I've long felt that decentralized work platforms are the "future of work", and that blockchain applications were the way to accomplish that. A huge challenge to making that happen is establishing trust amongst strangers: one population who has work to be done, and another that's capable of doing the work. In a perfect world, this would not be an issue -- users would always know what they claim to know, and bad actors would not try to game the system. However, this world is full of imperfect knowledge and motives; much of the work around decentralized work platforms like Mechanical Turk or Crowdflower have done focuses on preventing bad actors from gaming the system. There's been a great cost to doing that; those systems are expensive to use and additionally lock out many qualified workers from using their services, especially in the developing world (unfortunate, because these are the populations who could most benefit from providing work on these platforms).

Gems is a cryptocurrency built on top of the Ethereum network (as an ERC-20 token) designed to address some of those flaws, and make true work decentralization a reality.

A few reasons why Gems is so exciting:

1. Problem space

The Gems whitepaper is informative, and readable over a single cup of coffee. The team has wisely chosen to focus their initial efforts in AI training. For those of you not familiar with current state of art in intelligence programming; a technique called "human in the loop" is commonly used to help train algorithms with tasks like image recognition, text recognition, and more. It works by using human brainpower to label corpora of data so that software has a working set of correctly labeled data. You've been a human in the loop of an AI training regime if you've ever filled out a Re-CAPTCHA, or selected tiles that contain dogs when logging in to a Google powered service. Companies like Crowdflower are primarily focused on human-in-the-loop services for companies employing AI and machine learning.

Let's imagine a startup called Puppr that is training an AI to discover pictures of cute puppies. Puppr could use the Gems module designed for AI tasks (helpfully provided by the Gems team) in their own systems; removing the need for the company to deal with Mechanical Turk or Crowdflower, and allowing any person to participate in training. They could focus on building their application and plug in the Gems module to allow trainers to be compensated for their time.

Another use-case not specified in the whitepaper, but applicable to the token is for crowdsourced fact finding. Last year, I built a product called Docsift, which was designed to help citizen journalists review large document sets in order to find "interesting facts" inside them. While the system is good at extracting hard facts like names, places, and references from the document, it's a very hard problem to infer information from raw text without a person being involved (as an example: some text contains "Alice punched Bob on the shoulder." -- did Alice do something wrong? Well, that depends on the context, which computers are generally terrible at. A human could read the surrounding text and trivially understand the context of that sentence). If Gems existed when Docsift (which has sadly been mothballed) was in development, I would have absolutely implemented it in order to allow document posters to include bounties for facts found, reviewers to get paid, and support crowdsourcing funding of document reviews (this would have been huge).

Once it's available, I will be using Gems on CoinAlerts.io -- I'm tracking 1,600+ tokens today and need descriptions written for each coin, plus validation that the data I already have is correct. I could create a job on Mechanical Turk to do this, but my guess is that Gems users will understand a lot more about cryptocurrencies than the average Mechanical Turk worker. The validation system will be especially useful for this work; I'd worry that someone will enter random text for a description, or I think I can use a validator to ensure the data I have right now is correct (e.g. Twitter handles, GitHub repositories, websites, etc).

2. Fit with Blockchain vs. traditional service

Decentralized work platforms are an excellent fit for the blockchain, especially if there's a way to deal with dispute resolution. Gems has thought this part of the problem out well, with a unique staking mechanism (the full details can be found in the whitepaper starting on page 12 / section 5) and a rating tied to an Ethereum wallet address. Gems' staking mechanism roughly states that users are paid for the work they've done, as long as the work seems legitimate. Job posters can specify that their jobs require verification (say, for text transcription tasks), and users on the Gems network can act as verifiers for those tasks. The verification loop gives workers the opportunities to fix any mistakes or risk a negative rating. Requesters are rated on the Gems network as well -- this creates an ecosystem where everyone can validate the score of everyone else they interact with; requesters, workers, and validators will all need to maintain positive ratings in order to use the Gems network successfully.

There are a few additional especially useful benefits enabled by building a decentralized work platform on the blockchain:

• Allowing users without bank accounts to trade their time for spendable money. Users without bank accounts, or with bank accounts in many countries around the world cannot participate on Mechanical Turk or Crowdflower today. Gems allows anyone with a computer and a connection to participate.
• Crowdsourcing for sensitive topics becomes possible -- a theoretical scenario is that Amazon and Crowdflower decide that pictures of cute puppies are now against their Terms of Service; leaving Puppr in a lurch. Gems would allow them to continue working because it doesn't have a mechanism to control the ways the token is used; only who the participants are (and only to the extent that users have ratings and those ratings control activity on the network). Perhaps less important if you live somewhere with a stable government and trust Amazon and Crowdflower completely; but this is a concern for many people who could use this system.
• No reliance on a company to operate a crowdsourcing service. Mechanical Turk was created over a decade ago, and it looks like it. There's nothing stopping Amazon from turning MTurk off tomorrow; Crowdflower could be acquired or go out of business next week; Gems will survive as long as Ethereum exists and there are 2 parties interested in transacting.
• Crowdsourcing the funding of work. Since this is an ERC-20 token, it should be possible to crowdsource the funding of work; if a decentralized community or company has a need to have work done but don't possess the time or ability, they could pay other people to help out. Using the document review example above, if a dozen people get together and want to pay to get people to help review a 500-page net neutrality bill, they could pool their Gems together to create a work bounty (and even take part in the review to get some of their Gems back, especially since Gems does not take a processing fee for transactions).

3. Team

I don't know much about the O'Reilly brothers (Rory and Kieran) other than what's listed on the Gems homepage. Their biographies look impressive; with a background of actually launching products. Their advisory team is extraordinarily strong -- Biz Stone is one of the founders of Twitter and a serially successful entrepreneur. Joey Krug and Joe Urgo both have successful projects under their belt (Augur and District0x respectively; Joey runs Pantera Capital and they are both advisors to a significant number of projects). It's safe to say they understand blockchain applications better than most. It's easy to imagine District0x implementing Gems in order to provide content policing if needed on any of their communities. Or a system that uses REP along with Gems scores to paint a fuller picture of a worker, requester, or validator.

Gems has the potential to become extraordinarily important; but there are a few risks to its success:

Will the network take off?
Are there enough jobs out there to support this token and the team in order to make sure it continues to grow? Does Mechanical Turk look unmaintained because it's not a significant revenue generator for Amazon? Did Crowdflower pivot from a general work platform to human-in-the-loop training services just to latch on to the AI hype train of 2016-2017? As mentioned above, as long as the token releases, I don't believe this will be an unsurmountable challenge as long as there are some people working and providing work on the platform in order to bootstrap.

Will unranked users be able to join the network successfully?
If requesters only specify the highest ranking workers to perform tasks, how will new users succeed as the network becomes more mature? Ostensibly this will be resolved in an economic way -- new users will end up being paid much less than established users until their ranking improves. Perhaps the Gems team will suggest newer users start out as validators in order to get some initial rank on the system.

How game-able will the system be?
Will it be possible to overwhelm the network with automated posts to low rep-required tasks with the hopes that some of them will either be approved, or not validated? If task payments follow a power-law curve, and the top 1% of rated users make 99% of the money, will people sell their accounts to new users to help them bypass low-rating issues? My sense from the whitepaper is that the ranking mechanism will resolve this quickly, but it's unclear how reactive the algorithm is to a series of bad ratings.

Overall, this is the most compelling project I've seen on the blockchain in the past 12 months, and am very excited about using it in my projects. The fact that the team has chosen to focus on AI applications initially also speaks well about how they're thinking about the problem space -- I don't think there's another focus they could have chosen that would have given them as much chance of success. I haven't seen the mechanics of their initial offering to know whether or not it makes sense from a financial perspective to get in as soon as possible, but it is project I'd like to support somehow because I'd love for it to be successful.