Stay up to date with the HUMAN project through our bi-monthly newsletter.
Five ways HUMAN Protocol can improve AI
Unavailable to most. Narrow in scope. Siloed with large companies. Unrepresentative of the diverse world we live in.
These are some of the problems with the data used to create AI products. AI is behind where many thought it would be; it seems generalized AI has been twenty years away for the last sixty.
HUMAN Protocol’s unique solution can help to provide the data that is needed to inspire a new generation of AI products.
To create truly useful AI services – both those which are very specific, such as a surgical robot, and those which are more generalized, such as personal assistants – scientists need specific data.
It may sound incorrect to claim that generalized AI needs specific data, but generalization only comes about once many specific uses are perfected: when a personal assistant can do the specific tasks of cooking, vacuuming, and washing up, for example.
HUMAN Protocol is a platform to power any kind of data-labeling API; once the API is looped-in to HUMAN Protocol, it can benefit from the expanded work pools and network of contributors it provides.
Not only can any kind of tool be used to create data, any number of workpools – to the scale of many millions – can be brought onto HUMAN.
Why is volume important? When it comes to machine learning datasets, quantity becomes a quality of its own. Not only does HUMAN access the hCaptcha workpool of hundreds of millions of data labelers, it allows any other workpool to loop into the software infrastructure. It sets the foundation for the largest workpools ever seen
In this, HUMAN is not only a competitor to service providers such as MTurk and Scale.AI. These services can loop-in their own APIs and workpools to add their already vast pools.
The above benefits both require a degree of work, even if the process is made easy. That said, HUMAN Protocol has already integrated with major off-chain and on-chain work pools, including hCaptcha, which provides voluminous data to ML practitioners. These are continuously expanding, along with the number of workers projects have access to through them.
But why would a scientist use HUMAN Protocol over competitors such as Scale.AI or MTurk? Beyond the possibilities of looping in the specific APIs they may require, there are many outlined in the two articles comparing HUMAN with Scale and MTurk. Namely, the current benefits include best-cost and quality allocation of funds, automation, and on-chain payment.
HUMAN Protocol is permissionless: anyone can request work through the Protocol, and benefit from the solutions it provides.
One of the inhibitors associated with the slow progress of AI is the limits data scientists have to data-labeling services. Most data-labeling services do not necessarily prohibit data scientists from requesting work, but rather, by being centralized, offer only a limited range of kinds of data services.There is a lot of data out there; but the ability for scientists to curate their own, specific data sets is very limited (and costly), such that scientists are often left to label the data themselves.
This relates to the ability to curate specified AI. HUMAN Protocol changes this; it provides democratic access to data with a price-quality guarantee, ensuring not only the relevant detail but the capability to scale to requirements.
By allowing anyone, of any budget, access to such a data-labeling service, HUMAN Protocol allows more voices to be involved in the creation of AI products. It brings democracy and meritocracy to an industry which has been very niche, siloed, and for the few large companies that can own data, run labeling services, and curate their own datasets.
More scientists, startups, and perspectives can offer a more diverse generation of AI products. Not only can meritocracy ensure that the best scientists get the data they need, but it can bring more voices into the industry. Just as quantity is a quality when it comes to the sample group, something similar could be said of the scientists who create the products themselves.
For more on the democratization of AI by HUMAN, the basics of bias, or the intricacies of bias, read our blog.
Such scale is tempered by intelligent software that automatically finds requesters of work the best possible deal, when it comes to price and quality. Composability between networks benefits practitioners in a number of ways:
By bringing together different tools under one Protocol, the ability for these applications to communicate and share their efforts communally can bring about a new era for AI services. Through it, the request for work changes dramatically; because the work that can be requested is now an accumulation of all the kinds of work available on the network. A request for work could be an entire document, with photos sent to one application, text to another, videos to another.
Together, the combinations of tasks result in infinitely more complex and useful data sets which can go towards building infinitely more sophisticated Ai products; through such a system, specific AI services are combined to facilitate generalized ones.
For the latest updates on HUMAN Protocol, follow us on Twitter or join our Discord. Alternatively, to enquire about integrations, usage, or to learn more about HUMAN Protocol, get in contact with the HUMAN team.
The HUMAN Protocol Foundation makes no representation, warranty, or undertaking, express or implied, as to the accuracy, reliability, completeness, or reasonableness of the information contained here. Any assumptions, opinions, and estimations expressed constitute the HUMAN Protocol Foundation’s judgment as of the time of publishing and are subject to change without notice. Any projection contained within the information presented here is based on a number of assumptions, and there can be no guarantee that any projected outcomes will be achieved.