Human Protocol: Why Marketplaces Like Uber Could Run on a Decentralized Engine

Ashton Addison: How did you get into blockchain, and how did you end up at Human Protocol?▶ 00:22

Harjyot Singh: I'm from India and moved to the UK in 2013 to study artificial intelligence and computer science at the University of Edinburgh. A few friends and I were having a beer at university when someone asked if I'd looked at this coin, Ethereum. I asked what the tech behind it was, and it suddenly got really interesting. Near graduation I started my own company, Provenance. It began with plagiarism detection for artists and musicians, but that track didn't go anywhere, so we moved to misinformation detection. Provenance now serves over 50 clients, from news channels to independent media outlets. When it was running well, I decided to step back and try new avenues. I met one of Human's co-founders, and I was hooked by the idea of diversifying jobs across the globe and making them accessible to everyone. I joined almost two years ago as a software architect. Now, as technology director, I manage the product, the core tech, some technical partnerships, ecosystem development and even our marketing.

Ashton: What is Human Protocol, and what is the vision behind it?▶ 04:09

Harjyot: We are building decentralized infrastructure that powers dual-sided marketplaces where humans and machines can collaborate to solve any fungible task. Take platforms like Uber and Uber Eats. Someone wants something done, whether it's a physical job, a food delivery, a ride or a logo design, and someone on the other side can deliver it well. We saw that all of these share a four-step process: creating the job, matching it to a person based on skills and availability, doing and verifying the work, and then settlement. You can abstract that into an engine that powers marketplaces. If Uber were being born today, they wouldn't have to build that tech. They would leverage Human Protocol and focus on business development, marketing and PR.

Ashton: Why build this in a decentralized way rather than a centralized one?▶ 06:33

Harjyot: Most of these companies are centralized. If you're the driver or the delivery person, they control your profile. They can decide not to pay you, charge you penalties or call your work obsolete. Should they be the judge of that? No. If you do it in an entirely decentralized, permissionless manner, we as a protocol, the technology engine, have no say in that part of the process. It's governed by multiple actors in different parts of the system, such as the oracles and the validators. Here's another example. Say you're a worker in Vietnam or another emerging country. Do you really trust the banking system to pay you on time? Or do you trust an automated software system that you can see in front of you, where within a few minutes you can go and spend that money?

Ashton: Where is the team at with the protocol? Are there apps running on it yet?▶ 08:49

Harjyot: The initial version of the protocol is already running on Ethereum mainnet, so the end-to-end process I described is live and available through an API. Given our domain expertise, we built our own first marketplace on top, the Human Protocol machine learning marketplace. It's thriving with one of our partner projects, hCaptcha, which delivers captchas that don't just stop bots but also reward the users who take part. Captchas are used to train machine learning models, so suppliers provide the image labeling and fine-tuning they need. We push it through the Human app on Ethereum mainnet, where people solve the captchas and earn money. We're adding more tools to the ML marketplace. CVAT, which is Intel software, is coming soon, and a text data-labeling tool called Inception should follow quite soon. We've also created the Human Grants program. If you have an interesting marketplace idea and the domain expertise to deliver it, you can apply for funding between $50,000 and $250,000, plus technical and business support.

Ashton: When a job is posted, is matching first come, first served, or does an algorithm decide who gets it?▶ 11:19

Harjyot: First come, first served is not a good business model here, and it is not a technologically sound solution either. We use a combination of parameters, and they're not controlled by us. Third-party actors run various oracles that first determine an earner's skill sets and what they specialize in. As people keep doing jobs, those skill levels are adjusted. The oracles also evaluate availability, delivery time and the quality of work, and adjust these per job. That gives us a rating system where the matching becomes stable and quite successful. There are penalties in place too. The amount of HMT, or any other token you hold on a platform, also influences it. It's quite a complex system, and of course this is machine intelligence optimizing the matching process.

Ashton: What role does the Human token play in the ecosystem?▶ 12:55

Harjyot: The Human token, HMT, is on Ethereum mainnet, and we're now diversifying into the first EVM-compatible sidechains. It's essentially a utility token that serves a part in multiple processes, whether for the various oracles, for validators or for settlement. It depends on what the creator and the earner of the job want to do, so there's a lot of flexibility. There's more on the full tokenomics on our website, and our papers are out there if you're curious.

Ashton: Are you focused on ambassadors, or also on developers and service providers?▶ 14:33

Harjyot: The ambassador program is our first foray into building the diverse communities we're looking for. Earners won't come only from English-speaking regions like North America and Europe. They should come from all over the world. Our ambassadors are educators, creators, moderators, organizers and campaigners who will reach deeply local, maybe even non-English-speaking communities. We've also just launched the dev bounty program. We want a healthy community of developers who actively contribute to the protocol's development, features and bug fixes, and get rewarded for it. Everything is on GitHub, and we're seeing a lot of traffic through these channels.

Ashton: What will we see from Human in 2022?▶ 15:48

Harjyot: The fundamental thing we're hell-bent on is demonstrating our utility and getting people to work and create jobs. The Human app is the first step. It's your one portal to the protocol as an earner, and a lot of users are already coming in and doing different kinds of jobs. Now it's about scaling that on the ground. We also realized early on that being on Ethereum mainnet isn't always scalable for a high-transaction-volume protocol like ours. So we're going deeper with partnerships with some of the biggest names, including Solana and Hedera, with Polygon and Avalanche coming up. We'll try to distribute the protocol onto as many networks as possible. The other part is ecosystem development, through the ambassador and dev bounty programs. The community is our foundation, and getting it to run the majority of the protocol is our goal.