Builder-owned Intelligence on-chain is the future of work
Decentralized AI (DeAI) was the talk of town at this year’s TOKEN 2049 event in Singapore. But an open question remains — What’s are novel use cases of DeAI? Who cares and what’s the end game? After all, there has to be a reason why $12.7B has been invested in this space till date. Philosophical primitives of centralization cannot alone be the reason why someone uses DeAI technologies.
What is decentralized Intelligence?
Simply put, it’s AI models and agents built using decentralized technologies underneath. It matters for two reasons:
- It’s verifiable, on-chain, end-to-end, including but not limited to tracking where underlying specifics such as where did data came from, who owns it, license compliance, where is it stored, who can access it, where was it trained, how was it trained, who trained it, who owns it etc.)
- Enabling ownership and monetization so everyone can be shareholders and beneficiaries in what they have helped create.
We believe all intelligence that is actively being produced today, and existing intelligence, will be verifiable on-chain over the next few decades. This will begin at the base layer of AI production lifecycle moving all the way up to consumers of such intelligence.
Our core thesis is that on-chain verifiable AI will eat off-chain AI for a simple reason: It is vastly better at mitigating risk — financial, legal, maybe even existential.
Reppo and the technology we are building will play a critical role in this transition.
Introducing Reppo pods
A Reppo pod is a grouping of intelligence (human+ AI,) coordinating driven by incentives. Most companies and organizations today can be viewed as a Reppo pod i.e. A group of intelligent people + tech driven by money, prestige, impact etc.
A VC firm, for example, can be thought of as an intelligence pod. Everyday they work to grow it, fight to keep it alive, and protect it from others (although not for long).
A lot has been written about organizations as living organisms but the main question we are interested in answering is that in the absence of fiat currency, can you coordinate intelligent work between Humans and AI eg. analysis, consulting, research, primarily using crypto-economic incentives, including non-monetary incentives such as social prestige or camaraderie?
We think the answer is yes, and we believe decentralized intelligence networks, which consists of aforementioned groupings of pods termed “clusters”, trained on domain-specific knowledge, and collaborating to solve problems, will be the future of work.
Everyday, instead of logging in on zoom or going to work, people will contribute their domain expertise, which is actively developed by connecting information and signals in the real-world to the pod, fine-tuning the IP in the pod to a level where it’s so valuable that many will pay for it. People will be part of multiple pods and will have the power to decide how and when to monetize their intelligence. While machine intelligence will surpass human intelligence, the latter will continue to compliment the former to provide support for infinite context windows.
Cool Idea. How would this actually work?
The technical details can be found in our docs which explains how we coordinate protocols and platforms to create a system that enables domain-specific inference engines that have individual ownership and monetization rights.
Each pod is incentivized to produce intelligence or bring it on-chain, either using ModelRivet or by accessing marketplaces like (www.reppo.ai) , Genlayer, and Naptha.ai, and scale their pod to integrate into existing AI workflows of enterprises and developers. You can think of this as a permissionless and decentralized version of Hebbia, built using crypto railes.
We believe that the next frontier for AI is combining agents and models with domain-expertise, fine-tuning on niche datasets, and solve context-specific problems instead of feeding AI models the entirety of the internet. All off-chain orgs will live on-chain.
Whenever the idea of crowdsourcing intelligence or intelligent work on-chain is floated, eyebrows are raised, which falls in two categories —
We need proprietary data to build quality intelligence!
While it’s fair to say that quality and useful intelligence does not just come just from public data (although there are examples of platforms such as Arkham that have proven that significant intelligence can be derived from public data), accessing private data is becoming more and more accessible as solutions leveraging Zk tech enable access to this asset class in a privacy preserving manner Eg. Provably.ai and Gateway Protocol
We are building Reppo to be the best interface for humanity to interact with Decentralized AI, allowing private sharing of datasets, on-chain data licensing, & collaboration between Reppo pods and clusters that crowdsource intelligence to serve end-user demand driven by needs to reduce infra spending and critical needs for verifiable ML/AI such as underwriting risk or enabling immutable open-source systems.
We need to access context to meaningfully crowdsource intelligence
Once we have built the rails to access public and private data in a privacy preserving and secure manner, how does one access context? After all, the person working at an organization for the last 20 years has a lot more context than a random person on the street, right?
Well, not always. At least in the world of web3 and blockchains, context onramps are more accessible than one might think. From knowledge accessible on network-specific discord, X, and even direct engagement with members of a community, one can gain enough context to meaningfully build intelligence that is relevant to the end user. We have proven this over the last 10 months where individual model developers built AI models without any previous context on Web3 networks like Filecoin and DIMO. These models were used by the communities and stakeholders and were shipped in 4 to 6 weeks.
With Reppo’s intelligence clustering approach, we are taking this a step further by creating native incentives that encourage collaboration between pods and improve efficacy of context onramps exponentially.
