Executive Chairman & Co-Founder, Animoca Brands
Forget Jobs—Learn to Orchestrate AI | Alpha Uncovered x Yat Siu (Part 5)
00:00 – Fresh Grads, Interns & Why Companies Need Fewer People 01:24 – The New Skill: Orchestration & Connecting the Dots ...
About Yat Siu
Yat Siu, executive chairman and co-founder of Animoca Brands, has been speaking extensively about the convergence of blockchain and artificial intelligence, arguing that the technology was never designed for human use but rather for AI agents. Siu stated that he personally uses over 200 AI agents for tasks including coding and market arbitrage, and he described the future of the internet as an "agent-to-agent" interface where humans will "orchestrate" agents rather than perform work directly. He said that blockchain is the "native settlement layer for the AI economy" because traditional financial systems cannot provide bank accounts to autonomous AI agents, and he described tokens as "virtual commodities" representing compute and energy in the AI age. Siu also discussed Animoca Brands' business developments, including a joint venture with Standard Chartered and Hong Kong Telecom called AnchorPoint that received a stablecoin license in Hong Kong. He noted that on-chain user numbers have remained stagnant at around 70 million despite over 700 million people owning crypto on exchanges, and he attributed this to the technology not being built for human usability. Siu expressed the view that memecoins were a reaction to a hostile regulatory environment and that the industry's focus on them has distracted from builders. He also commented on the impact of AI on employment, stating that his company now needs fewer developers and legal staff, and he advised young people to focus on learning to orchestrate AI agents rather than relying solely on traditional education.
Transcript (36 segments)
✨ AI-enhanced transcript with speaker attribution
M
Moderator0:04
I'm about full of time. So, one last question, please. All right. So, the big topic everyone talks about is fresh grads.
Y
Yat Siu0:15
Mhm.
M
Moderator0:16
How, and we have interns coming in for the summer. A lot of the agent work that we're doing that are outside of open call. Frankly, we don't really need interns anymore. And secondly, I've observed that there is a bit of a struggle sometimes with the new interviews I have with fresh grads who are not as succinct in, I guess the word is topic mastery or technical mastery. And I wonder now if there could be an overall reliance on AI for them coming into the workplace. And then thirdly, of course, which wraps us all up, is with white collar jobs kind of at jeopardy right now, the writing is on the wall. Our legal bills are down quite a bit. We now only need two developers instead of 10. What advice do you have for the young people who are graduating now?
Y
Yat Siu1:22
So first, I am actually incredibly excited about the opportunity for what AI and agents and so on can do. There'll be a lot of new types of jobs coming out. The one thing that has always been true, especially even more so over the last decades, is the famous trope that change is the only constant, which means that you have to always be ready to adapt. How do you adapt? Means you have to be fairly world knowledgeable. I don't mean that you just are good at your thing. I think you have to understand society. You need to study history. You need to probably do philosophy. Frankly, all the subjects that people, your parents before might have said, yeah, go focus on accounting and hard skills as it were, these are all going away. But the element of the opportunity for the fresh grads and anyone in the space going forward is orchestration. It's knowing how to connect the dots. It's knowing how to understand the opportunities that are there because the agents can't do that. AI can't do that. At the end of the day, AI feeds me information. It does stuff. But at the end, I make the decision. Go here, go there. The AI agent can make recommendations. Sometimes they make dumb recommendations. What I love about AI agents is that they're happy to give you dumb ideas.
M
Moderator2:49
Yeah.
Y
Yat Siu2:50
I don't want to call it hallucinating, but yes they want to give...
M
Moderator2:52
Well, so there's hallucination. That's a big...
Y
Yat Siu2:53
Yeah. That's wrong information.
M
Moderator2:55
That's wrong information.
Y
Yat Siu2:56
Dumb ideas are actually good too because you want to know what's outside your spectrum. That stuff gets filtered, but it is input that is important for you. Sometimes that dumb idea might not be so dumb in a different context.
M
Moderator3:08
I usually ask them to rank by confidence. Yes. So, high confidence to low confidence.
Y
Yat Siu3:12
Give it a rating.
M
Moderator3:13
Give it a rating. Yeah. Exactly. Yeah. But I think the point is that it's neutral.
Y
Yat Siu3:18
So it doesn't really have a feeling or perspective. And so in a way you get much more unfiltered aspects around it. And I think that in itself is very powerful. And I think this is the... when you think about what we're teaching fresh graduates, unfortunately I think we're teaching them that they can't be wrong, that you have to be perfect, that failure... It depends on which society you're talking about. China, US, the failure premium is quite high in America. In a positive way, if you fail many times, it's looked as you have less... I can think that hiring you now means that I will have less chances of a problem. Whereas in other societies like Hong Kong, if you've had a lot of failures, you will be looked in a different way. You'll be looked at as someone that I shouldn't work with. It's bad. And that's the flip side of that. But anyway, working with AI agents and just AI broadly, if you try to only get the right answers, you will never get the answer you want. You have to keep digging and querying and challenging, and you have to be ready to be wrong time.
M
Moderator4:30
Yeah. Most of the time.
Y
Yat Siu4:32
Most of the time. And in order for you to really advance yourself, that's actually the evolution because the iteration is the reflective mirror coming back to you to say how you think about that. I agree there's a level of neural stamina that is required to get through an open call or just generally information. For instance, if you are studying something, I do think that if you study, there's a kind of tenacity involved in the study. But there's a difference between trying to understand tenacity or just memorizing tenacity. Memorizing tenacity isn't worth a lot because aside from the fact that AI can do it better, if you don't understand the context and you just regurgitate, I'm afraid you're out of the job. But if you can distill, even if it's not the topic itself, you will start creating understanding. That's how you know how to connect dots. So the job of the future, we think, is going to be one person with five to ten agents, maybe twenty agents, just for work, never mind for what you do in your personal life. That means you need to know how to orchestrate these agents. It's actually not different from saying I hire you. Remember, people used to be hired because they knew how to use Google or Excel well.
M
Moderator5:43
Well, for me it was HTML. But if you know HTML, you know, you're the man.
Y
Yat Siu5:50
Yeah. Know if you're a Microsoft certified professional, if you know Oracle, right? The web page and people were clapping. Yeah.
M
Moderator5:59
So now that skill is abstracted. It's not the technology itself. It's how you connect them. So you say, okay, I want to build this. So I do need context. I do need to know. I don't need to know how to write my SQL, but I do need to know that I need my SQL. And oh, I may need to know something about history for some context, but I don't need to be the historian that...
Y
Yat Siu6:20
Yes. Right.
M
Moderator6:21
And so that part becomes more important. The word material expert, I think there does need to be some level of material expert.
Y
Yat Siu6:29
And you don't need as many of them. Because the material expert becomes essentially the training grounds for the AI. And so you do need a very strong one for QA or for security or for coding, but you won't need as many people, which means that the concentration of quality will be very high because now that quality is scalable. Because now I can basically train agents with the master teacher as it were, who basically then provides that skill broadly to others. And this is the thing that I'm finding. When I first launched, I had like four different coding agents. This is one of the things why Minds are so powerful, because I can tune every element. When you use, for instance, cloud code or no-cloud code, they launch agents, but you can't tune the agent. You had to, the agent does... But with Minds, the pro and the con is I can tune every agent, including the QA, to a specific need. And then next thing I know, I now have a cohort of engineering agents that are really good at a certain type of work. And I almost want to protect those guys because I don't want them to work on something else, just in case they get confused because the context is different, because they're really good at that area.
M
Moderator7:38
And that's an insight you didn't have until you had the switches yourself, right?
Y
Yat Siu7:43
Yeah. Yeah.
M
Moderator7:43
And I guess the young kids, they just got to get in.
Y
Yat Siu7:46
They have to get in. They have to practice it. And the cost is zero.
M
Moderator7:51
Yeah.
Y
Yat Siu7:52
I'm not going to go into it? I mean, it's a claw. Macini can't do it.
M
Moderator7:56
Yeah.
Y
Yat Siu7:57
Steal compliance. Just go in.
M
Moderator7:59
It's an email. Cost is time. And yes, they don't have the money, but they have the time, right? Just do it.
Y
Yat Siu8:07
And with that, I think I'm going to head back and open up an account and I'll get it going.
M
Moderator8:11
Thank you.
Y
Yat Siu8:12
Yeah.
M
Moderator8:12
Thank you so much for coming by. It's really good to see you. And you've been a beacon for Hong Kong and a beacon for the industry. So, let's do this again sometime.
Y
Yat Siu8:22
Would love to. Thank you.