We recently connected with Harjyot Singh, Technology Director at Human Protocol, which has been specifically designed for tokenizing work.
HUMAN Protocol connects projects with millions of international respondents, offering not only access to a global Q&A, but the blockchain or distributed ledger tech (DLT) infrastructure to record, verify, and “reward any contribution.” To date, it has reportedly been applied to data-labelling markets, and to human-verification solutions like Proof of HUMANity, which provides bot protection to decentralized applications (dApps).
According to its developers, these are the only first use-cases, however; they are the beginning of (what they claim to be) “a new future of work, and demonstrate how human contribution can be verified and represented on-chain, and requested and rewarded through an automated protocol.”
Crowdfund Insider: We’ve seen descriptions of how HUMAN’s native token gets awarded to stakeholders, either websites and the people who run them, or individual users. In either case, though, how exactly does this work?
What’s the criteria for getting money for the kinds of tasks that people would do for free at reCAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart)?
Harjyot Singh: The fact that users do the work for free via reCAPTCHA does not diminish from the value of the work; it is simply that the value of the work never reaches the websites, or individuals.
HUMAN Protocol was the vision to appeal to both intermediaries, such as websites, and individuals, who host data labeling services. It was the vision to bring CAPTCHA on-chain, and, in doing so, to bring other forms of work and contribution on-chain to power new and existing markets. In the case of the individual, the HUMAN App is a perfect example of individuals getting paid for CAPTCHA-style work.
Crowdfund Insider: Do you think that the popularity of the term “human in the loop” reflects some of your philosophies at HUMAN Protocol, or is HITL in the mainstream more just the province of talking about making sure people keep an eye on the work that AI systems do, or collaborate with them in a practical way in some business process?
In other words, do you think that the general rank-and-file of tech audiences think about making things more human-centric for other reasons like those you describe?
Harjyot Singh: I think we are talking about two distinct things: human-in-the-loop refers to a strand of ML practice whereby a human verifies machine-labeling work. This is certainly an area that HUMAN Protocol can help with.
Humans collaborating with machines on broader projects may not be so much about checking machine work; in fact, a human may never know they are working with a machine, and they may not need to. So there are two things: work, and checking that work. HUMAN Protocol can do both.
Crowdfund Insider: It seems that Proof-of-HUMANity would level the playing field when it comes to small investors and the kinds of huge ones that we often call “whales.”
Can you describe this equity a little more and why you’re pushing so hard for a Proof-of-HUMANity model for blockchains?
Harjyot Singh: A very good point! Proof of HUMANity is a generic, mainstream bot-blocker for the Web 3.0 world.
One application of it could be to support a “one-vote-per-human” governance system, as proposed by Vitalik Buterin. That would certainly level the playing field and move us beyond coin voting as it stands.
Crowdfund Insider: Going back to human in the loop, how do you expect this trend to impact the AI market in the future?
Harjyot Singh: It is going to be massively important. As ML technologies become more advanced, they will begin to label increasing kinds and quantities of data themselves; this will lead to more complex, and useful, forms of AI in our daily lives.
However, this will also increase demand for the human to check the work of AI labeling services.
Crowdfund Insider: You’ve talked about HMT task designations using something called “proof-of-balance” in addition to proof-of-humanity. Is this just some sort of algorithmic parsing of workloads? Can you elaborate a little bit on its use?
Harjyot Singh: Proof of balance refers to the mechanism by which the Protocol ranks contributors to determine the likelihood of good behavior, accurate answers and, therefore, priority of payment.
It does not, however, stand alone, but operates as one aspect of a formula to determine priority of users in receiving tasks and payment. Other factors include their reputation, and their past work.
Crowdfund Insider: In terms of ethical AI, when you say that your machines are doing a lot of task-mastering, would you say that it’s true that the big-picture stuff, the discrimination of content, the problem-solving – is still being done by humans?
How much more do we need in terms of guardrails for developing AIML?
Harjyot Singh: It depends on the context, on the kind of data being labeled. Machines are labeling huge quantities of data, but humans remain important. The vital point to make is that humans will be needed for the foreseeable future; once ML masters one form of work, it simply moves HUMANs up the value chain to complete more subtle, complex, and human-centric work, such as emotion recognition.
When machines eventually do that comprehensively – and this is a while off – HUMANs will move onto the next thing, and so on. This is how we get truly brilliant AI services.
Crowdfund Insider: What steps is HUMAN taking on a tangible level to implement Proof of HUMANity (PoH) into the mainstream?
Harjyot Singh: Last year, we successfully integrated PoH into an instance of MetaMask. This was developed as a standard, openly available through our GitHub, that other ERC-20 wallets can use it to implement PoH into their own solutions.
That’s really our main goal: to make it easy for any project to integrate with PoH, and to deliver its unique bot-blocking benefits to their applications and networks. Every implementation of PoH also adds to the aggregation of data.
For each user validation, data – such as an image – is labeled, and that contributes to the detail and diversification of the datasets we make available on-chain. For Layer 1s, this also significantly increases on-chain activity and transaction volume.
The adoption of PoH is an effort foremost led by the HUMAN community; we created the grants program – with an allocation of $10 million – to support other projects in designing their own PoH integrations and bringing it to more networks, wallets, NFT auctions, etc.
We are also funding a lot of other implementations through our active DevBounty program (available on Gitcoin).