Transcript
cyberpunkmetalhead: So, do you mean like my my metal, my favorite metal band or just ah favorite band in general?
Ally: I guess I have a punk metal headband.
cyberpunkmetalhead: Yeah, fair very good question. um I've been listening to Behemoth a lot lately.
Ally: Ahh, right back?
cyberpunkmetalhead: They're a really good like black metal band and they have a tour coming up with another very friendly name band called Rotting Christ ah next year in April.
Ally: Mm -hmm.
Ally: I don't like this one.
cyberpunkmetalhead: um But they're really good.
Ally: Awesome. Uh, I used to listen to like kind of a lot of clutch and, um, his dancing people like that. Some of those bands, some good old school.
cyberpunkmetalhead: Yeah, yeah. Yeah, theyre like the old classics, right?
Ally: Yeah. Yeah. Definitely. Well, you know, I'm a, I yeah was born in the eighties, so grew up in the bad news.
cyberpunkmetalhead: Right.
Ally: I needed a good era for that kind of music, I think as well.
cyberpunkmetalhead: Like Pantera as well. I think they were ado back in the 80s, right?
Ally: Yeah. Yeah. Skias was a big favourite for a while, yep. Lots of different...
cyberpunkmetalhead: Yeah, Venom, I think.
Stanley Bishop: It was a peak culture.
Ally: take ah you sir so did I can't see you being a metalhead.
cyberpunkmetalhead: Big culture.
Ally: Did you get any any of that?
Stanley Bishop: I love everything, um but ah i'm I'm more of a jazz a jazz guy, not to ah date myself.
Ally: Yeah?
Stanley Bishop: but um
Ally: that ah
cyberpunkmetalhead: No, that's cool. I mean, I love i love jazz myself. um There's this really cool, like dark, ambiental jazz that I like to play on like late at night. It's called Boren, b o r B -O -H -R -E -N. um It's like really like dark and ambiental. It's really nice.
Stanley Bishop: I'm gonna need a link to that sir. Genuinely though, I really have moods where I enjoy almost any kind of music and um in in in terms of like research, my my main area is like sequence processing and and music is a sequence too.
cyberpunkmetalhead: Right.
Stanley Bishop: So I actually really enjoy some of the the and ah the math of of math of ah music is really interesting.
cyberpunkmetalhead: Oh yeah, there's a lot of it in there, right? And I think that's also probably one of the things that kind of attracted me to something like the the metal genre is that it's it's so incredibly precise, right?
cyberpunkmetalhead: You're working with like 180 plus BPM. You have to make sure that, you know, if you're hitting both pedals, you got to make sure that it kind of lines up with the voice and with the drums and like with the guitar track as well.
cyberpunkmetalhead: It's very precise, right? But also I like the like the
Stanley Bishop: You know, sounds almost like running a Web3 protocol.
cyberpunkmetalhead: ah
Stanley Bishop: I'm just kidding.
cyberpunkmetalhead: I mean, Web3 protocols, they're not always known for how precise they are. Because if you're talking about oracles, we've got like gaps in there that we still need to to fill in, and we kind of need to come up with new solutions for this. There have been actually some great innovations in in oracles lately, like Chronicle. They've been doing some amazing work in into oracles and making that more like robust. Yeah.
Ally: Yeah, I'll take it.
Stanley Bishop: So what you're kind of saying is that Ali would be the Pantera of Web 3.
cyberpunkmetalhead: That's a pretty cool way of putting it. Yeah. A hundred percent.
cyberpunkmetalhead: Just just coining the genre, just just redefining the genre. Why not? Yeah.
Ally: I'm a big fan of.
cyberpunkmetalhead: First it was Satoshi, right? Black Sabbath.
Ally: I knew you heard of the band Apocalyptica that does those like concert band covers for bands.
cyberpunkmetalhead: I did. Is it the one with the cello, like the cello band?
Ally: Yes, yes, because they were classically trained and they do all metal covers. They're amazing. So yeah, maybe that's what we're doing in crypto changing changing the track up.
cyberpunkmetalhead: Yeah, they're really good.
cyberpunkmetalhead: I think that's a really good way of looking at it, just kind of changing the tempo, but just remixing the old.
Ally: with
cyberpunkmetalhead: um But genuinely excited to have you both here.
Ally: ah so
cyberpunkmetalhead: I think there's you guys are doing some amazing things. And um before we get into Lilypad itself, I kind of want to get a bit of an understanding from both of you, like where do you come from and how do you actually got into crypto?
cyberpunkmetalhead: What's your journey been like? And and how did you even come across crypto in the first place?
Ally: Yeah, absolutely. How do we meet each other is another good story, I think, as well. ah raally So my name's Allie. I'm Australian. I live in Melbourne currently, but I used to ah live in ah Sydney and I've lived all over Australia, really.
Ally: I had a bit of a widening journey to get into tech itself to begin with and then into k crypto. So I actually owned a cafe for most of my 20s.
Ally: I ran and owned a cafe on the coast here um for most of my 20s and I sold that and I went back
cyberpunkmetalhead: Oh wow.
Ally: uni a university or college depending on where you come from and I studied mechatronics engineering and computer science at university and while I was there I think you know part of it was I'd already run a business I kind of had that entrepreneurial um Vibe so to speak so when I was a university I took advantage of a lot of the groups and helped organize hackathons I came across all sorts of cool people ah in the space doing really amazing things and it just really you know the tech space was just something that
Ally: um really made a lot of sense to me and that was trying to, you know, do something to solve problems in the world. So I i went into the software engineering side of it and even ran a startup while I was at university, actually. I was ah writing my thesis on electronics engineering and a business partner and I were going through an accelerator doing a startup and I was doing three full -time subjects. I don't know how I did it.
Ally: but was kind of a crazy time but like lots of learnings from that we decided not to continue with that business because we would have my business fund I didn't want to take funding and I didn't really want to bootstrap for four years ah straight out of ah college so I went on to to court yeah
cyberpunkmetalhead: Yeah.
Ally: to corporate life. um I actually took a job at Accenture and we were building out, we were part of kind of a specialized team that was building out an AI ops actual product for Accenture and then implementing it into like the Fortune 500s, which is a lot of banks and mining companies in Australia, ah generally. um So, you know, a lot of both hands on the tools work as well as understanding the kind of user's perspective and and what they needed out of the tooling that we were providing.
Ally: And then I moved into a job doing R and &D work, more more um software engineering, but leading up software engineering for a development at the University of Sydney here that was working with our Air Force on some sensors. So working with the physics um physics department or on some really cutting edge sensor material there including they had a wet dry room sorry and everything that we we used to like look at so it was really great working with some an amazing amazing academics there that were working on kind of some cutting edge products.
Ally: And then got poached by IBM actually, so I went and worked at IBM and that's kind of where I went down the blockchain ah rabbit hole as well. This was after Hyperledger had come out, I saw about 2020. But still looking back on on IBM's journey with Hyperledger was super interesting.
Ally: And also, you know, kind of did a lot of kind of pre sales work at IBM so I was working across their range of products across. They just bought Red Hat across AI and seeing the AI history as well and implementing that.
Ally: for clients. um But one of the the big one of the interesting things for me was IBM had just bought Red Hat. um And one of their big narratives was around how you should have no vendor lock in, you know, you should be able to take your data anywhere you should be able to move your compute loads.
Ally: And Red Hat's an amazing product and team, don't get me wrong. But it felt like in this context, it was like, well, why aren't we designing that from the beginning? Because at the same time, looking at this ecosystem, this public open blockchain ecosystem, and particularly people like Filecoin and Protocol Labs,
cyberpunkmetalhead: Right.
Ally: and and seeing this same narrative but building from the ground up. So I actually went and took a job at Falcoin Protocol Labs then for two years and and that's where Lilifat actually began. So um yeah, that sorry, that was a long story. but Yeah, I started working on a few different products at Protocol Labs, so very deep down the blockchain rabbit hole by this stage, obviously. And we were working on decentralized storage and then decentralized compute on a project called Barquea, which was a peer -to -peer protocol for compute execution um headed by David Aroncik, who was one of the original founders on Kubernetes.
cyberpunkmetalhead: Mm.
Ally: original was on Kubernetes, one of the leads there. And that's actually where I met Stan. Stan was playing around in the back of our ecosystem. And we had a few chats about what was going on in the compute space. And yeah, I'll let Stan tell his story now as well. We kind of vibed. It wasn't until a while later when we launched Lilypad, which is kind of an incentivization layer on top of, well,
Ally: baka originally it was, it's kind of developed into a bit more than that. That, you know, I asked Stan to come aboard and and lead up our our research division here and haven't regretted it, so it's been amazing.
Stanley Bishop: I'll continue to do my best not to piss you off, Ally. And I have to say, ah so excited to hear the full Ally story.
Ally: ah
Stanley Bishop: I'm kind of taking notes. I hadn't heard kind of your whole arc. and And you're going to have to send me a copy of that thesis. I want to read that.
Ally: Oh no.
cyberpunkmetalhead: Same.
Ally: ah ah the Yeah, will do.
Stanley Bishop: um and Yeah, I'm a Stanley bishop in in the web through world. I sometimes go by ah science Stanley. That's my kind of um decentralized science guy character.
Stanley Bishop: um But I am a sort of scientific
cyberpunkmetalhead: love it
Stanley Bishop: technologists in real life too. I design software that runs on high performance computing clusters and specifically in the area of bioinformatics. I'm i'm really ah passionate about the application of um machine learning to biology.
Stanley Bishop: um and and yeah Most of my career over the past 10 years has been helping you know companies that design medicines build big computers. like i'm I'm just like, let's hook up as many computers as we can. That's my kind of energy when it when it comes to this stuff. um have Have I would say like felt in my my heart a deep pain?
Stanley Bishop: um working in this area seeing how decisions are made and have seen so many systems that really had the potential to change you know lives ah just just get thrown out because it it didn't have the 10 ,000X profit motive. and And at the same time, when when researchers can't get access to compute, we we have literally 20 % to 40 % of the world's compute um sitting there doing nothing. So when Allie let me know that she was working on a solution to this like kind of tragic ah misalignment of resources, I was like, let's let's do it. Let's do it. um and And then that's, I think, one of the things I'm i'm really passionate about is connecting you know the compute on the Lilypad network to you know researchers out there who need it.
cyberpunkmetalhead: I think that's a fantastic story from from both sides. I actually took some notes because I have two questions, one one question for each. So I'm going to start with Ali because you mentioned you started off, you used to run a coffee shop.
cyberpunkmetalhead: My first question is, A, did you sell that coffee shop?
Ally: Yeah.
cyberpunkmetalhead: B, do you still own it? And C, does that coffee shop accept cryptocurrency currently?
Ally: I sold it actually before I went off to rud to university, but You know, I make the joke that I did everything the wrong way around, right? You're supposed to have the startup and then you're retired to the coffee shop.
Ally: I just, I lied all wrong.
cyberpunkmetalhead: video out here.
Ally: It's a great learning experience though to have when you're your younger, although I would also recommend traveling the world folks. ah But, you know, great a great experience to have when you're younger and gain some leadership skills, I think on the ground leadership skills.
Ally: I think, you know, there's There's no other way to learn how to lead other than to do it and to sort of think about it.
cyberpunkmetalhead: 100%. I think it's one of the best ways to learn. ah you just you You acquire so much more like actual practical knowledge when you do stuff, rather than when you you know read about stuff or just or get passively exposed to it. I think there's a balance, maybe 80 % do 20 % read, you're probably going to be in a much better place than just you know be you know theoretical all the time.
Ally: hundred percent
cyberpunkmetalhead: um and And my follow -up question is that you said you studied ah software engineering. and um What exactly was kind of like your your expertise in software engineering? What was the thing that attracted you the most about software engineering?
Ally: Yeah sure sure so I actually studied so I studied mechatronics engineering and software engineering in Australia it's kind of you can do a double degree it's called so I started out in actually mechatronics engineering or just engineering actually and moved into mechatronics which is robotics and as part of that we did a software course because you need to know how to program your robots and that was actually my first foray really other than
cyberpunkmetalhead: Right.
Ally: when I was young typing out dr DIR ah slash W on our old IBM computer. But other than that, that was kind of my first real foray into software engineering. So it was quite late to software engineering, well comparatively, like people grow up knowing it these days.
Ally: um But yeah, quite like to software engineering. And I just like really loved that you could do something that would affect how things ran. And you could codify that.
Ally: And then it would show up in your screen or you would you know be able to ah code up an algorithm that did the same thing each time or um you know would sort things for you.
cyberpunkmetalhead: right
Ally: like That was kind of just black magic to me at the time. I was like, this is amazing. I feel like I'm in control of the world. And, you know, that was one of my digital feelings. And that and being really annoyed at my professor who like took marks off from me colon or something like that, I remember. ah Anyway, but but just like, you know, that kind of you can codify the world feeling I think I got when I first, you know, wrote some Hello World code and then wrote some, you know, sorting algorithms. I think we did something like that.
Ally: And then I learned more and more about it and and got into the startup and hackathon scene a fair bit. I helped organize some of those, was part of some of those societies in my local town.
Ally: um We actually ran one of the biggest hackathons in in New South Wales, one of Australia's biggest states.
cyberpunkmetalhead: Oh wow.
Ally: Yeah, that year with the with the IT Society so that was no really cool and met a lot of really great contacts that I still have actually one of the people that also works with me. It's a contact that I met at that hackathon so some really seminal moments I think and just seeing how much um Well, firstly, how much innovation was happening, how much people had ideas and wanted to try and change things and wanted to try and solve things. Like that was amazing. And the other one for me as kind of a woman was like seeing how much flexibility a career in software engineering gave me at the time as well. So I was quite attracted to to the fact that, you know, there were jobs that had flexible hours that you could work from anywhere. And here I am running a completely distributed crypto company from my, you know, lounge room. so ah in Australia so you know that was that was definitely something that attracted me as well just the the conditions especially for for someone like a woman or someone with kids I thought that was that was brilliant at the time as well.
cyberpunkmetalhead: Absolutely. Yeah. And I think I completely agree with the sentiment that it feels like magic because first time you you you do some back -end stuff. And I'm guessing because of robotics, it had to have been back -end.
cyberpunkmetalhead: It had to have been low level like C++ plus plus or something like that.
Ally: Yes.
cyberpunkmetalhead: It is, it is yeah.
Ally: My best language.
Ally: Everything's been easy since.
cyberpunkmetalhead: Exactly. like if you If you start off with C++, plus plus there's not like you you go down to assembly and then there's nowhere lower that you can go down the stack.
Ally: yes
cyberpunkmetalhead: like That's it.
Ally: Exactly, exactly. We did assembly two, electrical engineering.
cyberpunkmetalhead: really wow
Ally: We did some assembly, just I think to give an idea of how much the code affects the hardware or the hardware affects the code as well.
Stanley Bishop: Thank you.
Ally: So um yeah, it was a great background, it turns out, for running a distributed computer company.
cyberpunkmetalhead: yeah Yeah, absolutely. And I like how like you both of you kind of synergize on this and I can see the passion coming you know from from both of you regarding software engineering because Stanley, you've also you also you said you've mentioned that I meant to ask you about this, like bioinformatics. I really want to know more about that because I haven't heard that term being thrown around too often and it sounds like there's a potential goldmine in there in terms of like what do you actually do or did.
Stanley Bishop: Oh absolutely and and quite frankly I think this is a topic you're you're going to be hearing a lot about in coming years. There's a number of technologies that are lining up to really change what's possible with bioinformatics in terms of impact on our health, but but also in terms of some other areas, and and actually love the kind of topic at the you know intersection of of hardware and software, because it it was actually in that context, I i met Allie at Protocol Labs and and um learned learned about Bacalow, actually doing a project that was looking to um help the world's oceans with um genetic testing.
Stanley Bishop: It turns out that since the ocean is such a deeply interwoven food web, if you um take just a glass of ocean water and you sequence the DNA of every single species that has a DNA floating around in that glass of water, which can be thousands and thousands of different species, you you actually get a fingerprint of what's going on in the ecosystem. and And we think that that data is going to be a key part of um Let's just say not dying due to climate change.
cyberpunkmetalhead: Wow, that's mind -blowing.
Stanley Bishop: um Yeah, but there there is one thing that that we say in in this project where we're attempting to kind of like operationalize the world's ocean data, and that is there's nothing more decentralized than the ocean.
Stanley Bishop: Right. It's ah all over the place. And so really is a use case where, you you know, having technology that can run, you know, verifiable software containers on demand anywhere is is kind of critical.
cyberpunkmetalhead: Yeah, that that's super cool. And was bioinformatics what you were what you used to do before Lillipad or does bioinformatics still kind of play a role in your in your ah life, I guess?
Stanley Bishop: Yeah, bioinformatics is most of what I do. um I work on projects usually that are trying to design medicines or understand rare genetic illness.
Stanley Bishop: um But just like Ali, I'm a big open science person, love to contribute in different areas. um Interestingly, though, I started off in language processing for Google like 10 years ago.
cyberpunkmetalhead: Yeah.
Stanley Bishop: And like I was saying, though, a sequence is a sequence is a sequence to an AI, whether it's a linguistic musical or genomic.
cyberpunkmetalhead: Got it. That's fantastic. I think that's some some really unique set of skills. So I'm guessing a big part or or at least ah a substantial part of something like bioinformatics is machine learning, right? Like you're you're looking at a set of data and then you're trying to understand how this interacts yeah under various conditions. Like, you know, if you change one variable within that subset of like that cup in the ocean, then what are the implications or like plotting the future of you know, humanity or ocean ah health based on changing variables or so something like that.
cyberpunkmetalhead: Don't know if I'm making sense.
Stanley Bishop: No, that's totally correct.
cyberpunkmetalhead: i'm yeah
Stanley Bishop: No, it's totally correct. And I just have to say, I actually particularly love talking about this subject with computer scientists such as yourselves. um Because really bioinformatics is just an informatic subject.
Stanley Bishop: You know, you might call computer science computer informatics. It's about data in one form, transforming to another form and you know a dance being carried out in that process and yeah hundred percent um in bioinformatics there's you know actually multiple different types of data and they all interact with each other and so it's a very hard problem for humans to understand unaided and this is actually the year when we've had the key breakthroughs like literally in the past three to six months we have. What are called scientific foundation models that you can say speak genomes.
Stanley Bishop: And yeah, we're absolutely using those to understand what's going on in the ocean um and what's going on in a ah patient's body.
cyberpunkmetalhead: yeah
Stanley Bishop: And on some level, those are not two different things, you know, just a bunch of molecules floating around in a soup.
cyberpunkmetalhead: Absolutely. it's just It's just data points. It's just you have a ah whole bunch of data points. If you can somehow make sense of them with an algorithm, with a machine learning process or something, then you can glean some insights within that ecosystem, whether it's a human body, the ocean, you know a tree, ah you know ah a fungi farm or whatever.
Ally: but yeah
cyberpunkmetalhead: Yeah.
Ally: One of my favourite stories actually that I've learned from Stan is around, um so he introduced me to a character called Michael Levin, I don't know, have you ever heard of him?
Ally: at all, Andre. ah Michael Levin that works on is an AI scientist that works in in the biological field.
cyberpunkmetalhead: No.
Ally: And he's working on some fascinating, fascinating as an interview with Lex Friedman that you should listen to if you're interested. ah But he works
cyberpunkmetalhead: I will definitely be checking that out.
Ally: like yeah he's using both uh kind of his understanding of ah software engineering and his understanding of biology because I believe he did both um to model out or not model out even to kind of fundamentally propose yeah I mean you should talk about that as well open fold as well Stanley but the fundamentally proposed that maybe uh We're going about some of the training of human biomes. I think this is right, Stan. You can interrupt me if I get this wrong, but it was fast. Instead of giving them instructions, like a recipe list, what we really need to do is just empower them
Ally: to be an arm or be a leg and let them do it and I think we're coming around to doing the same things in a lot of AI as well or we have different understandings of of AI depending on what you want to apply it to but I think that's one of the ways we can do it and I think another really interesting thing and this that was specific to this Michael Levin character who's like a leader in this field and just super fascinating seriously recommend it um But another thing that Stan's introduced me to is OpenFold, which is this protein folding, and this requires masses and masses of data and masses and masses of compute. But this will fundamentally be able to map out what goes wrong where and us be able to really change the human trajectory. Like this is what's so exciting, being able to enable compute for for applications like that, downstream applications like that,
Ally: um That's where the real breakthroughs are happening. That's where we you know solve cancer. That's where we you know have fundamentally new ways of um you know making sure we're healthy. I don't know. It just is going to really change. AI is going to change everything, but that's one of the things like that and climate change. Those are some big subjects that I think AI can really, really have an impact on. and And hopefully, like by giving access to compute, to people, to SMEs, to subject matter experts around the world that don't necessarily have that on demand compute, we can like kind of help enable that.
Ally: That's one of the exciting things for me about building this project anyway. Sorry, I went on a bit of a rant, but I just...
Stanley Bishop: Oh, Ali, you know we we we love your rants.
Ally: Oh,
Stanley Bishop: Can I just take a second to shout out, though? Ali is wearing the coolest t -shirt in the universe. I'm still waiting on my lily pad Pac -Man t -shirt.
Ally: oh yeah.
Stanley Bishop: Lily pad goblin up all the world's GPUs.
Ally: ah
Stanley Bishop: I hope you won't mind, though, if if maybe before jumping in to talk about Open Fold, because I think this is really important. I would love to say just a word or two about Dr. Levin.
Stanley Bishop: who runs a big lab at Tufts University. and and yeah His appearance on the Lex Friedman podcast is a must -watch. If I can get you to watch that, my my job for the day will be done. um but I think at the beginning of the podcast, he he he talks about something pretty incredible. There's an organism called a planarian.
Stanley Bishop: and It's a weird little invertebrate marine creature and if you cut it in half both halves regrow into a full creature and Both halves have the memories of the original creature Somehow
cyberpunkmetalhead: Wow.
cyberpunkmetalhead: That's insane.
Ally: So which is the original?
Stanley Bishop: and and How does the data get there?
Ally: ah
Stanley Bishop: Where is the data stored in the body that when you cut it in half, the data that was in the brain is able to repopulate in a completely new brain that grows?
cyberpunkmetalhead: Exactly.
Stanley Bishop: and and This is what I think Ali was homing in on. um Dr. Levin has some theories about a certain type of inner cellular informatics. that could be almost a hidden firmware layer of our bodies. And and this is what Ali was referring to that like, there might be a signal, you know, you could give your arm to tell it to be a leg. Now that doesn't sound very desirable, but if you can kind of extrapolate what would be um possible. And Ali, I love to, it's so fun kind of sharing ideas and hearing them back and and you got it 100 % perfect, except for one small thing.
Stanley Bishop: um And this is just ah something i'm I'm sharing because it's important.
Ally: Okay.
Stanley Bishop: Dr. Levin, actually not an AI scientist. He's ah a biologist and a synthetic biologist, but he's an interesting person because he is a world leading expert who needs access to these new AI and machine learning tools and doesn't have it.
Stanley Bishop: And and so he was introduced to AI. some of my collaborators in the open source world and and were're maybe we're going to be working with him to create some ah new molecules.
Stanley Bishop: And I think it just goes to show that there is an enormous opportunity in making this hardware available. And and then just shifting to OpenFold real quick, um have you heard of Folding at Home?
cyberpunkmetalhead: No.
Stanley Bishop: um ah this This is where I'm going to have to apologize for ranting, but Folding at Home is the OG collaborative computing project.
Ally: We'll see you next time.
Stanley Bishop: It was launched from the Panday Labs at Stanford University, which is where I learned a lot of my chemical machine learning. um And it is a ah ah program that you can download on your computer, and then you you contribute your excess CPU cycles to ah folding proteins.
Stanley Bishop: um There was a ah technology discovered maybe three or four years ago called Alpha Fold that um really changed ah protein folding forever. It used to be something that needed thousands of CPUs, and now you can do it on one single GPU.
Stanley Bishop: So, folding is one of the the verticals that we're really excited about mobilizing the LilyPad network for.
cyberpunkmetalhead: Oh wow.
Stanley Bishop: um But yeah, back in the day, the only way that a researcher trying to make a medicine could kind of create the protein to understand if if it could be a good medicine was to submit it to this folding at home network. And and anyway, i'm all I'm kind of homing in on on just one of my favorite anecdotes.
Stanley Bishop: um I actually got started in ah bioinformatics as a ah middle schooler. Me and my friends lived in Palo Alto, and and every day after school, we would go take our rollerblades up Sandhill Road, and we dumpster dive for any computers that the VCs threw out. And then we had this big wall. I'm i'm thinking of that as lilypad .0000001, Ali, at this point.
Stanley Bishop: But yeah, we would just hook up every computer we could and we were just trying to get as many folding points as possible.
Ally: but love it
cyberpunkmetalhead: That's so cool. That's such a cool story. Also, it's fascinating. I did not know about Folding at Home, and it's fascinating to see that there's a network that's dedicated to that. and It feeds into, right and correct me if I'm wrong, but it feeds into, like you were saying, Ali, potentially curing cancer, but also ah figuring out why it is that at some point our sales or our our
Ally: I love it.
cyberpunkmetalhead: regrowing and reproducing of cells kind of breaks down, which is why you age, which is eventually why you die.
Ally: I love it.
cyberpunkmetalhead: So technically, you could keep running this machine learning algorithm forever and ever until it finds out exactly where does that breakdown actually happen within the DNA, within the actual cells and how to mitigate that.
Ally: Yeah.
Ally: I mean, absolutely. Hopefully in the next 10 years, so I can reverse it, that would be great.
cyberpunkmetalhead: Yeah. So, so then do you guys think that if we're thinking in terms of right blue sky thinking, let's say immortality, can we achieve immortality? Is it more likely to achieve it biologically or, uh, you know, through a perfect body that just doesn't age or, uh, through a, uh, or consciously like uploading your consciousness or, or extracting your consciousness and just having it live, you know, in a, in a machine that doesn't age by by design.
Ally: Big question. Are you you if you're in a machine? Um, I mean, I think there's already a lot of companies providing kind of AI afterlives. Um, like there's enough information and on, well, on some current day people at least, um, around to provide pretty realistic AI models of what they would do and say in certain scenarios. And I'm fairly sure there's, well, um um I've heard them. I can't think of any names off the top of my head, but there's companies already creating pretty lifelike AI
Ally: ah versions of that person. I guess you would need to, you know, ah maybe evaluate some of the arguments from The Singularity Is Near ah to decide if if that is a person or if it is not. Have you heard of that book?
cyberpunkmetalhead: ah Is it Kurtzweiler? Robert Kurtzweil?
Stanley Bishop: Oh, I mean, if i have I heard of it? I read it 100 times when I was 10 years old.
cyberpunkmetalhead: Yeah.
Stanley Bishop: a
Ally: I only fairly recently read it. So um yeah, it's fresh in mind as as well.
Stanley Bishop: it
Ally: ah Yeah.
Stanley Bishop: Incredibly prescient though, right? Like, I feel like he's just calling everything we're walking through right now, you know?
Ally: yeah
Ally: absolutely Well, he I mean, this is an updated version as well. So he has put in know the other things that we're seeing recently, but he also specifically refers to this is what I said.
cyberpunkmetalhead: Right.
Ally: Here it is, so ah so to speak. But yeah, amazing. Sorry, we've
cyberpunkmetalhead: I think I've heard. No, no, it's all good. I think I think i so. I haven't read this recent iteration that you're mentioning, ah but I heard him speak about it on a Joe Rogan podcast a few months ago.
Ally: oh
cyberpunkmetalhead: And he was saying that at the time, like we currently get, her up you know, because of ah medical advances, we get ah about four months in a year back. four out of 12. So we only age about the equivalent of eight months a year because of all the medical advances.
Ally: well
cyberpunkmetalhead: And he proposes in that book that we'll we'll actually get to the point where we'll get all of our 12 months within the next 15 years or something like that.
cyberpunkmetalhead: And beyond that point, you'll get more than 12 months.
Ally: Oh, I see.
Stanley Bishop: yeah they What do they call it? ah like Longevity escape velocity? That's the phrase, I think.
cyberpunkmetalhead: Yeah, right. Yeah, exactly.
Stanley Bishop: and It's an interesting one, and and and definitely like this synthetic um consciousness is something um I'm a little less qualified to comment on, although I'm actually doing some pretty cool work on brain data.
Stanley Bishop: um
cyberpunkmetalhead: Mm hmm.
Stanley Bishop: I do have to say though, and and um for forgive me for for sharing something a little personal, but a a little earlier this year, maybe my my closest friend passed away.
cyberpunkmetalhead: Sorry to hear that.
Stanley Bishop: um He was one of one of my main collaborators in rare genetic illness and it has been um such a loss. There's just so many moments where I miss him, but then so many moments where I'm working on something and I i want to talk to him. And I actually did, um I sort of specialized in post training. So i I took every conversation we've had over the past 10 years. It added up to like thousands of pages.
Stanley Bishop: and I trained a model and i've been talking to it every day and it's been helping me with problems and and i have to say it does feel like. Talking to him. you know and and I don't know how to really understand it much more than that at this point, but it's, man, a whole can of worms that we're opening.
Ally: too much.
Stanley Bishop: On the longevity front too, the biological longevity front, um with with things like curing cancer or curing aging, I'm like very careful with with saying too much or expressing too much.
cyberpunkmetalhead: Yeah.
Stanley Bishop: but like Man, and the pieces are coming together. The pieces are really coming together. And aging isn't something that has to happen. Cancer isn't something that has to happen. ah Whales, for example, don't get cancer. There's zero cancer in whales. And it it just happens to be that there's something in their genome that's set a little differently. So I really do think these problems are solvable. And it's a matter of getting the data and getting the tools I'm really excited actually right now to be working on a pilot with a group called the Human Pan Genome Consortium and their international consortium of genetics departments that um kind of established the best practices and standards around genetic technology. And they're right now working on something pretty interesting that you you might actually enjoy reading about if if longevity is interesting to you. It's called the T2T genome.
Stanley Bishop: or the telomere to telomere genome.
cyberpunkmetalhead: Mm.
Stanley Bishop: um The telomere is like the end cap of the genome or you might also think of it as like the the header of the genome.
cyberpunkmetalhead: Right.
Stanley Bishop: It kind of expresses how the the sort of genome should be should be read and it's believed to be one of the main ah parts of our genetic data that accumulates defects leading to to aging. and so you know First step is is really seeing how the telomere changes as we age and you know other epigenetically determined parts of the genome, and and then yeah seeing how do we repair that damage, how do we prevent that damage, I think it will be possible. um But right now, the telomere to telomere, exactly, exactly, exactly, Ali, that's what it is. And that's what's so funny too is because when we you know had the human genome project, we thought that was it, that you know that we had the source code of biology. But yeah, it actually turns out there's many layers of data and they all kind of interact with each other. um Anyway, big punchline, um sequencing one patient T to T right now costs half a million dollars.
Ally: Wow.
Stanley Bishop: and There's only been fifty patients sequenced. um ah really Fortunately, actually three of the patients I work with have been have been sequenced. um and it was actually um ah This is kind of cool, Allie.
Stanley Bishop: It was the University of Western ah Australia that donated the sequencing. And it turns out that they actually have a really big mission in making sure indigenous people are represented in um our genetic data.
Ally: but so
Stanley Bishop: and And then they they view rare disease patients as like brothers and sisters in that that struggle for representation.
cyberpunkmetalhead: Right.
Stanley Bishop: So anyway, we just got to get that half a million down to 50 bucks, though, and I think we'll be ready to roll.
cyberpunkmetalhead: So when you say sequencing, you mean that their DNA the their dna data was mapped out into this ah machine learning system?
Stanley Bishop: um Yeah, their their DNA was sequenced. um Out of a sequencer, you get all these little fragments and then you put the fragments together. or You think about it as like putting together a puzzle. um Most genomes are like 98 % accurate, like if you go to your doctor and get a genome.
cyberpunkmetalhead: Right.
Stanley Bishop: But it turns out that 2 % that you miss is actually where some of the most interesting stuff happens. so
cyberpunkmetalhead: Oh, wow. I did not know that. um I did get a 23andMe test recently. And and it was it was quite interesting. but But I recently just read about 23andMe and how their stock price just kept just get plummeting for the last few, ever since they gone public. um And I'm not sure if there's any kind of big scandal there. I wouldn't be surprised if there were, though.
Stanley Bishop: I mean, I think we saw their entire board resign last week.
Ally: Wow.
Stanley Bishop: And so, yeah, there's going to be some majorly bad stories coming out of that one.
cyberpunkmetalhead: Oh fuck, yes!
cyberpunkmetalhead: Oh, I know you're absolutely right, I forgot about that, yes! Oh, jeez!
Stanley Bishop: And worth mentioning, too, that the only other company with a similar dataset, which is called ah Decode Genetics, was bought for like $8 billion dollars by Amgen. So I would say 23andMe, an all -time epic F -up.
cyberpunkmetalhead: yeah Yeah, honestly, to have that much have access to that much information, and and just it's it's a shame really, if anything.
Stanley Bishop: But hey, doesn't it make you wish we had good cryptographic protocols for data sharing?
Ally: Surely they'll - And the ability to -
cyberpunkmetalhead: i feel I feel like there's ah there's ah there's a punch in there somewhere. there's you're you're you're You're trying to say something.
Stanley Bishop: Ugh, Ali, this guy's too smart for me.
Ally: No, he is. I mean, it's a point though, right? It is a really big point. One of the reasons that I refuse to do that eyeball thing and haven't like gone and got DNA testing is because where's my data go?
Ally: I'm not sure if I'm getting asked permission for my data being included actually.
Stanley Bishop: Oh.
Ally: It's really interesting because
cyberpunkmetalhead: Hmm.
Ally: My family is actually fragile ex -carriers which means we can have slower children. It seems to be stable in our line and it's actually really interesting because we've had enough of our family tested that now they use our line for um comparison to other people because our line is so stable it appears like the gene doesn't mutate it just continues stably which is really interesting.
cyberpunkmetalhead: Right.
Ally: Um, so yeah, I think in, in, in terms of data, we need more of it, but also I think it needs to be permissioned at the same, same by the same token. Right. Uh, I agreed to do ask me for permission to use it.
cyberpunkmetalhead: Mmm.
Ally: And I said, yes, but you know, I think there's a definitely a lot of companies out there that are are really just making money off your data rather than doing, doing the right thing by it. Like we want them to be.
cyberpunkmetalhead: 100%, yeah. So having gone through the 23ME process, that the questions that I was asked was, are you happy to um have your anonymized data be used for cancer research? I go like, yeah, sure, go ahead. um Are you happy for us to store your your genome data and and details in case someone with a similar DNA is looking for like you know people from their like family? And I was like, you know what? No.
Ally: We know enough about it, but that's so that's great. so
cyberpunkmetalhead: Yeah, exactly. um But yeah, obviously that that just brings us to um how do you take how you take all that power and all that you know all that learning and and use a system that's provably fair, but which is obviously a lot of what you guys have been working on. So maybe I think having a ah high level overview over how LilyPads solve some of these issues, I think that would be super interesting.
Ally: ah Yeah, absolutely. So I think um you know at its core, ah we're not going to say solve all of those those issues, all in one protocol. um But what we can enable is on -demand compute for for the SMEs in the field that know you know what they want to build.
Ally: and that know, you know, where the problems are really, we definitely don't want kind of centralized versions of all of our AI models because that just leads to biases being inherently spread until we're basically just one big ah culture. We need diversity to thrive and I think this is something that an on -demand distributed compute network can help provide by providing that kind of democratic access to the resources you need to build these kind of models to
Ally: to you know solve these kind of problems that we've been speaking about for the last 20 minutes now, I think, or really hashing out. So at its core, we're a distributed compute protocol, but we're really hoping to build out from there, you know ah add storage, add those protocols, building out great data layers, great privacy layers, and become you know part of a decentralized AI ah system in the world as well.
cyberpunkmetalhead: Right. So I think it's really interesting. You mentioned distributed, and not decentralized. ah is that Is that distinction on purpose? the Decentralized could be um you know anyone can own this token, but one person may own up to 90%. Distributed means everyone on the network owns the same amount.
Ally: Yeah, I think it's an interesting distinction. So how I distinguish it is from a tech view rather than from an own ownership view necessarily because we're but decentralized. So the nodes aren't don't belong to us. We're just building out the protocol. We we take a fee per job because we need to continue to build this system.
Ally: But otherwise, like the nodes don't belong to us, the models don't belong to us, the data doesn't belong to us. So in in quite a lot of ways, we're completely decentralized and actually building, we're just helping to build trust in a distributed network.
cyberpunkmetalhead: Right.
Ally: So actually, to kind of use them a bit interchangeably, they're interesting to know, you're correct.
cyberpunkmetalhead: Yeah, cool.
cyberpunkmetalhead: Awesome.
Ally: ah yeah In terms of the token, we we will be releasing a token. We're not releasing that until early next year alongside our mainnet, it will be.
cyberpunkmetalhead: So what what network is it on? i'm I'm sure you've mentioned it at the beginning. There's something starting with B, but i've kind of I think I've forgotten the name of it.
Ally: i No, but so we're actually, so compute doesn't happen on chain. It happens off chain, obviously, but just just in case anyone was wondering.
cyberpunkmetalhead: Right.
Ally: And then we provide these on -chain guarantee guarantees. Now we're completely EVM compatible, so we'll be launching a multi -chain. We want to be where users are, right?
Ally: We want them to be able to access it in whichever application they decide to. That includes across the different Web3 chains and that then also includes the Web2 world, so providing those APIs, those SDKs and a SAS layer even for our folks in um the Web2 or traditional tech world ah to be able to access this without needing to know about tokens, without needing to know about cryptography without needing to know how anything under the hood works and just needing to know that they can have an on that demand, um you know, GPU available to run their jobs on the network. Sorry, I think I like completely went away from your question.
cyberpunkmetalhead: na No, I think it's cool. I think i have i have a like a lot of questions around the way like your the the nodes, the the way they work, the way you choose, kind of um how do you decide like what kind of job to give a certain node?
Ally: So we're carrying
cyberpunkmetalhead: Is it based is it based on their specs? Do you run different models or do different algorithms based on the kind of ah you know like parts that a node has? Or do we have specifications ah in terms of you know what kind of compute CPU, GPU nodes should have to be accepted onto the network?
Ally: Yeah, absolutely. So currently we're in an incentivized testnet phase. So like most most marketplaces, you need to bootstrap both sides of the network until you get to ah to ah an inflection point. This is true of pretty much any any marketplace. ah So at the moment we're incentivizing or you know providing rewards for GPUs to join our network before there is a lot of demand or a lot of paying demand on the network.
Ally: At the moment, we're only encouraging the video GPUs because of the way that we're providing proofs. But we certainly have a plan to make sure that we are GPU neutral and even CPUs like Stan shared before the the folding project that is just using spare CPU power. We can certainly ah provide jobs that are CPU or that run on an AMD or that run on Intel as well in future.
Ally: um So currently, if a job comes into the network, what happens is it's offered up as a a job like it's broadcast to the network, hi, hey, a user here wants to run this protein folding job, let's say, um and then the network goes, oh, like,
Ally: um it sends a broadcast out and multiple GPUs go, oh, I can do that job. I am a 40 -90, so I meet the requirements for this this job. um And also, I have a great reputation, so there will be a reputation system.
Ally: ah And you know also, I'm close to where you are or have low latency so for those regions, for example.
cyberpunkmetalhead: Right.
Ally: At the moment the matching is just by the requirements of the job, so needing say a $40 .90 to run or needing ah you know whatever it is to run. ah In future though they'll be more sophisticated matching that occurs. um Then the job is launched and the GPE providers can can name their price, so the marketplace.
Ally: um And then the job is run. And then the big problem in a decentralized network like this is how do you know that this node, this GPU that you don't know anything about and could be anywhere in the world has run the job that you wanted it to run. ah Now it's easy in a deterministic, well, easier in a deterministic network because you can rerun the job and you can check it.
Ally: but in non -determinism, it's a little bit more difficult. So we're using kind of a blended verification method and it'll be a multi -verification method as well. It will definitely depend on what type of job. If we're looking at AI inference, ah we're kind of using a blended method that checks how much GPU was run and does kind of some sampling of the job along the way alongside a reputation for this node to verify that that job was run.
Ally: um and
cyberpunkmetalhead: Right.
Ally: and So the person that proposed the job in the first place can go, I don't think you did the right job, and then a mediator, so a separate node to the the um hardware provider, the GPU that ran the job, can step in and do some verification on whether the job ran or not.
Ally: um that also will happen in a random sampling of different times so yeah so that was a that was a long the long discussion but one of the big problems in these kind of networks is is providing trust um so that's what starts being complicated in some of these uh decentralized marketplace situations
cyberpunkmetalhead: Of course, yeah. And thank you for the explanation. I think that was actually really, it was really fascinating to hear you speak about it. um The only one question that I have is you've mentioned ah the a job gets matched with a node.
cyberpunkmetalhead: um Is it a plan to always be like one -to -one or or do you guys have plans to fragment the job into clusters so that you can have a pool of nodes running the same job and then just giving you the result back?
Ally: Mm
Ally: Really good question. ah So currently it's one -to -one, so embarrassingly parallel jobs, but actually due to the executor, which I spoke about earlier back there. So we use back allow as our compute executor under the hood still.
cyberpunkmetalhead: Mm hmm.
Ally: quote It is cluster enabled. That means, for example, on board at a data center, that job could get sent to this node that says it has a hundred A100s.
cyberpunkmetalhead: Right.
Ally: and you could run a machine learning job for many days in this data center on demand just about, you know, provided do you have the right resources on the network, ah then that's absolutely a possibility to spit up clusters.
Stanley Bishop: you know This is a ah use case we're particularly excited about because in um the enterprise space, there isn't an enormous demand for um you know large language models that are conscious of a company's data and records.
Stanley Bishop: And there's um in a lot of areas either an inability or a large hesitancy to send all that data to open ai so um lots of people need fine -tuned custom open source models like the llama four or five billion model that you know are trained on.
Stanley Bishop: proprietary data, but but remain in total control of the you know the entity. and and so It's actually one one of these um offerings that that we think is going to be extremely high impact, you know allowing um companies to you know have access to these models without you know giving away all of their data.
cyberpunkmetalhead: one hundred percent
Ally: Yeah. but Well, I mean, I put in my deck to actually open AI, so I haven't made the joke before.
cyberpunkmetalhead: 100%. I feel like there's way too much reliance right now on open AI and chat GPT. um and like Don't even get me started on on how it used it was founded as an open source AI for everyone kind of solution and it's a closed source for profit, even hinting at spending two thousand ah charging $2 ,000 a month on the enterprise or one plan. um I think it's a weird turn.
Ally: ah No, I totally agree. So we shouldn't start down that hole, though. We'll be here for another hour.
cyberpunkmetalhead: Yeah, a hundred percent. um So when it comes to pricing, then how how competitive is is, you know, Lilypad against something like using um AWS's recent AI um ah farms?
Ally: ah Yeah, absolutely. So I think where our competitiveness comes in is because we're a peer -to -peer marketplace, there really isn't a middleman. So that means you can, other than the small fee we take, um there's the no reason to think this wouldn't be up to 20 times cheaper than going through something like AWS, which is a lot of layers of different things they need to support.
cyberpunkmetalhead: Oh wow.
Ally: So the build out of the protocol, the like data centers, the people running the data centers, the people like working on their next product or stealing it from open source. Did I say that out loud? Or um you know, the the list kind of um goes on. ah So there's no reason to think that. We ah don't have a benchmark on pricing, but it will definitely be cheaper than running on AWS.
cyberpunkmetalhead: Nice, that's great to hear. Stanley, I think you want to talk about FHE and I'm really looking forward to hear you talk about it because i I've recently got acquainted with FHE and I think it's fantastic technology, but it's not without issues.
Stanley Bishop: Oh my gosh, I'm so excited um to talk about this, but can you guys hear me okay? I'm actually having some some lag on this the line.
cyberpunkmetalhead: All good, my end.
Ally: You're all good.
Stanley Bishop: Okay, well ah stop me if it's it's if it's at all laggy. I'm having a little trouble hearing you guys. um
Ally: Yeah, I can just jump in there for a second while Stan refreshes a little bit as well. So we've actually been talking to him. We have an advisor who was working with XAMR, which is the French Hemomorphic Encryption Company.
Stanley Bishop: Hey, hello?
cyberpunkmetalhead: Ah!
cyberpunkmetalhead: I know Zama, I spoke to them.
Ally: and i believe yeah Yeah, I listened to that podcast, actually. That was really well, really good. So we're very excited for fully homomorphic encryption and very open to adding it as a privacy measure in our protocol.
cyberpunkmetalhead: Thank you.
Ally: At the moment, my understanding is it's quite difficult to fully encrypt a full model, but you could encrypt, all that you could FHA your data at the moment and and send that in and out.
Ally: so Full privacy is probably too slow at the moment, but they're getting there and I'm very excited to see and follow along with where that goes. And we're ready to put that in as soon as it's ready to go, basically.
Ally: So our protocol is quite modular in that respect. And I'll let now that Stan's back, I'm just talking about Zavir, please, Stan.
Stanley Bishop: Again. um so sorry So sorry too, guys. I don't know why the video is lagging out, but hopefully you can hear me. And and I want to say about FHE, I believe um we we will look back on this moment as part of the pre -FHE era, that when FHE reaches the level that Ali is alluding to, where we can actually conduct all of our data and machine learning operations through encryption,
Stanley Bishop: it's It's going to change the world. like it's It's going to be something where you know it isn't just there's some cool tricks that we can do on Lilypad. It's like every mainframe in the whole country will be FHE by default.
Ally: more the internet as well.
Stanley Bishop: I can talk a little bit more about why I think that, but then also I feel like maybe you should talk about like what FHE is or what it allows us to do. and I remember that box you checked on the 23andMe, right?
Stanley Bishop: like Is it okay for my data to be used? in cancer research, and I um do a lot of data science for oncology. like I've been active at um ah Stanford Medicine for a number of years working on patient cases, and um I get yelled at by our oncologists a lot. um A lot of times, like we need to find certain data, or we're trying to get certain data from a biobank, and we get the data, and there's something wrong. like They mislabeled it, or or it's sort of like the data is contaminated.
Stanley Bishop: And it's a really challenging process, like getting the data, navigating health and privacy issues, as Ali alludes to. And so I would like genuinely say that for every day of useful work um contributing to the development of a particular precision therapeutic for a cancer patient, we waste seven days chasing or cleaning ah data. And it's just a pain. And foundation models have this potential to totally change that. um If I train a foundation model on, say, um every breast cancer case that has been sequenced, I can generate data from that model. I can generate synthetic data that can be used in the design of a therapeutic without ever seeing the data of the underlying patient.
Stanley Bishop: um So, we we have potentially a way of making you know all of the benefits of data availability happen with none of the downsides.
cyberpunkmetalhead: Right.
Stanley Bishop: But you do have to pool all that data for the training, right? And and that is an incredibly difficult cat herding exercise. But with FHE, these groups could put their data together pooled for training with trust.
Stanley Bishop: and you know train the model through FHE, delete the underlying data after, and then you have the model that can be you know used by clinicians. And you know again, that's just one example of what I think FHE is going to allow.
cyberpunkmetalhead: Yeah. Well, I think it will also ah youll get people to be more comfortable with sharing ah sensitive information. But when I say sharing, allowing sensitive information to to have computation run on it on it. And that will in turn just create a richer understanding of of you know anything really. like you're not You're no longer feeling um uncomfortable having someone perform computation on that piece of data because it never it it is never unencrypted at any point, right?
Ally: That's
cyberpunkmetalhead: That's that's the magic of it. Though I do understand that one limitation, and I'm not sure if this is um this is bearing any any weight on Lilypad at the moment on your guy's solution, is that it's a lot slower than plain text computation by 1 ,000 times or up to 1 ,000 times slower than plain text.
Ally: fine.
cyberpunkmetalhead: I know that people at Zamaa have been making ah great progress in in getting around that. It is still significantly slower, ah though I do expect and I think everyone expects that it could get a lot faster as as they have new breakthroughs in that technology.
Ally: Absolutely, and I think one of our theses as well, or one of the things, one of the trends you can see happening in AI as well is that, ah you know, models are getting more and more optimized for particular tasks, and so they're getting smaller while still being more powerful. So I think, ah you know, there's a possible meat in the middle there for some essential models, not necessarily, you know, your your foundational model for and oncology there, Stan, but ah for many popular models,
Ally: um They're getting optimised at the same time as FHE is getting faster so I'm sure there'll be kind of an inflection point there where they they kind of meet and it makes a lot of sense. as well. But not everything needs to be private either I would suggest so you know definitely looking forward to having FHE be rolled out and I think even when it's slow um I think that's still valuable like if it takes say what three months to train up with FHE instead of you know a few days or something
Ally: Is that better than not having the data at all? I i would say yes. So maybe it's still a breakthrough even if it's slow and maybe that doesn't matter. Depends on the use case.
cyberpunkmetalhead: Yeah? No, 100%. I'm really looking forward to just seeing more of that across the web and and compute.
Ally: absolutely
cyberpunkmetalhead: I think it's going to be magical.
Ally: yeah I think our fundamental thesis at Lillipad is to make sure that we have a underlying ah marketplace in play ready to go on demand um when you want to use it and then these are layers that we add on top so FHE will be almost like a plugin for Lillipad that you can choose to use or that's That's privacy. You might also want more verification on the compute that you're using. So you might put in ZK into your job or or things like that. So these are definitely modular value layers that can plug into to the Lilypad compute network.
cyberpunkmetalhead: That's super cool. So one big question. um When is Lilypad going to be deployed in the mainnet?
Ally: Yes, so we're looking at a release date in probably late Q1, early Q2 next year. So we're currently in incentivized testnet phase though. So if you want to run a GPU that's currently already being incentivized and we're aiming to release our next phase of incentivized testnet, which is folks building models and and modules on Lilypad for other folks to use.
Ally: um That'll be coming in the next month. And then, yeah, there'll be lots of other community activities so you can help educate or you can, you know, help contribute to the network in many other ways as well if it's something that's of interest to you. We'd love to hear from people that provide storage or data or people that think that, you know, they could use a network like this as as a part of an overall decentralized AI strategy too. We're always looking to hear from cool projects that would want to use compute or would want to plug into creating a decentralized AI ecosystem.
cyberpunkmetalhead: So it's it's not exclusively for decentralized AI though, is it? You could also run any sort of computation. Could you use it to run a ah web application or a sorting algorithm or anything like that?
Ally: Absolutely. That will definitely be possible and already is possible actually. So yes, you can already do that. One of our fun little hello worlds is actually, you know, cows say the little, you give it a prompt and it comes up with a cow saying that prompt in ASCII art.
Ally: It's actually our hello.
Stanley Bishop: Yeah.
Ally: It's actually a little hello world and it runs on the CPU. So yes, absolutely. You can run other jobs on the network.
Stanley Bishop: You know, Ali, I hope you won't mind if I if i jump into plug um one one of my favorite Lilypad subnetworks under development.
Ally: i
Stanley Bishop: And that would be, yeah, and in a non -machine learning vertical, in the quantum computing vertical.
Ally: Oh, of course, yeah.
Stanley Bishop: um It turns out we have way more quantum ah computing researchers out there than we do have then we do quantum computers for for them to work with.
cyberpunkmetalhead: Oh.
Stanley Bishop: But um many steps of the quantum algorithm design process can actually be simulated on GPU. So we're we're actually hoping that one of the big applications directions of of LilyPad will be you know where a quantum researcher goes to test their their algorithm first time.
cyberpunkmetalhead: That is so cool. Yeah, I think i think quantum computing is is not as explored as it should be at the current time.
Stanley Bishop: So.
cyberpunkmetalhead: um Also, one cool thing about FHEs that is quantum proof, by the way, so there is no way to to once you know quantum computing is going to be everywhere, FHE will remain encrypted and will remain safe, which is a nice added bonus for that.
Ally: Just have to make sure we sell out the Bitcoin, right? No, just...
cyberpunkmetalhead: Yeah, exactly.
Ally: No,
cyberpunkmetalhead: just Just jump ship as soon as you can. ah
Ally: I'm kidding. I'm sure they'll be enough.
cyberpunkmetalhead: you know I was actually thinking about that. It's surprising that you know that it hasn't happened yet. like There are quantum computers out there on the market. Is it maybe just a matter of like they're way too slow to break ah and as like the encryption algorithm?
Stanley Bishop: Yeah, well, I think it's a demand issue. there There are certainly a number of quantum computers, but there are many researchers. And I believe that for non -government researchers, there's often queues of two to four months to get ah time on on a quantum system. um and And yeah, like i yeah ah many many people I work with and in this space whose opinions I trust describe this as ah and a national security issue.
Stanley Bishop: that um you know we really need to be making sure that researchers are blocked and and that they're able to kind of um move forward and and yeah i think it's just such a cool high impact opportunity for lily pad.
cyberpunkmetalhead: Right.
cyberpunkmetalhead: Do you guys have, ah are you working with anyone that's running quantum computers at all? or Or is there anyone on your network that's running them?
Stanley Bishop: Yeah, absolutely. in In fact, I think this is um maybe maybe a little too early to fully let the count out of the bag, but we we have a pretty interesting partnership coming together with some very um large hardware companies and and an incubator called Q -Force.
Stanley Bishop: that actually is a really beautiful community of quantum researchers who are kind of like working together to you know support each other's products and research. And you know we're hoping to maybe up ah place some ah quantum compute simulation grants in that ecosystem of researchers.
cyberpunkmetalhead: Wow, that's super cool. I love the idea of quantum computing as much as it eludes me. like I think about qubits and superposition and I'm like, how is that possible? And the only way to understand it as a programmer, is as a software engineer is to get your hand onto one of those quantum computers and just to write something and see how it compiles, which I'm yet to do so.
Ally: file.
Stanley Bishop: And you are right, except it's also possible to do some of that on GPUs. And you can use a framework that NVIDIA makes called QQuantum, C -U Quantum, which stands for CUDA Quantum.
Stanley Bishop: And yeah, it's amazing.
cyberpunkmetalhead: Okay.
Stanley Bishop: like You can actually start to develop some intuition about these these elements and how they they interact informatically, although we must Always remember the words of Richard Feynman, the true quantum field theory master, who said, if you think you understand quantum, you don't understand quantum.
Stanley Bishop: Because his his illusion is that it isn't understandable with our our silly little monkey brains.
cyberpunkmetalhead: Right.
Ally: i
cyberpunkmetalhead: yeah Yeah, Jamie, pull up some DMT. that that
Stanley Bishop: it's It's so funny though. Actually, it's it's um my original training was in that quantum informatics. and it was actually like um I came to bioinformatics when I developed a rare disease. I became very passionate about ah but you know disease and care and stuff. But but hey, that's the that's the fun part is it's all just like bits flipping, right? It's just information dancing.
cyberpunkmetalhead: Yeah, exactly. It's also super fascinating. And and i have I have so many things to to research after this podcast. It's probably is much the the most i've I've learned out of any episode I've done. And I absolutely love this.
Ally: I love it, love it. That's amazing.
Stanley Bishop: You know, I have to say too, you'll have to have us back on. And I think this is like really reflective of the lily pad culture. And I hope you won't mind if I take a second to plug the lily pad community and encourage anyone who likes this kind of stuff to come ah check out the discord. I really think that, you know, we've built a community that's driven by curiosity and exploration and mutual support, um really in a way that mirrors the intention of the network.
cyberpunkmetalhead: and And thank you so much guys for coming on. I think, um like I said, I think you guys are doing some amazing work. I would love to have you both back in the future. ah You've said you're you're planning the launch in around Q1, Q2 next year. I think it would be super interesting to to hear from you guys how the launch went and and and what your plans are once once you're up on the mainnet. So you're always welcome to come back here. And I would love to talk more ah bioinformatics, decentralized computing with you guys. It's been an absolute pleasure.
Ally: Absolutely. We'd love to come back and thank you for hosting us, Andre. That's an amazing conversation.
cyberpunkmetalhead: Oh, in closing, do you have any socials that you wanted to shout? ah So I don't forget.
Ally: ah You can find Lilypad across Twitter, YouTube, LinkedIn under Lilypad underscore tech. Most of our socials are on and our Discord is also at discord dot .lilypad .tech, which should give you a quick link to our Discord invite if you'd like. We mostly hang out in there, including Stan. So ah we're pretty public about what we're doing as much as possible. And we're, you know, pretty active in the Discord. So you you'll find us there if you you want to talk to us one to one.
cyberpunkmetalhead: Amazing. Thank you so much, guys. um And Stan, have a lovely rest of your day. I hope caffeine caffeine kicked in. And Ali, have a good night.
Ally: like you andre of what i get some
cyberpunkmetalhead: but Yeah.
Ally: ah great talk thank
cyberpunkmetalhead: Lovely talking to you both. Nice meeting you.
Stanley Bishop: Oh, this was really a treat, man.
Ally: okay
Stanley Bishop: Thank you so much for having us and Ali always like really love when we get to chat about this stuff. It's the best.
Ally: doing me
cyberpunkmetalhead: Fantastic energy, guys. Thank you again.
Ally: like
