge up to clone your agent that might be more advanced. And the rate of staking will be controlled by the DAO so we don’t get one person coming and flooding the market with heaps of their brains. That’s the two mechanics for production of new brains. And the third is introduce a new spec. You might have a new use case, maybe digital humans. We need 7 billion brains out there to map to the human population. And so the mechanics around how those are created and claimed and minted can all be proposed in the spec. And then that’s built into the protocol mechanics for creating those brain specs for each of the use cases that come out. Layered approach giving control to the DAO as a principle, making sure that we get that supply demand balance right, making sure we get utility for ASTO out of the gate, all of those things affect the economics of the brain model.

Tom (00:42:00):

Geez. No, that’s incredible. Just to recap for everyone listening and let me know if I’m off here but you guys are going to have basically a pack where people can mint a pack. They get four characters, which is basically four forms, so the appearance and one brain, which are the base attributes. You basically have one player, you have their brain, their form. The other three forms, you’re going to have to mine to get a brain for each one by staking ASTO.

Aaron (00:42:26):

Yeah, yeah.

Tom (00:42:27):

Got it. Okay. That’s pretty cool. And I guess, just to zoom out though, the brains that somebody has to play AFA, the football game … Or sorry, soccer game, can that brain ever be used in another game or another … Can I feed that brain stuff to learn and fill it for something else?

Aaron (00:42:47):

Yeah, yeah.

David (00:42:49):

Yeah. The brains themselves are intended to be interoperable across different games and metaverse spaces for this brain specification. And it’s the memories that are going to be game specific. You might have a memory for AFA and a memory for a metaverse space, a memory for a conversational AI and a hundred different other memories for different tasks. And they all get loaded up into the brain.

Aaron (00:43:25):

Like that scene in The Matrix where Neo goes into the training environment and they upload Kung Fu into his head.

Tom (00:43:32):

Lawrence Fishburne, yeah.

Aaron (00:43:34):

Yeah. It’s exactly like that.

David (00:43:36):


Aaron (00:43:36):

You’re able to inject memories into the brain and then use those memories in a specific application but having said that, if you produce a player that’s really good at playing AFA and you go through this variation cloning mechanism to create a variant of your really good player, it will already know how to play at a level that’s comparable to the parent. It doesn’t have to start from scratch and that memory is cloned in the process too. Or if you create another game that has really similar mechanics to AFA, then the agent could go into that environment and know somewhat how to play it. You don’t have to start from scratch, like how as a human, we learn a task and that task … Catching a tennis ball gives us some notion of how to catch a football but we still have to learn specifically how to master the catching the football versus catching a tennis ball.

Tom (00:44:57):

No, that makes a lot of sense. The only question though I have on that is, you said that the brains are created with a base set of attributes that correspond to their use case. I can use an AFA brain for some new use case if I train it but will it be somewhat pigeonholed because of these base attributes that it’s minted with or what’s the …

Aaron (00:45:19):

Yeah. The personality matrix or the genome matrix that comes with this brain spec is randomly assigned. And just like humans, when we mint humans, they have individual DNA. The DNA of these brains are unique and they’re minted randomly. And so the idea behind that is twofold. One, it provides variability into the environment. If you came out with the same specs over and over again, you put those in a metaverse, it would become the bad side of The Matrix, where Mr. Smith is just the only character in the metaverse. We want to avoid that, so we inject a lot of different randomness into the personality creation, which produces lots of different outcomes in game style play and all that stuff.

Aaron (00:46:13):

And then the other thing, is that it encourages people to discover new brains, so keep them minting and keep them staking to mint new brains because that brain might come out with a stat that’s really important for a specific type of task or skill. And then the other layer that sits on top of that, is that each game environment can choose what parts of the matrix are important for its desired outcomes and gameplay mechanics. If you have something that’s really good at AFA and another game comes out, it might use a different part of the matrix for what it decides is important to that game. And so you are not always going to end up with these super players that are good at everything.

David (00:47:02):


Tom (00:47:03):

No, that’s pretty cool because it changes the way that they have to learn from game to game.

Aaron (00:47:07):

Yeah, yeah.

David (00:47:09):

Yeah but you might … For instance, if you make a game similar to AFA, I don’t know, a hockey game or a rugby game, you might draw the same stats from the same area as the AFA game did. We’ll make that publicly available when people create these filters for their games to go over the genome matrix. You’ll be able to see where people are picking different traits from. If everybody’s picking speed from this corner of the genome matrix and you want to add a trait for speed within your game, you might look at that and say, “Okay, I’ll get it from the same area because that’ll keep their experience constant.”

David (00:47:58):

People will know what to expect when they come into my game, that if they’ve got a fast character, it’s going to be fast in my game as well.

Tom (00:48:07):

That’s cool. Let’s take a … Just to dig into the brain spec again, let’s take a random sport like underwater hockey, just because I know Anatoly from Solana loves it and it’s always blew my mind that he plays but if I’m the game creator, I can come to the Altered State protocol and I could say, “Hey, here’s the matrix that’s important. I want 10,000 brains, let people go mint them.”

Tom (00:48:29):

Is it that simple or is it …

Aaron (00:48:32):

You could do, yeah. You would propose a new game spec and say that … Anew brain spec to the DAO and say, “We want to min 10,000 brains that look like this.”

Aaron (00:48:43):

And then the DAO would vote to accept that new specification and whatever minting mechanics that come with it. And then they can take it and use it in their environment. Or they could say, “Actually, there are a hundred thousand brains out there now already. We’re going to optimize our gameplay for ones that have this heat map.”

Aaron (00:49:04):

And so that anyone who’s got one already can start to play inside of their game.

Tom (00:49:07):

Got it.

David (00:49:07):

That’s a really cool thing for developers, is that you can go on and inspect what the distribution curve of heat maps are. You can play around with the filters and say, “Okay, if I put strength filter of this size in this area, how many of the brains that are out there are going to have a high value? And what does the distribution curve look like?”

David (00:49:33):

And you can use that to see what your distribution of characters with certain traits are before you even launch the game.

Tom (00:49:44):

That’s awesome. And what’s stopping a developer from … People are playing AFA, they’re throwing around, they’re scoring goals. I’m playing Aaron, we’re taking bets. People are loving it, right. Now, my partner comes along at Delphi and he goes, “I’m just going to go mint a brain that has 10 out of 10 attributes on everything.”

Tom (00:50:03):

How do we police against that?

Aaron (00:50:05):

Yeah. I think it’ll be hard for a start off because it’s random. It’s like minting new private keys, it’s very difficult to get the one you want with guessing private keys. And that’s part of the idea of behind randomness, is that you can’t just go and do that if you want to. Or if you try to do that, it’s going to be prohibitively expensive, like breaking proof of work. That’s one trigger.

Aaron (00:50:33):

And then the other trigger, is just like in humans, you might be born with natural intelligence. You might have high scores in all your genetic makeup but at actually, how you perform in the real world is a combination of that plus your experience, plus other factors in the environment. And so in the same way that you could have a genius be born out of the womb and not go on to be a genius in the future because of other factors, the same thing can happen with your ASM agent. You’ve got a higher chance of having a really high performing character across multiple tasks if you get that ideal genome matrix but it’s not guaranteed. There are other random factors that take place during the learning process, that means that you’ve got the ability to level the playing field even if you score high.

Tom (00:51:29):

No, that’s very important. It’s a mix of parameters but learning … Is there a way to restrict only the brains NFTs to play in the game that were part of the original minting contract?

Aaron (00:51:43):

Yeah. This is an important part about how the protocol works. Each of the brains, when the memory is saved, it’s signed by the protocol and it’s signed by the game. The game makes a choice about which agents or which NFTs it wants in its environment through that signing process. If I load up into AFA an NFT that I’ve trained on a bootleg version or something like that or I’ve stolen someone else’s brain or whatever it happens to be, AFA will look at that and say, “Actually dude, that’s a bootleg. You’re not allowed to upload that memory into this game environment.”

Tom (00:52:23):

That’s really cool.

Aaron (00:52:24):

Yeah. Game developers control that dynamic.

Tom (00:52:28):

That’s really cool. And to switch gears a bit, and this is incredible, a key part of your protocol is training, right.

Aaron (00:52:36):


Tom (00:52:36):

You can train them slow, you can train them fast. How exactly does training happen? How do I make my NFT …I’ll say NFT, I keep going back and forth, correct me if I’m wrong here but how do I make it smarter?

David (00:52:51):

We built out the platform to be agnostic to the environments that you can train in but if I take the AFA example that … Think of the hyperbolic time chain but from Dragon Ball Z. Basically, you send your little dude in there. He runs around the world in hyper time, across many copies of that world all at the same time and gets this life experience condensed into a very short period, takes those learnings, applies his genome matrix to them, his world view to them and then outputs a result at the end.

Tom (00:53:38):

Now, who’s running these training places, these farms or these gyms? How is this hosted? How do I go about as a user, training my AI? What’s that process like?

David (00:53:49):

Yeah, initially … Oh, off you go, Aaron.

Aaron (00:53:55):

No, you go. You go.

David (00:53:58):

Initially, we’re going to bootstrap with our own infrastructure for this. We want our early adopters to have a really good experience and smooth experience with it. As we go through testing decentralizing the training. We actually already have a proof of concept up and running around doing that. And it works well but we want to roll that out with [Kia, 00:54:23] because the last thing you want is people to have a bad experience. There’s a massive opportunity for decentralizing the training here. With Ethereum moving to proof of stake, we’re going to have a whole bunch of GPU compute out there not doing anything. And one of the best things to use GPU compute for besides playing games, is training machine learning models. And we want to be able to take advantage of that and give people extra utility for their sunk capital and these GPUs that are out there at the moment.

Tom (00:55:01):

That’s pretty cool. And is there a difference in learning through experience or training in a gym, which is faster? Does it have a different effect on how the NPC learns or its attributes?

Aaron (00:55:14):

Yeah. Well [crosstalk 00:55:15]. Yeah but in the gym, it’s playing against a simulated player environment. And in the real world, it’s playing against actual player environment. There is a nuance there that you might find something that performs super, super well in the simulated environment, doesn’t expect something that happens in the real environment. It’s important to train in both the real time, as well as the hyper time. Hyper time gives you a boost to get to a certain level. And you’re refining your skills in the real time.

Tom (00:55:54):

Pretty cool. Yeah, you have that randomness of the real world in your learning.

Aaron (00:55:58):

Just like training for a game of soccer, right. You might be training with your teammates or if you’re a premier league team, against the B side or whatever that’s inside of your environment but when you get on the real field with the real players, some of that goes out the window because that’s a totally new set of dynamics. Yeah, I’ve still got to kick the ball around and get it into the goal and the skills are still broadly applicable but there’s new intelligence in there that makes it different.

Tom (00:56:30):

No, I love that. And from a user perspective, do I visit a website, link my NFT, link in data, go to a gym? Sorry for the naive question but I think it’s important. How easy is it or hard is it for the training to actually happen for a user?

Aaron (00:56:45):

Yeah. For AFA, it’ll be as easy as clicking a button. Similar to a wallet experience, you’ll have your assets in the wallet. You’ll load them up into the environment and press a button to go train. Other people might implement different models. Different AI training models might require different kinds of user experience to connect to data points or to different model environments but at least for the first drop, ours is going to be really simple. We really feel there’s a big opportunity out there for people who have investments in capital that are in hardware and if any of your investors are watching and you’re running a mining operation or something like that, there is a huge opportunity here to take those assets and use them for intelligent mining and work with us to embed your infrastructure into our gym environments and decentralize this part of the process and make it broadly applicable for different kinds of use cases.

Aaron (00:57:51):

I think that in the same way that hardware and capital owners of hardware in the Filecoin ecosystem really made that ecosystem pop, this could be a new opportunity for people who have that sunk asset cost, especially with proof of stake coming on the horizon to put those assets to work doing something else that’s interesting.

Tom (00:58:13):

Yeah. No, that’s incredible. I guess, it’ll be cool for you guys. It’s a totally different education here, right. It’s just a different market. For you guys to target old … Once we go to proof of stake fully, that’s a big market to train AIs with.

Aaron (00:58:28):

Yeah. [crosstalk 00:58:30]. And if we look at lift up to another layer, like societal good, there are lots and lots of tasks which AI can help us as a human race be better at. And at the moment, I think a factor that could improve or enhance the way we go about investing in developing those capabilities, is connecting them to these same mechanics that have been wildly successful for NFTs. NFTs as a content distribution mechanism, is super charging the creative economy and it’s giving them a new way to connect to audiences to distribute but also importantly, to connect content to money in a way that’s never been possible before. Capital flowing at the speed of information is what blockchain environments are about.

Aaron (00:59:19):

If we can do the same thing for AI, where we can make it possible for people all over the world to connect capital that flows at the speed of information to economic outcomes for training AIs, then we can open up this door that’s been possibly closed for a little while, where we had … Back in the day, you could lend your computer for example, to city, to search for aliens. Can we take that concept and connect these mechanics of capital flows and DeFi, two incentives for people to search for problems, to solve big issues for humanity and take that infrastructure which was being used for mining and seen as a bit of a dirty thing, and turn it into something that’s used for good for humanity with these incentive mechanisms built into it? Because now, if I train that thing, it’s valuable. Everyone can monetize that work instantly. And that’s a new dynamic I think we can add to the AI race overall.

Tom (01:00:19):

Damn. I thought the AI side and owning it via NFT was cool but you guys are literally giving new meaning to the work expended that used to be for mining. It’s incredible.

Aaron (01:00:30):


Tom (01:00:31):

And just the two quick questions on this topic, it would appear from the outside that all this learning, all this time spent in the gym or experience builds up a large file or something that’s heavy and clunky that people have to pay for. And everyone thinks that and they say, “Oh, on chain will be so expensive.”

Tom (01:00:50):

Right. What exactly is stored on chain? And I guess, a second or related question there is, what exactly does the NFT own, right? I guess, I think that’s important for people

Aaron (01:01:02):

Yeah. Go for it, Dave.

David (01:01:06):

There’s actually very little stored on chain. There’s a hash to the file that sits behind that. And that file, if I take the AFA example, is not actually very big. The storage requirements behind that’s quite small. At the moment, they’re hovering around two or three mark, which is not onerous at all as storage file storage goes. And what is stored in there is the neural network. You don’t have to store the data of the game in that file, you just have to store how it reacts to situations in that file. And that is actually quite a small thing to store. When it gets in the game, the game loads in the world, it loads in the soccer ball and the net and the other players and whatever the obstacles that might be on the field and the agent sees those things, recognizes them and then checks through its neural network to say, “How do I react in this situation?”

David (01:02:13):

And then reacts accordingly. You don’t actually have to upload a whole bunch of data to these things to make them work. And we will be using decentralized storage for the AFA example, so IPFS. Although, we do recognize that different use cases might require different types of storage for different reasons. The platform is designed to not be prescriptive about how you store that. If you need to store it in a certain way, you’re able to store it in a certain way.

Aaron (01:02:47):

Yeah and …

Tom (01:02:50):

Yeah, that’s helpful. Oh, sorry.

Aaron (01:02:51):

And the AI basically owns the right to use that file. Sorry, the NFT owns the right to use that file. That’s what the protocol is doing. And so the right to use that file in a specific application.

Tom (01:03:07):

Got it. Okay. The NFT owns the right to use the file. And the other side of this is as the AI learns, how is that file updated cryptographically?

Aaron (01:03:18):

Yeah, same thing. As you progress and you save a memory, that calls the protocol, you pay some fees in ASTO and then it updates the hash reference to that file.

Tom (01:03:33):

Got it, that’s cool. Yeah. No, it’s weird. It’s like owning AlphaGo, imagine trading that on OpenSea.

Aaron (01:03:40):

Yeah, yeah, yeah. Totally, totally.

Tom (01:03:44):

Crazy to think about. Yeah, it’s wild. Guys, last two question areas for you. First one, just on the ASTO token, what are your thoughts for what the token underlying your protocol will do? Because I think that would be really helpful for people that are new to the project.

Aaron (01:03:59):

Yeah. Out of the box, the initial utility is governance. This is the ability for the community to make decisions about the economics of the protocol. And then as we go down the path and the community builds these things out, the obvious things that come to mind would be things like setting the fee structures for saving memories or the staking of ASTO to produce new brains, mint those NFTs. Or the ability for games like ASTO to use … Sorry, like AFA to use the ASTO token in its play to earn mechanics. If you look at the universe of play to earn, there’s probably two choices you can make as a developer. And I think because we’re early days, everyone rushes to choice one but that isn’t probably scalable. And choice one is, “I’ve created my own token and I build my own community around that token.”

Aaron (01:05:07):

Everyone thinks they’re going to be the next Axie, right but what we’ll see and what we’ve seen before in other protocols and other spheres of the space, is that there is a liquidity gravity factor in the success of your protocol or game and … Sorry, in your protocol in DeFi. And in the future, that will apply to play to earn gaming too. And so the opportunity for you to go and launch your own token and create successful gravity is going to be lower and lower and lower over time. And we think that the ASTO token in that context can act as a gravity bootstrap. If games start to build out using that as a play to earn mechanic, then you can have this meta play to earn ecosystem. And when you build your new independent game and bring it in, you’ve already got some gravity and some liquidity there to tap into, that connects to these agents.

Aaron (01:06:06):

I think that will be a very interesting one. Obviously, in the future as the protocol is used, there’ll be fees for different things. There’ll be fees for the games, all of that go back to treasury and then treasury will need to make decisions about what it does with those treasury assets over time as well, which will be another function for governance in the protocol.

Tom (01:06:28):

That’s incredible. Yeah. No, it’s cool to have it underpin your protocol in a meaningful way. That’s important.

Aaron (01:06:34):


Tom (01:06:35):

One of the last things I want to talk to you guys about, because it’s come up a lot, is you see a lot of raises. You’re two founders and you have a fantastic team, a squad, right but you’re not just … I think it’s really important to detail how you guys didn’t do an equity around, “We’re going to go away for six months, then launch something.”

Tom (01:06:53):

Right. You’ve taken a very innovative approach. You’ve launched previous projects that you’ll tie into this. You’re well known in the community. You’re continuing to iterate. I feel like that’s so important, right. It’s the number one aspect of being known in the community and having this base to sell into, right. How did you guys do that? And was that a conscious decision, I guess?

Aaron (01:07:15):

Yeah. I think it was conscious. We wanted to be able to bootstrap it to the point where it really was a … Because it’s such a big concept, right. And you go out there and you go to someone and say, “Oh, we are doing AI for NFTs.”

Aaron (01:07:35):

And that’s already two buzzwords that people are going to be like, “What the hell?”

Aaron (01:07:38):

And on blockchains. And so I think it was important to get it to the point where we had these proof points that we could go and show people. We’ve showed you these proof of concepts and stuff like that. And being more detailed about the way that mechanics work and we can get into a due diligence conversation and not look like idiots. We’ve got all the bases covered. And so that was really important for us because it was a novel thing, because it was touching some areas that people were already like bats in outer space, to do the homework to get it to a point where it was robust enough to have those challenging conversations. And to be frank, you guys were the poster child of those conversations. Delphi was right out there with the due diligence process you went through. If you’re a founder and you want your idea to be battle tested, talk to these guys because it’s a really important value add. Yeah.

Tom (01:08:29):

You guys answered all of our questions all night long, really appreciate that. And you guys are an incredible team. We’re thrilled to invest, so thank you.

Aaron (01:08:36):

And also, the second point you make is ecosystem right, because the super power of the blockchain and crypto world is this collective collaboration and everyone standing on everyone else’s shoulders. We are building these ideas on top of the things that giants who’ve come before us have already got out there and developed and it makes our process faster. And so when we look at building in this space, we’re consistently looking for that opportunity to collaborate with people we know are good builders, with projects who can add utility, with communities that can boost our own community and we can boost the value of their community. And from an outsider looking into this space, they miss this point about what makes it special. And this is the killer app for blockchain, it’s the thing that makes it successful, is this collaboration. And so we double down on that in terms of our process.

Aaron (01:09:41):

And the third element I think, is making it a DAO led project because I think it’s structurally important and also, corporations are zombies. How can you possibly can compete in the future world with a structure that’s opaque to its users, that its users aren’t economically involved in the process, the power of their collective brain power is not helping you become successful every day, versus a DAO environment where economic incentives align, community contributions are immediate, where governance is transparent to those who are involved in the process. There’s no competition for that model in the future. Corporations are zombies.

Tom (01:10:29):

Well, the crazy part about all this is you can create a business solely around one AI NPC, right.

Aaron (01:10:39):

Yeah, yeah.

Tom (01:10:39):

I can funnel all of Delphi’s reports, all of our podcasts into this thing and I could sell it. There’s real value behind all that.

Aaron (01:10:47):

Yeah, yeah. Totally. Yeah, totally.

Tom (01:10:48):

[inaudible 01:10:48].

Aaron (01:10:48):

And this micro … I was on another podcast earlier today and we were talking about the NFT space more broadly. And again, one of the generally misunderstood things about this space for content creators, is in the past, to be successful and to have a good life by being a content creator required scale. You had to have millions of followers to get enough impressions to make your advertising revenue make sense, right. Now, I can build a billion dollar community from 10,000 followers. If you look at something like Bored Apes, right.

Tom (01:11:30):


Aaron (01:11:31):

And so the idea of micro scale viable communities is realized by connecting these new content mediums to instant capital flows. And that’s never been done in history before. And so for creators out there or businesses out there, you’ve now not only got lower barriers to entry, you’ve got lower barriers to success because these micro communities can be viable.

Tom (01:11:58):

No, that’s incredible, man. I’m a huge sci-fi fan. This is like owning Ex Machina as … I don’t know, that’s physical. This is more unphysical but [inaudible 01:12:07].

Aaron (01:12:07):

No but imagine it, right. Let’s play the sci-fi angle out because I think it’s probably a good way to close this conversation. Not that I want to rush away from it but I think it’s a good point to lean into, is you’ve got an emergence of artificial intelligence becoming more important in our daily lives and that’s not slowing down. Laziness is the mother of all invention and AI’s helped us be lazier. And to a certain extent, it’s an extension of human evolution as we go down this path. Indirectly, by those things making us more knowledgeable about the world around us and automating labor so we can spend more time on thinking but also directly, with things like neural link, where this is going to augment our own physical intelligence.

Aaron (01:13:02):

If we imagine a far enough out future where AI plays a much more active role in society, either through machines or through augmented humans, I sure as hell want to be the person who owns my AI. I don’t give that to a corporation to be in charge of. Well, how can I possibly do that if I don’t own the infrastructure which controls that AI and how can you possibly do that without distributed systems? This is like the blueprint for how humanity actually evolves into this blended metaverse space, is through owning intelligence. And ASM is a good blueprint for a model in which artificial intelligence can coexist with humans in the future and the [inaudible 01:13:49].

Tom (01:13:49):

It’s super wild to think of that. Like an adversary trying to attack Altered State Machine just to hack my DNA makeup or something, that’s insane.

Aaron (01:13:57):

Yeah, yeah, yeah.

Tom (01:13:58):

Yeah. Wow, that’s pretty cool. I’m sure you guys probably get pretty scared talking about this all night. Yeah. There’s got to be some areas that scare the hell out of you.

Aaron (01:14:08):

Hey, you can either be owned by the AI or you can own the AI. Come join us and own the AI.

Tom (01:14:15):

Yeah. No, I love it. Elon Musk always makes me scared about this stuff. He’s a smart dude that is scared of it, you guys are embracing it. It’s awesome.

Aaron (01:14:24):

Well, it’s not going away. This is the thing, you can be scared of it but that’s not going to change the outcome.

Tom (01:14:29):


Aaron (01:14:30):

You’ve got to get in there and work to find solutions to help navigate that path and maybe accelerate the singularity, who knows but at the same time, it’s inevitable anyway, so let’s get on the train.

Tom (01:14:43):

Yeah. No, we’ll have fun with it in the meantime. Aaron, David, you guys are incredible founders. You’re literally taking NFTs to a level that is nowhere imaginable in the space right now. Embedding intelligence, being able to train and trade them and to be able to bootstrap that with an understandable game and let that expand the space into any use case is super cool. We’re thrilled to back you guys and really appreciate your time tonight or morning by you guys.

Aaron (01:15:12):

Yeah, yeah. Thanks, Tom. It’s been a real pleasure as always to talk to you and can’t wait to see this come out and looking forward to the next chats.

Tom (01:15:22):

Me too, man. Thank you, David too.

David (01:15:24):

Thanks, Tom.

Aaron (01:15:24):

Keep an eye out for the drop this month happening around the middle of the month. Get your first brain. Get part of the ASM community on Discord. Come check us out, ask us questions.

Tom (01:15:35):

Hell, yeah. Cool, man. Well, thanks a lot guys.

Aaron (01:15:38):

Cheers, thank you.

Show Notes: 

(00:00:00) – Introduction.

(00:00:20) – Aaron and David’s background.

(00:02:15) – Elevator pitch for Altered State Machine (ASM).

(00:05:00) – The issues with NFTs today.

(00:07:42) – Integrating Artificial Intelligence (AI) with NFTs.

(00:09:07) – Overview of AI in games.

(00:18:17) – The experience of gaming against an intelligent AI. 

(00:29:09) – Use cases for AI in DeFi. 

(00:34:19) – The ASM brain infrastructure. 

(00:42:30) – The interoperability of brains.

(00:49:45) – Preventing getting the “perfect” brain. 

(00:52:29) – How brain training works. 

(00:57:14) – The opportunity for miners with ASM.

(01:00:31) – Data management for the brains.

(01:06:35) – The process behind ASM’s launch.

(01:11:58) – The sci-fi angle on ASM.