Effect.AI Mechanical Turk Demo with Charlie Shrem
On February 28, Chris Dawe and Jesse Eises of Effect.AI had a video call with Charlie Shrem to discuss the project’s progress. Of course, a short update on the Effect.AI Whitelist was part of the conversation, but more importantly, Jesse showed Charlie a short demonstration of the Effect Mechanical Turk that will be launching in 2018.
Charlie Shrem, a veteran in the field of blockchain and decentralized technology, recently joined Effect.AI’s Board of Advisors to aid in development and the acquisition of strategic partners. Read more about Charlie in this update.
The Call
The following transcript has been edited for content and clarity:
Charlie Shrem: Hey guys, it’s Charlie here.
Chris Dawe: It’s Chris and Jesse Eises from Effect.AI. How are you doing, man?
Charlie: Hey Jesse, hey Chris, how’s your day going so far?
Chris: It’s been good. You’re in New York City, right? How did that go?
Charlie: New York went well. I gave a nice talk and I did the money show. At the first one, they gave me a small room where you could see that even with just 30 people it was too small, it was like standing room only. They fixed it by giving me a room where there were a lot more people. I think there were a few hundred people in the audience, for sure.
Chris: Sounds great.
Charlie: It went well. How did the launch of the Whitelist go?
Chris: Yes! That was a few hours ago. Obviously, it was the middle of the night for you, but it went fantastically well. All the security measures that our team had in place went off without a hitch. We had a big spike in traffic — about fifteen or twenty thousand jumped on the website at 9:00 in the morning — but still everything went smoothly. The dev team told me everything would go smoothly, and it totally did.
Jesse: At 09:00, there were thousands of people entering the system. It was exciting, but everything went well: we had sub-second response times, quick page loads and not a single problem. It was great.
Charlie: I really like the concept that you guys are working on. You are bringing in all these people, doing the KYC, essentially enabling them to participate in the Token Sale in the fairest way. Compare that with a lot of projects that will sell out in seconds and have like two or three whales buying up the whole thing. What you’re doing is the best way to do it. It’s just a better system.
Chris: This space is about distributed technology, right? We felt the distribution of the tokens that will be used in our system should be the same. We had a lot of community members shouting at us, saying things are unfair in this space. Then we said, “how do we not only distribute this evenly, but how do we make this as fair as possible?” That’s where we decided on the 25,000-euro cap. And that cap comes into play after what we have dubbed “Fair Share Distribution”. What that means is this: we take our hard cap and divide that by the number of qualified Whitelist participants. We then give our participants the opportunity to purchase a fair share distribution of this. If they take part, great. If they don’t, that’s fine. Whatever is left over, we’ll allow people to purchase in bigger allocations.
Charlie: That’s the fairest way of doing it — When I was first introduced to you guys by John over at Beetoken, what really caught my eye and what really excited me most (knowing that you’re building a whole decentralized ecosystem) was that you guys are building Mechanical Turk in a decentralized system. That’s where it got exciting for me. I really think building a system like that can go towards realizing my ideology of helping to change the world by enabling millions of people around the world to do tasks, to work, and to get paid for them. How’s that part of it going?
Chris: Very well. We’ve been working around the clock. Obviously, we are not an ICO marketing firm, we are a tech company. We’re from tech backgrounds and we build technologies. There’s been a big learning curve with designing the token sale, distribution and the whitelist and related projects, but behind the scene we have developers and a large group of guys that are working around the clock on the technology we want to create. So, and this might come as a bit of a surprise to you, but we have a prototype we want to show you today. This will probably please our community as well, because they’ve all been asking about it.
If Jesse doesn’t mind sharing his screen and showing you a small walkthrough, you can get an idea of where we’re at. This is not a prototype yet, it’s an MVP that we’re working with. We would love to show you though if you’re willing to take a look.
Charlie: Let’s freaking see it! I’m stoked.
Jesse: Ha-ha. Alright, I have it here.
Charlie: Would you look at that…
Jesse: In addition to the Whitelist and the actual Token Sale, my team has been working hard at making a working prototype of our phase one product. We’re about to share with you a small impression to give so you an idea of the look, the feel and how it will work.
Of course, people can log in. For this demonstration, we’re doing that from a Worker’s perspective. A Worker on the system will have to log in like this:
I’ve set up an account. This will work like any normal login screen: after logging in, you end up in the system’s dashboard.
Immediately you see the list of tasks that are available on the system. For now, it’s just dummy data, not real tasks. But this screen provides a list of everything that’s available.
Charlie: Is it so that the more data Workers give about themselves, the more tasks they can complete? Because the more demographic and geographical data you have with the person, the better it is. Of course, a Worker can decide for him or herself how much data they want to share, I assume?
Jesse: Definitely. The most important thing is the reputation score of a Worker. It will determine how many tasks you will be able to accept. A Requester on the system that’s creating jobs, can decide a certain reputation level that a Worker must have to qualify for a job. The better your reputation score is (and at a later point also depending on the quality of the profile information you supplied) the more jobs you will see, and the more things you can do on the Effect Mechanical Turk.
Right, now I will accept a task and do some work. I joined the first task here.
In this prototype, we have the image classification task type where you see an image that you’ll have to classify. This means there is a short description about exactly the task that must be done. In this case, we’ll have to select two categories for this image (with a maximu of three) and we will earn a total of 150 EFX in this example case.
Let’s check out the options we have.
We see a small market with people, so let’s just go with “pedestrians”, “small store” and “market.”
That gives us three, which should be enough to proceed to the next question.
This is a very basic type of task. It’s an image that Workers enrich with data consisting of semantic information and a classification. By doing this, we can earn rewards.
For this picture, let’s go with fruits.
This was just a quick demonstration, but it should give you some idea of what The Effect Mechanical Turk will look like and how Workers will be able to interact with it.
Charlie: How do you prevent people from gaming the system? For instance: putting in data that’s not correct. Or trying to game the system by using artificial intelligence, how do you discourage that?
Jesse: That’s one of the most important questions we (and any Mechanical Turk) are dealing with. There can be fraud in the system. You can earn money doing tasks, so you can do a very poor job and still earn money. We do have to counter that, however. There’s multiple ways of doing this. One thing we’re doing to combat this is by having redundant tasks on the system, meaning one task will be completed by multiple Workers. By averaging the results, you will get a more stable classification.
Charlie: That’s genius. It allows you to see the outliers.
Jesse: Exactly, you see the outliers. And you can also see which Workers performed badly in the system and give their reputation a loss. This discourages them by lowering their reputation score. Eventually, bad Workers will be pushed out of the system because they perform worse than others and have access to fewer tasks as a result.
Charlie: That’s what the reputation point is that I see on the top there.
Jesse: Precisely, you see your reputation, the rewards you’ve earned and there’s going to be an account page here.
That’s basically it. We will release this in the near future.
Chris: We can’t promise anything, but we’re looking at a full working prototype by the end of March.
Charlie: You’re already further ahead than most other projects.
Chris: We’re not a whitepaper project. That’s not what we wanted to be. We wanted to show our community, our supporters, and everybody involved, that we build technology. There’s been some questions about the age of the team, but I promise you: we can compete with any platform.
Charlie: Hey, I’m just 28 years old and I feel like I’ve accomplished more than people three times my age.
Chris: Exactly! Not to name-drop a competitive platform, but look at Vitalik Buterin: 18 years old, a kid essentially, but absolutely making a name for himself in this space. Look at Steve Jobs! look at all these guys! They did it when they were young. I’m sure everybody had similar questions when they were first starting and look where they ended up. The same goes for our team at Effect.AI. These guys and girls are driven, ready to compete, and convinced they can be at the forefront of this technology.
Charlie: I’m super proud of you guys. You’re doing a killer job and I can’t wait to continue doing these live updates for all the people that are watching. Talk to you guys tomorrow, have a good day!
Chris: Yeah thanks, man! See you soon and take care.
Jessie: Nice talking to you, Charlie!
If you want to see the full video, you can follow this link: https://www.youtube.com/watch?v=L0vGd8SWwlw&feature=youtu.be
Effect.AI Whitelist
As of this moment, the Effect.AI Whitelist is still open for registration. Head to the website for more information on the project and detailed instructions on how to participate.
Effect.ai/
T.me/effectai
Facebook.com/effectai
Twitter.com/effectaix
Github.com/effectai
