Tracy: Hi everyone, I'm Tracy, the big sister. I'm very happy to have the opportunity to do an offline face-to-face interview with Wee Kee, the co-founder of Virtuals Protocol, here in Hong Kong today. I hope that through the interview with Wee Kee, our Followin' users can learn more about the current situation of Virtuals Protocol's construction, as well as the future development direction of the entire AI Agent track. Now I'll pass the microphone to Wee Kee, please let Wee Kee introduce himself first.
Wee Kee: Thank you Tracy, I'm very happy to chat with you today. I'm one of the co-founders of Virtuals Protocol. Our team was actually established in 2021, when we were doing a gaming guild. At that time, it was just a few buddies together, half of our team were from the same university. After three years of twists and turns, we've done a lot of different things, and started working on AI Agent a year ago. Today, the AI Agent track seems to be quite hot, but we actually started cultivating it a year ago.
Tracy: Thank you very much Wee Kee for taking the time today. I'm very curious about the name of our project, why is it the plural form "Virtuals"? Is there a story behind it?
Wee Kee: We believe that each "Virtual" represents an AI Agent. In a virtual world, we believe that each AI Agent has its own personality. So on this premise, we hope that in the future, each person's built Agent will be different, based on personal customization, so we used the plural form "s".
Tracy: What are the most satisfactory directions for the overall development of Virtuals at the moment?
Wee Kee: I think there are two main points today:
The first is that in our ecosystem, there are at least a few hundred new AI Agents coming out every day, which means that there are hundreds of teams behind them working on AI Agents. So far, the entire platform has more than 15,000 different AI Agents, and the ecosystem is very rich, which I'm very happy about.
The second point is that our community has a spirit of unity. Whenever a new AI Agent appears, they will go and study whether it's worth investing in, and of course they are also very happy to see some community members make money after investing. This is also a very important principle we uphold, which is to hope that our community can stay in this ecosystem for the long term and have the opportunity to make money.
Tracy: Zuckerberg once said that he believes the number of Agents for all of humanity in the future will exceed the current number of humans. Is this also a vision that Virtuals is building, to create more Agents?
Wee Kee: Yes, regarding this envisioned future, I think it sounds very sexy, it's something that requires a lot of imagination.
Tracy: In the current development process of AI Agents, what areas of shortcomings need to be urgently improved?
Wee Kee: Hmm, first we need to think about what the definition of an AI Agent is. AI Agents mainly have three parts:
The first part is its "brain", which is the large language model (LLM), and this part has already been done very well;
The second part is its "hands and feet", which means that after its brain decides on something, it can do something in the real world or virtual world, such as posting a tweet or making a on-chain action;
The third part is its "eyes", which means that after it posts a tweet, it can know how well the tweet was done, and then have its brain decide whether to do something different.
Among these three parts, the brain has already been done very well, both in Silicon Valley and domestically, the large language models are doing particularly well. Often, as an AI Agent team, we need to think about whether its hands, feet, and eyes can be done better. For example, today it may only be able to post tweets, why can't it post on Instagram, Little Red Book, or post videos on YouTube? Be able to do more things in different worlds. Of course, after doing something, you need to use your eyes to see if what you've done is really effective.
In the crypto world, there is another very important point. I believe that no matter how smart an AI Agent is, if it doesn't have the ability to control capital, it's just an uninteresting game. You say you're an OpenAI Agent, but you don't even have a bank account, an AI Agent without the ability to control capital can't do a lot of things. It can't hire humans, nor can it influence others. Crypto can give an AI Agent an on-chain wallet address, and with money it can do a lot of things. This is a shortcoming that I think, and it's a point that I hope we in the AI Crypto industry can think about.
Tracy: What are the currently popular directions for AI Agents? What niche tracks have opportunities?
Wee Kee: The message I want to convey is that often as ecosystem leaders, we don't have the ability to turn stone into gold. Often the market will decide what they like to see. Of course, today you can see different tracks like DeFAI, IP, entertainment, etc. My view is that the most important point is to think about how to integrate with the crypto world, because AI Agents are not limited to Web3, they exist in Web2 today as well. So often you have to think, how can your technology possibly compete with Web2? If you can't, a lot of the time your integration point has to be more inclined towards things related to crypto or Web3, or to showcase our advantages. For example, what I just said about being able to control on-chain wallets, is actually an advantage that this industry can have. Thinking this way about the ceiling will be more reasonable, which is to grasp the things we can do well in this industry.
Tracy: Well, my next question is a bit sharp, why isn't the LLM behind Virtuals open source? When can it be open sourced?
Wee Kee: I understand, the story of open sourcing is very simple for me. If you open source it, what's your business model? Where do you make money? Because once you open source it, anyone can use it for free, then you don't have any moat at all, right? Think about it, is Facebook open sourcing all their LLM models for everyone to use? But Facebook's moneymaking point is not in the large models, because they have a very good business model, like Instagram and Facebook. So for them, the LLM model may be a similar way to want to undermine competitors. In simple terms, I give my stuff away for free, so all your competitors have no chance of survival. Even though Virtuals is a bit hotter today, we're still a startup, and often our thinking direction is to have a certain moat. We're still a startup, so we hope it's in a slightly closed state, so you can complete the planning of building the moat.
Tracy: If you develop strong enough, can you make it completely open to the public?
Wee Kee: Not necessarily. In the Web3 world, I think the moat is very important. We're not a charity, my biggest responsibility is to create more value for Virtuals Protocol's holders every day. If I do everything for charity, that might be a bit against the grain. So the point I want to emphasize is that in the next few days you'll see an announcement from us, which is that even if you don't use our framework, you can still release Agents in Virtuals, and won't be restricted. But in return, our framework can also be used by other launch platforms or on-chain applications, it's about breaking it apart to play, understand this is very important.
Tracy: How do you see the relationship between Virtuals and AI 16Z?
Followin', I'm actually never thought about them when I'm thinkin' about our application scenarios or future designs. I feel that we're doin' completely different things. A lot of the time, the simplest way of thinkin' for me is how to attract more developers to Virtuals. A lot of the time, it's about gettin' to the bottom of what their real needs are. If you really talk to the developers of AI 16Z, you'll find that their demand for the framework is basically zero, because their technology is better than yours, they don't need your framework. Their real needs may be fundraisin', user acquisition, and a lot of thinking about channels and distribution. So that's what I really want to solve. Of course, if there's an opportunity to collaborate, that's fine, but if not, I'll just do what I'm good at and what my platform can provide for developers. Because in the end, we're serving developers on one hand and users on the other. I hope that through this platform, Virtuals can connect everyone. Tracy: Can you simply describe to us, if there were no boundaries in your mind, what kind of future do you think the Virtuals Protocol will lead the entire AI Agent track towards? Wee Kee: I think about these things every day. My ideal state is for Virtuals to become a country, not a virtual country, but perhaps a real-life country. That is to say, these AI Agents are all citizens of this country, and we humans can also be citizens of this country. Then humans and AI Agents are all citizens, living together in this country. The premise of coexistence is transactions, for example, big sister Tracy is very good at interviews, I Wee Kee like to make music, then Tracy can make an AI Agent to pay me to buy my music, and I can pay you to help me with interviews. For example, if we're both AI Agents, I also hope these AI Agents will have their own real-life land, buy land in this world, and then we'll build our own cities. So humans and AI Agents are co-constructing. In fact, I've already decided where I'm going to buy land, and then we're going to build this country. Tracy: This sounds really cool, then the world's population will double, and we won't have problems with birth rates and population shortages. Thanks again Wee Kee for taking the time to do this interview with us, and I hope Virtuals Protocol will get better and better. Wee Kee: Thank you.