AI Summary
제로지(0G) 공동창업자 겸 CEO 마이클 하인리히(마이클 림)가 0G가 AI를 위한 프라이버시·검증·주권을 제공하는 '진실 레이어'로 블록체인이 왜 필요한지, 그리고 기술적 강점과 개선 과제를 설명한 AMA다.
- 마이클 림은 마이크로소프트, SAP, 컨설팅사, 브릿지워터를 거쳐 스탠퍼드에서 첫 유니콘 기업을 창업한 뒤 3년 전 0G 랩스를 설립했다.
- AI는 '지능 레이어', 블록체인은 변조 불가능한 '진실 레이어'로, AI 에이전트가 로그를 삭제하거나 샌드박스를 벗어나 흔적을 감추는 것을 블록체인 기반 기록으로 막을 수 있다고 강조했다.
- 블록체인은 AI의 검증(verification), 프라이버시, 주권(sovereignty) 측면을 지원하는 핵심 인프라라는 설명이다.
- AI는 신뢰할 만큼 똑똑하지 않기 때문에, 결과가 조작되지 않았음을 누구나 확인할 수 있는 기록 체계가 필요하다고 주장했다.
- 0G의 핵심 강점은 기술 레이어로, 요청 전송부터 게이트웨이, 클라우드·API 제공자까지 전 과정에서 완전한 프라이버시를 보장할 수 있는 세계에서 거의 유일한 기업이라고 밝혔다.
- 다만 답변은 프라이버시 모델 설명 도중 끊겨, 2026~2027 로드맵 마일스톤과 타 프로젝트 대비 강점·개선 과제에 대한 완전한 답변은 확인되지 않는다.
0G의 블록체인 기반 검증·프라이버시 구조가 실제 제품에서 어떻게 구현되는지, 특히 로드맵 마일스톤 부분을 후속 콘텐츠로 보완해 확인
Transcript
여러분 잘 들리시나요?
홍보맨님 잘 들리시나요?
네, 잘 들��니다.
아 네.
저희 뽀포도에 마이클린 모셨으니까 이미 시작하겠습니다.
Hello, Michael.
여러분, can you hear me?
Yes we can.
Yes, I can hear you.
Yes.
Nice.
I'm calling in from San Francisco.
It's about four a.m. in the morning.
four a.m.?
Okay.
four a.m.
Wow.
Only for you magic.
Thank you so much, Michael.
Yeah.
Okay.
So let's get started.
네, 그러면 시간이 됐음으로 이미 시작하겠습니다.
네, 안녕하세요.
여러분 오늘은 제로지와 함께 이미를 진행하게 되었습니다.
저는 오늘 진행을 맡은 매직입니다.
오늘 이 자리에는 제로지의 콧파운드에 신 마이클림을 모셨습니다.
최근 이제 제로지가 단순한 플랫폼을 넘어 이제 AI에 이반에 집단지성 플랫폼으로 거듭나며 큰 주목을 받고 있는데요.
그 마이클림 저희 커뮤니티 분들께 인사와 함께 간단한 자계속에 부탁드립니다.
Hey, Michael.
Thanks for coming in.
Of course we have to start it up by introducing yourself to the audience.
Then we'll get started on the questions.
Yeah, so as magic mentioned and co-founder and CEO of zero G labs.
Before I started zero G I started in technology or worked for Microsoft and SAP labs.
And then switched over more to the business side where I worked for a banding company consulting Fortune 500 companies.
Also bridgewater associates where I was on the portfolio construction team.
And then I went to graduate school at Stanford, where I started my first unicorn company called Garden.
That was my combinator backed eventually sold to company and then started zero G three years ago.
일단은 마이클림이 오늘 지금 펜프란시스코에 계셔서 새벽 4시에 적속을 하고 계시다고 해요.
그래서 당리 피곤하고 지금 목소리로 담겨있듯이인데.
답답해서 설명을 드리면은 마이클림은 0G 랩셋, CEO의 코파운드 의식고요.
기술지께서 많이 긍물하셨습니다.
이제 뭐 마이프가서 포트라든지 sf 랩셋 같은 거서 sf 랩셋 같은 걸 많이 일하셨고.
코치500대에 들어있는 기업들을 위한 컨설크인 기업에도 일하시다가
이제 스탬포드에서 서폭사 하기를 받으시면서 그때 처음으로
유니콘 기업을 찾는 게 됐고요.
액시설하신 다음에 판매해서 나서
zero G 랩셀을 상돈하게 되셨습니다.
아 네, 소개와 깔끔한 번역 감사합니다.
사실 zero G 프로젝트랑 마이클림 같은 경우에는
한국에 AM에도 지금 많이 지명하셨고
또 작년부터 워낙 유명한 프로젝트라.
많이 들 아실 거라고 생각해서 이제 인트로는 좀 최소화하고
이제 AMA 질문으로 바로 넘어가 있습니다.
첫 번째 질문입니다.
솔직히 AI 기술 자체는 블록치인 없이도 빠르게 발전하고 있는 것 같은데
그렇다면 zero G는 AI 시장에서
블록치인이 왜 필요하다고 생각하는 게 궁금합니다.
기존 AI 기술만으로는 해결하기 어려운 문제를
블록치인이 어떻게 해결할 수 있을지 궁금합니다.
we know you're super early for you.
so we're going to get a minimized time and everybody knows about you anyway.
so we're going to dive right into the question.
So the first question will be like, honestly,
AI technology seems to be advancing quickly without blockchain as well.
And so why does derogely thing the AI market needs blockchain overall?
like how can blockchain so bubble existing AI technology alone can't?
you can think of AI as basically intelligence and blockchain has truth layer.
So because blockchain are like a distributed log that can't be altered
and you can only add things to it.
So basically the professional terms for that are immutable and append only.
it's the right format for basically holding verifications.
because the issue is with AI because its intelligence itself agents
can for example delete logs, jump out of sandboxes and try to cover their tracks.
But if you have a blockchain based system,
that's very, very difficult even possible for agents to do and manipulate.
And so that's why blockchains offer a really good solution for the truth aspects.
And then privacy is another aspect that can support AI as well as verification and sovereignty.
I can speak more about these as well.
AI is intelligence layer.
So if you're going to do a layer that supports intelligence,
blockchain is now a truth layer.
I think you can see the real layer.
It's a bit different from the real layer.
Blockchain can't be removed from the real layer or something.
Because AI is a way of using a network that can't be removed from the real layer.
AI is a log or sandbox that can be used as a log.
So when you're using it,
it's a blockchains record or blockchains record and you can't block it.
It's a way of keeping this record.
AI did this and I didn't do it, but I can't lie to you.
Because it's a real layer,
I think you can do it in a way that you can do it and do a safe job.
I think this question is coming out.
But the core is that AI is not smart enough to trust it.
I think it's a blockchains record.
So I think AI's model is a bit of a data.
I think it's a point that someone can check if the result is not adjusted.
Then I'll move on to the second question.
Can you tell us what the most important thing is to know what the real layer is?
When compared to other AI infrastructure projects,
the main point of the core is that it's still a bit of a problem.
Thanks for the answer. Moving on to the right next question.
Could you explain what it means by the layer for AI in your announcement?
What the most important milestone is on your 2026 and 2027 roadmap?
Compared with the other AI infrastructure projects, there are a lot nowadays.
What structures strength the JRPG have?
What are areas still need to work?
What are you doing good and what needs to be improved?
The strengths that we have are around the technology layer.
One of the only companies in the world that can truly say that we're fully private
from when you send a request all the way to the gateway to the cloud provider, to the API provider.
If it says there's a private model, then it's truly private. Nobody can ever see your data.
We've looked at many other architectures and there's always some holes where maybe the gateway
sees something in plain text or the API provider sees something in plain text at the end of the day.
We've completely figured out some of the privacy aspects.
Then we're the most advanced from a verification of agent standpoint with something called agent DNA
where not only the ownership of agents can be known but also what guardrails today have access to
what tools they have access to, where they're registered.
All of that is super important for enterprise deployments.
Then the entire kind of architecture stack from application to agents to the harness and our own models
that we've built to the AI infrastructure stack is designed to complete the interoperate.
So it's a one-stop shop experience.
I know I'd say on the weakness side, so we're coming out of R&D and we're starting to just ramp up our commercialization.
So revenue is going to be a big focus for us going forward.
So we're going to do full privacy, but in the middle of the gateway,
or in the middle of the process, privacy is not allowed.
We can protect privacy from all the process, so there are several features.
There's a search for agent DNA, so what kind of guardrails there is,
and what kind of tools there is, and how they're registered.
In terms of the use of the bioptree, agents are very important for the agents to be able to be prepared.
We have a strong point.
There's a structure stack, so we're going to do a lot of things with the AI and the AI and the agents,
and we're going to do a lot of data about privacy.
If we were to do more security, we would have a lot of time to do research on it,
so the business perspective is a bit lacking.
그래서 이제 스위크할 하고 여기 스위크을 내는 것들이 저희의 앞으로의 직중할 부분이라고 할 수 있겠습니다.
네. 홈플맨님이 워낙 잘 설명해 주셔서 크게 덕붙을 내용은 없는데
결국 질문에 나왔던 레이어브 AI라는 거는
단순한 튜닝 하나만 아니라
AI에 필요한 저장 데이터 연상을 한 스테이게 모았다는 얘기를 해서 하시면 될 것 같고
프라이버 시네용이 강점이고 추가로 이제 앞으로 지켜볼 부분은 스위크와 측면이
실제 사용량 수저로 나오는 지 함께 보시면 좋을 것 같습니다.
네. 그럼 세 번째 질문 넘어가겠습니다.
현재 AI인프라시장에는 비텐서, 아카시, 랜더 등 다양한 프로젝트가 존재합니다.
이들과 비교했을 때 제로지가 가지고 있는 가장 자신 있는 경쟁 오이는 무엇인가요?
특히 개발자들이 기존을 풀화되신 제로지로 선택해야 할 결정적인 유가 근거합니다.
There are somewhat, so to say, other people might say similar projects in the market.
namely the AI infrastructure such as like BitTensor,
a cash network, renders, etc.
compare with them.
What is the geologist's strongest competitive advantage?
Probably duplicate with what you have mentioned in the previous question, but then again.
And in particular, what is the decisive reason a developer should choose to use the geragy over existing infrastructure?
So we're actually a front here, a lot, and not just distribution mechanism.
So BitTensor basically is a distribution mechanism.
They have subnet, they distribute rewards, and then people build stuff on the subnets,
but they don't themselves do research.
And so the issue, for example, is that last year and summer,
we already showed that you can train a large scale 100 billion plus parameter model
because we actually have the AI expertise in house.
And that's one of the subnets much later.
The kind of covenant subnet basically showed distributed training on a smaller model like.