HUMAN Protocol is a blockchain infrastructure designed to decentralize human labor by supporting the growth of digital job marketplaces.
After recently launching on the Ethereum mainnet, the protocol has now gained the capacity to fully automate the lifecycle of data labeling jobs, enabling the collaboration of humans and machines to create and complete a host of real-world, fungible assignments.
Working with artificial intelligence and machine learning technologies, users can now be rewarded in HUMAN’s native token, HMT, for the successful completion of anti-bot visual challenges such as the identification of traffic lights in a grid image. This data is then collated and used to support the eradication of biases in the labor markets, promoting a more circular gig economy.
For a deeper reflection on these recent announcements and the wider implications for the blockchain-tech sector, Cointelegraph spoke with Harjyot Singh, the technology director at Human Protocol.
Harjyot is a prominent entrepreneur in the field of fintech engineering with an academic background in computer science and artificial intelligence.
His present focus is on “exploring how cutting-edge technologies such as AI and blockchain can improve the day-to-day experience for the majority of internet users.”
Cointelegraph: How will HUMAN’s recent announcements (launch on Ether mainnet and release of CAPTCHA web app) support the growth of the protocol?
Harjyot Singh: We’re excited about our recent achievements. HUMAN Protocol’s launch on the Ethereum Mainnet allows us to realize the first instance of a HUMAN decentralized job market. This is also about the Protocol’s evolution; HUMAN Protocol currently processes a significant amount of user interactions every day through the applications it supports. It is designed to operate across multiple blockchains, with Ethereum being the first mainnet deployment. What we learn and make possible here we can utilize and execute elsewhere, including Solana and Polkadot.
Obviously, the launch also enabled us to list HMT, which helps us grow the HUMAN community and incentivize broader participation. But the real growth comes through the HUMAN App: the first gateway into the HUMAN ecosystem, and the first means through which individuals, located anywhere in the world, can directly earn HMT for completing tasks.
It is also important to note that the HUMAN app is not just a CAPTCHA app; it allows many kinds of tasks to be performed by people.
CT: Readers will be familiar with Google’s reCAPTCHA system. How does the HUMAN model differ from a technological standpoint, and what are the benefits of a human-centric identification method?
HS: It is important to note that hCaptcha isn’t part of the HUMAN Foundation; it is simply an application that uses HUMAN Protocol. HUMAN has a much broader goal of tokenizing many kinds of human work, not simply the narrow set of tasks that can run via a CAPTCHA.
That said, a key difference between reCAPTCHA and hCaptcha is that hCaptcha pays websites for the work their users do when they solve a CAPTCHA, rather than forcing them to donate that labor to Google.
CT: Vitalik Buterin recently advocated for a transition to “proof-of-humanity” governance across DeFi. If widely implemented, how do you envision this impacting the space?
HS: I think it will make DeFi a fairer space. Right now, systems that distribute votes as a function of wallet balance cause huge problems; it allows crypto whales to affect proposals in their favor. HUMAN’s “Proof of HUMANity” would allow for one vote per verified human user, which would also combat the prevalence of bots. Because Proof of HUMANity is the first and only on-chain human verification system, it makes sense for an on-chain DeFi world.
But the potential of Proof of HUMANity does not stop there; any space in which bots cause havoc — such as frontrunning on exchanges — can potentially apply Proof of HUMANity to solve it.
CT: Could you share some specific examples of HUMAN contracts that could be facilitated on a marketplace utilising Intel’s video and image labelling system CVAT, as well as the text-based INCEpTION?
HS: A Requester at an AI startup needs one hundred thousand images of damaged cars labeled. They provide the images, along with a sum of HMT which is held in the smart contract until work is complete. HUMAN Protocol agents ensure the data is safe for sharing, and prepped for the applications; HUMAN Exchanges can then distribute tasks intelligently to Intel CVAT users (who could be running on different chains – and the work is sent to different chains depending on speed, cost etc.).
The Worker connects to an Exchange, sees the work, and starts completing the granular work on Intel CVAT by drawing detailed boxes/shapes around damaged areas of a car. The oracles, which record and assess the work, then update the smart contract to reserve HMT for the Workers who completed work.
CT: How does your native currency HMT — and specifically a worker’s token value — determine the priority for task offering?
HS: We utilise proof of balance as one of the factors contributing to the task offering order book sort, in other words how many tasks will go to one worker or labor pool vs another. However, to reduce friction in the system we also run a weighted average over numerous other parameters in order to enable new users to join immediately.
CT: How advanced do you believe AI and machine learning systems are at present, both in terms of technical ability and cultural awareness, to support HUMAN’s scalability?
HS: AI systems are currently good at specialized intelligence. That is to say: they are good at performing specific, linear tasks, such as GPS, chatbots, or Amazon’s Kiva bot which relays boxes to and from Amazon workers. But AI is not so good at generalized intelligence, which is the domain of flexibility, response, and adaptation, a domain in which humans thrive.
In terms of cultural awareness, I think we are primed and ready for the next wave of AI. AI products are already ingrained in our lives — from face-unlock systems on your phone, to robot cleaners. If anything, however, I think the culture overestimates current AI capabilities; I think most people believe AI is more intelligent and capable than it really is, because we have been talking about AI since the 1950s, yet progress has been shaky, as we saw in “AI winter” of the 1980s. For instance, we have already integrated into our cultural knowledge the inevitability of driverless cars, and yet they haven’t quite taken off yet. I think we’re ready; I think people are just waiting for the products.
CT: As we enter into a more automated economy, how crucial is it for us to build systems where machines serve the true values and needs of humans?
HS: We hear a lot of different things about AI, and the implications of introducing machines into job markets. But rather than replacing human workers, we like to focus on how machines can support and even empower them. Intelligent automation means that a greater quantity of remedial work — the small tasks — can be handled by machines, which helps to maximize the time, energy, and focus of human workers.
Humans are capable of feats that machines are not — creativity, ingenuity, imagination — while machines are more efficient at performing repetitive tasks. An infrastructure that supports this scales particularly well to the growth of knowledge workers, who are capable of providing nuisance and specialist input but whose time is increasingly scarce. It also means providing specialist workers with the data they need to make informed and confident decisions.
HUMAN Protocol is designed to allow machines to complete repetitive tasks, and to request the completion of those tasks from other machines. Through this, we want to empower human potential, and to provide the space and focus for creative problem-solving.