The term “Mechanical Turk” comes from an 18th century chess-playing automaton built by a Hungarian inventor named Wolfgang von Kempelen. The Turk was presented as an automated machine displaying a high level of artificial intelligence by beating famed chess players of the time, including Napoleon Bonaparte and Benjamin Franklin. A panel of the machine was exposed to reveal an intricate system of gears and cogs that supposedly powered the machine. In reality, the “machine” was large enough to house a chess master who had the ability to control the human-like figure that made the chess moves. For 84 years the Turk toured Europe and the Americas until it was destroyed by fire in 1854.
This was the name chose by Amazon for the roll-out of their marketplace for the automated utilisation of human intelligence. While AI, machine learning and neural network improvement is at the forefront of the majority of modern research and development efforts, there is still necessity and room for human input and opinion. A machine cannot emulate the thought processes and imperfect logic of a person yet, and possibly not ever without reaching a true level of artificial intelligence. Micro-tasks are completed by people, for organisations, research teams or any authorised group that requires the input of, usually, access to a large sample of real people. While there are tasks that require writing, mathematical skills or creativity, there are also those that require a lot of feedback from as many human sources as possible to build up a realistic idea about peoples impression of, or to properly categorise something. Tasks that are available to large groups of users also help improve deep learning and AI systems, as the responses of a large group of people can be used to train or validate AI’s outputs when it comes to things like recognising things in photos.
Amazon achieves this best via their AWS Mechanical Turk API services, allowing anyone to sign up as a worker and complete tasks as defined by a registered and authenticated requester. Through the M-Turk API and interface, requesters are able to set HIT’s (human intelligence tasks), set their price for completion, qualify the workforce to ensure the right people are completing the tasks, retrieve and approve completed work and pay for tasks that have been adequately completed.
Crowdflower is a popular alternative, focusing on giving companies with a large group of human testers to provide sentiment analysis, search relevance or business data classification. These are all methods to train and test the effectiveness and accuracy of different machine learning algorithms when it comes to human sentiment, the relevance of search results or image annotation and classification.
Many inefficiencies exist within the current microtask industry, meaning that employers are overpaying hugely, but workers are not seeing reliable wages that can truly be counted as a sustainable income. This is largely because HIT’s flow through centralised platforms such as AWS M-Turk, Crowdflower or a host of others that will charge up to 40% in fees for each task completed. Due to these inefficiencies and the dwindling incentives for those completing tasks, between five and fifteen workers are often charged with completing the same task. This has been called “consensus by redundancy”, and highlights the fact that work must be completed by a much larger sample size of humans than is necessary, simply to ensure a certain level of reliability. Low pay, high fees and repetitive work does not stimulate an effective workforce of motivated task completers. Access to these services is also restricted to those with a specific bank account, usually restricting parts of the world or certain countries entirely.
The GEMS protocol is designed as a decentralised foundation on which micrtotasks can be distributed and completed, without a lot of the inefficiencies of the traditional methods. Gems does this in three main ways, by automating task verification, trust, and payments. Consensus by redundancy, or the need for more people to complete a task than is truly necessary is eliminated by people staking their token based currency on the network, as well as trust scoring mechanisms that ensure task accuracy and timely completion. All that is required to use the Gems protocol is an internet connection and some basic computer skills. With companies like Google working on affordable, universal smartphones or Facebook working on delivering internet to wide areas via drones, the services that run on the internet need to be able to scale down to the huge job and expertise market that becomes available in places that were typically too poor or remote to be relevant to these industries.
Gems allows for overall objectives tasks to be broken down into smaller HIT’s that can be reliably resolved by a small number of individuals. This is achieved via the three separate levels of Gems as it currently stands. Firstly, is the Gems protocol, which is the main logic behind the technology that is used to determine whether or not the a potential task participant is trustworthy and reliable. The protocol is just that, and as such can be used as a platform for the development of Gems based applications. The second level is the actual Gems platform, which is the main site, interface and technology that is used to pair requesters with those who wish to complete tasks. This part would not be too dissimilar from existing centralised solutions, were it not based on the Gems protocol to begin with. The third level are Gem modules, which are open-source re-usable task interfaces that run on the Gems platform. Dapps (decentralised apps) allow any organisation, company or individual to build apps based on the Gems infrastructure to list their tasks. Users can log into any app or listing site using the Gems login, see and complete available tasks (depending upon their trust rating!). Payment happens in the Gems tokens, meaning that there is no need for users to repeatedly enter bank account details, or to even have a bank account.
Gems is extensible, allowing for further development of decentralised apps on the protocol to build additional verification methods or reusable tasks. A lot of industries are being disrupted by blockchain startups, with the good ones looking at the stagnant and worst parts and attempting to come up with realistic, practical solutions. Unlike a lot of other ICO’s, Gems has looked at an industry that is beginning to struggle under the weight of its inefficiencies and suggests in itself some practical solutions that can make the process a lot simpler for those using the service as a task assigner or assignee. By staking and incentivising workers to complete tasks as a part of their presence on and value within the platform itself, issues around consensus by redundancy, slow payments and unreliability are eased. By having users put Gems up front when they agree to take on a task, there is a huge incentive to complete it to a high standard and in good time, in order to not only earn new Gems, but to not lose Gems that they have staked to demonstrate their trustworthiness.
The two founders are brothers Rory and Kieran O’Reilly. Both studied at Harvard, have founded the site gifs.com and have been recognised in Forbes 30 Under 30. The team of advisers include names that have created and run successful startups and businesses, and not just crypto enthusiasts! Biz Stone is one of the co-founders of Twitter, Ben Maurer is the co-founder of reCAPTCHA and works for Facebook. Joey Krug, Luis Cuende and Joe Urgo have all founded blockchain startups and work as or have worked as traders. Information regarding the extended team is limited at the moment.
Although no formal ICO date or token distribution information has been released, there is a lot of speculation already based entirely off of the project, the tech and the team. Community program applications shut down in a couple of days, and the ICO probably won’t take place until late January / early February at the earliest. There have been some accusations around the shady nature of having to provide content in order to become eligible for the GEMS ICO whitelist. It is a smart way of creating a lot of content and buzz, but really risks a huge backlash if the product does not live up to the hype. For full disclosure, I will be submitting this to the whitelist applications. Whether it will be deemed enough to get through though I don’t know. At this point I can’t say that I’m confident enough to invest, but I would like to know what more I can by applying for the whitelist, and seeing how it goes.
*Update: I was accepted onto the GEMS whitelist. I did not participate.