You can help drive the future of artificial intelligence (AI) by labeling content like images and videos.

 Telegram Group Links | Join, Share

How is labeled data used in artificial intelligence?
Let's identify the cars in this picture below:

Course

Let's identify the cars in this picture below:

You'll point out the cars and give them a "car" label. With enough data, AI systems “learn” from the labeled data in order to make predictions and decisions. This leads to incredible things, like driverless cars!

Tas is when you point out objects of interest in the picture, video, or 3D scene, and give a description of what the objects are. On AI, you can get staffed on a project and get paid to accurately label data.

How to Get Work (1/2)

Each project is different! The things that make each project unique include:

What kind of data are we labeling? You might see pictures, videos, and even objects represented by points in 3D space, sort of like a video game.

How many different labels are we finding and pointing out? Sometimes, we are looking to label just a few types of things, like "person" and "car." Sometimes, we are looking for more types of things, like 5+ types of cars and 5+ types of roadway obstacles. Other times, we are labeling everything in the scene.

How complex are the rules for when to use each label? Sometimes instructions for a tas are short and simple. Other times they are longer and have more complex guidelines on how to handle each situation differently.

Each project will be different. That's why we provide a Training Center with project courses, which teach you how to label each project correctly. Each course also tests your understanding of the rules and requires that you show your project skills before you can start tas - and earning money - in that project.

How to Get Work (2/2)
Courses are your key to getting work.

We want you to do the task right (which also helps maximize your pay!). Each project has a different set of instructions, so doing the task “right” can mean something different from project to project. Some projects are very similar to each other, but they still have their differences and it is very important that you follow the specific project rules. Courses teach you how to do each project. They also assess how well you understand the instructions.

Once you pass the courses for a project, you will be able to task and earn money on that project!

How do I start tasking and earning money?

Go to the “Training” tab in your Aiaccount.

Pick the project type you want from the options in Content and Language, 2D, and 3D.

Pick the project (for example, “Llama Cuboid”) you want to learn and complete the courses for that project to start tasking! Note that the project type you picked might have a “tools 101” course that you must complete before unlocking the project courses. These 101 courses help you learn the Ai interface so that you can task quickly and with high quality!

How to Get Paid
IMPORTANT: To receive payment, be sure to add your payment method (for example, PayPal) in your
When do I get paid?

Once you start doing tasks, your earnings will start flowing in after a couple of days. Once you are tasking regularly, payment will appear once a week. Here's why:

Task pay will be higher when your labels are higher quality (meaning, they closely follow instructions). This is why we calculate pay for each task carefully. Sometimes, we can calculate the pay for the task very quickly. Other times, it may take a few days to calculate the final amount. Generally speaking, more difficult tasks (such as LiDAR tasks) require more time - but those tasks also pay more. This is because we may ask another tasker — a reviewer — to review your task and fix anything you missed in the instructions. We calculate your pay once this review process is complete.

We trigger everyone's payout once a week. We do this on Thursday midnight UTC time. The cutoff is Tuesday midnight UTC time. In other words, if you did the task before Tuesday midnight, and we calculated the pay for that task, it will get paid out in the payout.

How much am I paid?

When we calculate pay for each task, we consider 3 things:

Difficulty. How hard is this task? A categorization task is very easy and quick to do, but it may only pay a few cents. A semantic segmentation task is harder and pays more per task. A LiDAR annotation task is even harder and takes longer and pays even more. Different tasks within the same type of task can vary in difficulty, depending on, for example, the number of labels that is required.

Accuracy. How good was your work? Did you label all the objects that needed to be labeled? Did you label them correctly? Was your work precise, or was it sloppy? We reward accuracy (eg, following instructions closely) — a 100% accuracy task pays much more than a 50% accuracy task.

Contribution. How much of the "completed" task, or final submission, was the result of work you did? Remember that other workers and reviewers may also add and fix labels on your task if you missed something in the instructions, so you always want to follow instructions carefully to maximize pay.

Last big topic: Security!
Learn how to keep your ai account and task data safe and secure! Everyone on aiis required to complete this security training before starting tas.

The purpose of this training is to help us develop an understanding of security risks and what we can do to tackle them when they arise.