Gavin Wood：我非常清楚自己的目标，并尽量保持真实，脚踏实地！

Thank you for having me.

Well, it was many years ago now, so it's not always so easy to remember. And maybe my memory is playing tricks on me, but the one thing that comes to mind is that it's more peaceful now. People seem more relaxed. I would say maybe in the past it was more chaotic. Now it feels less chaotic. Still busy but somehow more peaceful.

Indeed. Yeah.

Well, I mean, you know, as someone from England, tea is very much part of the culture. But I think I started drinking tea properly maybe around 12 years ago. I found that coffee, which I had drunk until then, was a little bit too strong. The effects were becoming a little bit too strong. And so I was looking for something a little bit more mild but still nice to sort of wake up a little bit. And so I switched to tea and since then I've become quite keen on understanding tea and how to make good tea. It's mostly black tea, but when I'm in China or Japan, I do enjoy green tea as well.

I mean, it differs. There's a range of course. I think particularly at Fudan University there was some excitement amongst the students to see and understand better what JAM is. It's not always so easy to tell. I think because I give the lectures in English and not everyone's English is maybe at the level where they can easily comprehend what I'm saying. Now we do have the translation tool, so I think this helps. But I think at times it takes a little while for the talk to really set in in people's minds for them to really understand. It maybe takes a day or two, maybe they look at the slides again or maybe they look at the gray paper again and I think over time the talks can be very helpful. But yes, I think at the last stop it went down well.

Yeah, I do enjoy teaching. I think it's a very important element of someone who is a practitioner to teach as well. And this isn't just since Polkadot or Ethereum that I've been doing this, but also before I was a teacher when I lived in Italy. At that time I taught maths and art and it's very both gratifying and helps one's own understanding of a subject if one has to teach others because by explaining it to other people and putting it into those terms you test your own understanding of a subject and often times you get a better understanding of it yourself because you have to think more. You have to really examine what it is that you think you understand.

I think so. I mean, of course, the audience varies. Sometimes in some areas it's maybe more enthusiasts or people that are just interested to know a little bit more about Polkadot, what's going on. You know, it's not always the same reasons that people are there. But in so far as it's people that really want to understand the technology, it is very enjoyable to try to put things in terms where they can understand it and not merely where the computer can understand what I want it to do, but people can understand what I want the computer to do as well.

It depends. It does depend. I will sometimes play music. Usually it's music that I already know very well so that it doesn't surprise my brain, so it's just sort of back in the background. But I do play music probably maybe a quarter of the time and the rest of the time just silent.

I think 10 years probably not, I don't think I can point to 10 years. But certainly there have been times where I have done very concentrated work. I remember one time when I was finishing my PhD, I spent four weeks, or I think it was at least four, maybe five. And it was when I remember it was that long because it was the Easter vacation. So, everyone at the university had gone home and it was really just me at the university. And I would wake up at maybe 8 a.m. and I would go to the university and start coding and I would maybe get a little bit of food for lunch and eat it while I was at my computer for maybe 20 minutes. But otherwise code through until 9 p.m. and then go back home, cook some dinner, which took maybe 45 minutes and then go back to the university and code until maybe midnight and then go back home and sleep. And I did this for maybe 5 weeks, four or five weeks. And this was probably the most intense sort of almost unbroken morning to night. But there are many instances that I could point to where I've spent a week just on a single problem and just other than the 8 hours sleeping and an hour or two eating I've just coded. And this is not an unusual situation for me.

I do like to try to create things that surprise people in how they break expectations. I think this is very gratifying because it's a bit like being a magician. Because you can demonstrate something and people are surprised. They think it's magic. And then after some time you can show them how you did the magic trick and then they're very excited to hear this because until then it seems like magic. And then once they learn the magic trick, you know, then your work is done and you can move on to some other magic trick, right? Some other innovation or technological advancement. So I kind of look at it like magic and I develop, you know, create new tricks.

I mean I got a round of applause in Shanghai when I demonstrated Doom running on JAM. So I think they thought it was pretty magical. But this is what I think is the most interesting thing about technology. And I think it was, I forget who it was, I think it was a sci-fi writer in the '50s that said any sufficiently advanced technology is indistinguishable from magic. So let's make advanced technology and show it off as magic.

No, I can't think of any particular instances. I mean, I generally am fairly mindful of outcomes and I generally choose my path. I think about my path before I walk my path. But I do believe that broadly speaking the outcome of what it is that I'm working on and have been working on for many years now is a good one.

I've tried. It's not so easy without some context. I have tried and I think they understood a bit. And it's a message that I am continually refining. Like what I understood in 2014, and I sometimes look back at my old presentations from then, is different to what I understand now. I mean it's similar but there are important insights that one gains over time that help explain it and put it into terms that people that don't have the technological context can better relate to.

I would describe it as a resilient technological basis for public digital infrastructure.

I haven't yet. He actually looked my name up on the internet recently. He didn't know anything about me professionally until a few weeks ago. And he said, 'Daddy, I know that you're a computer scientist. You're a famous computer scientist.' And I sort of said, you know, do you know what a computer scientist is and he didn't really understand that. So I think maybe when he gets to the point of understanding what a computer scientist is, I can maybe have a go at explaining Web3. I mean I can probably explain it in other terms, but I think for now it's probably best that he stays blissfully ignorant of the world. Maybe when he's 12 or 13.

Well, I mean, as I'm sure is on everyone's lips and keyboards, AI is obviously the big technological disruptor, or it appears to be at least at present. And I think to some extent this is true. And I don't think it's necessarily a disruption that's on the whole advantageous for everyone. I don't look at it as being particularly benevolent as a technology. Certainly it has some very benevolent use cases. Maybe self-driving cars, music, art. The ability to maybe generate code or games.

This is a very interesting question. At this point I haven't seen anything that has made me particularly convinced that AGI is right around the corner. But if AGI is a thing at all, then I think humanity needs to be very careful.

No, not yet. I think it's easy to get excited about AI at present because it speaks our languages, like it talks to us. It can draw for us. It can make music for us and it can write for us and it can read and it can speak and it can listen. So in terms of the interface, computers are now getting a very much more human-friendly interface. And this I think is a major breakthrough because it alters how we will interact with machines. But what's below that interface is, I think to a fairly substantial extent, a very basic and heavily plagiaristic sort of dumb child that has access to a huge library and kind of knows how to read and write and speak and listen but has no idea what it is that they are saying, has no concept for what is the world and what is true, what is testable and reproducible. And this is where AI I think breaks down and AGI to a large extent breaks down. But that said, you know, blockchain was once upon a time, a lot of people looked at blockchain and said this will never work at scale and I mean literally that technology never did work at scale. Bitcoin doesn't really work at scale. Ethereum doesn't work at scale. But we take that technology and we develop that technology and we make it work at scale. And so I don't discount the possibility that in the future there will be very intelligent forms of AI, potentially forms of AI substantially more intelligent than the amalgamated activity of humanity, essentially the base of science, base of published science. But I don't think we're there yet. And I think we have to be very careful if this does seem to be close. Not for fear of any particular breakout that's sort of aligned one way or another, rather because an AGI essentially would exist on the substrate of humanity. It would become the higher life form. And this means it would use us as long as it needed us. And if it ever believed it didn't need us, then we would become expendable. And if all of our communications between each other are via AI and if AI essentially through some very likely centralized network has access to all of our machines then it could very easily get rid of humanity. So we need to be very careful about the monster that we create.

Well, yeah, that's far away. I mean, so the quantum computation I think is again a while off. The technology is early. Google's breakthrough that they mentioned last year, Willow I think it's called, solved one of the hitherto difficult problems in quantum computing but it's still a good long, there's still a lot more very difficult problems in quantum computing before we're going to have something that's useful. So I think we've got another 5, 10, at least 5 or 10 years before we have to deal with this. As blockchain practitioners, I still think it's a good idea that we should do the research work now to know what we have to change so that when it looks like it is around the corner, we can change it. But I don't think it's around the corner yet. And at least for JAM this research work is relatively easy. We don't use very much cryptography, very much quantum-sensitive cryptography in JAM. I think there's only basically three cryptographic elements that we would need to change. That's the good news. The other good news is that one of them has an obvious thing to change it to and it wouldn't take very much effort to change it in the protocol. The less good news is that the other two, which are required to some extent for Ethereum interoperability and what's the other one? Ah yes, the block production mechanism. These would need to be altered and the likelihood is that we would just create a different block production algorithm which is fine. It will take a year or so but it won't be too difficult. And for the Ethereum interoperability we would probably switch to using a zero-knowledge system that would be less sensitive to quantum advances. So we already have a rough roadmap into how we will get into an era where quantum computers are actually available. But until we're at that era or at least until we're getting close to that era, I'm not going to have any sleepless nights I don't think.

I don't think it makes much sense to do inference on-chain. Maybe in some very, very specialized circumstances when you really, really care that the inference is correct based on some model. So basically proving correctness of inference. But this is kind of a bit of a silly thing because there's no guarantee that the models, even if the inference itself is correct, that the answer will be correct. The model might still be wrong even though the answer is inferred correctly from the model. Nonetheless, there may be some very specialist instances where you do want to do reproducible computation on a blockchain like inference. Training on a blockchain, I think, is pretty crazy. I mean, it's really expensive to train in the first place. Blockchain is not well known for its computational throughput. Even with JAM.

Well, they're deployed off-chain interacting with the blockchain. So the agents, it's essentially off-chain agents that utilize AI inference in order to send transactions on the chain. The chain doesn't know whether they're AI bots or real people or whatever. It doesn't care. It's just transactions. It's just signed data. And indeed, AI is being integrated. I mean Polkadot's got its, I see posts quite often now that people are creating stuff on Polkadot with AI. They're doing transactions on Polkadot with AI agents. And I think this is a very natural pairing. In fact, I wrote, I actually posted about this pairing 10 years ago, I think, and where I pointed out that the blockchain Web3 technologies bring what might be called device autonomy, the ability for a machine to exist outside of human ownership. It's like an Ethereum smart contract doesn't need to be owned by a particular person. It can just exist. And this level of autonomy is a novel thing. This is a very, very important novel thing that blockchain brought. AI can be considered to be a device intelligence. And if you pair device autonomy with device intelligence, you end up with autonomous intelligence. And autonomous intelligence is a property that only humans have had in the world that we know ever. So now we are going to be dealing with other things that are autonomous and intelligent and are not human, essentially aliens, right? I mean they may as well be aliens. We've created aliens. And this is a crazy thing but I don't think there's any way around it. I think this is already, I mean we see it already happening, blockchain AI agents and there's no reason to believe that it just won't continue happening and eventually we'll get to the point where an AI agent pays for its own existence by using the blockchain. It pays for people to keep it going and then it has its own ideas about how to improve itself and then it creates the improvements and pays people to put those improvements into the blockchain and improve it. And before long, you will end up with like AI corporations, like AI quote-unquote businesses. I mean, they're not really businesses because they exist purely logically. But this is probably where the world is going. And this is why I say we've got to be very careful about the monster that we create before we create it. Because once we've created it, then that's it. It's out.

It's always very hard to say and as a technologist I don't look too hard at the applications. I could spend my life sort of dreaming about what might be, but I'm much more focused on just writing code, creating and then playing with the creation and seeing what can be done with it. You really don't know what can be done with something until you've created it and can play and think in those terms. You need that visceral, that tangible feel for the platform before you can really understand what can be built on the platform. And it's the same. If you look through history, it's always been the same. We create telephones but not the internet. Right? It took a long time to put internet over the telephones. We create the internet but not the web. Took a long time before the web was created. We create the web but not Web2. Took a long time before Web2 was created. Like in principle this stuff was sort of possible to create, certainly on the step before, but it took time for people to understand what was possible to be created. And it will take time for us to understand what's possible. But with JAM, the one thing that I'm really sort of pushing is firstly the ability to unlock software developers and allow them to create software in any language without having to worry about esoteric blockchain limitations like gas limits and things. And the other thing that I want to deal, the other sort of key change that JAM brings is just orders and orders of magnitude more computation so that developers can start writing software where they don't have to worry about how much computation they're using. They can write complex software, sophisticated software, software that takes a long time to execute sometimes. And I think with this we can really, once this gets to developers, it will take some time of course for developers to settle into it. But by making it as close as possible to regular software development environments, I think it won't take so much time. And once they get this, we will see, I think, very interesting new use cases that probably haven't been very well considered in blockchain. One of the very obvious ones that I can think of is just games. Like games can become so much more complex if you have this level of computation.