Filter by tag

And now, the winners of our first AI For Everyone Challenge

A few months back, we announced our million dollar AI for Everyone challenge. Our goal was to award researchers, academics, and, non-profits who wanted to make a real difference in the world with artificial intelligence. Dozens of submissions later, we’re thrilled to introduce our first two winners: KivaFaces and DeepCell.

Now, we’ll be getting into each of these projects in fuller detail next week, but we wanted to give you a high-level look at what these teams will be doing over the next year. Let’s start with KivaFaces.

Kiva, as you may know, is a micro-lending platform focused on alleviating poverty. Borrowers from over 80 countries ask for loans covering everything from shelter and education to farming implements and food and lenders loan–not donate–the money to help. Over a billion dollars has moved through Kiva since their launch roughly a decade ago and they’ve funded well over a million loans.

So where’s the AI come in? Well, part of the Kiva process is that a borrower uploads a picture of his or herself. This process has created a rather immense store of diverse borrower images from around every corner of the globe. But if you run typical facial recognition algorithms on these images, the reality is, they’ll miss a lot of these people. As in, those models simply don’t see the borrowers the people. They don’t recognize them. That’s because, primarily, these models are trained on more homogenious (namely first-world) datasets.

Melissa Fabros’s KivaFaces project is looking to change that. They’ll be annotating these images through CrowdFlower in order to create a dataset to train more inclusive algorithms. It will also help Kiva itself approve loans more quickly so that lenders around the world can get helping faster.

DeepCell is another computer vision focused on people. Except DeepCell, as you might guess from the name, is concerned not with faces, but microscopic images.

Disease researchers spend of lot of their time looking over cellular images. They follow individual cells or groups of cells, analyzing their behavior over time to evaluate how a disease works, the efficacy of potential cures for a disease, or a whole host of other important factors.

This is time-consuming work. But DeepCell’s founder David Van Valen has already demonstrated that AI and deep learning systems can speed up this analysis. The problem is there simply isn’t enough quality data to use. And that’s where CrowdFlower will come in. DeepCell is looking to annotate cellular images to create a library of training data that can be used to perform experiments and analysis and, hopefully, help researchers make important discoveries about disease behavior.

We’re thrilled with both AI for Everyone winners and will be enthusiastically sharing their progress over the next year. Both KivaFaces and DeepCell will get full access to our platform, $25,000 in contributor costs, and CrowdFlower support from start to finish. And, since this is called AI for Everyone, rest assured these datasets will be available when they’re labeled and annotated.

While they’re our inaugural winners, they won’t be the only ones. In fact, we’re hosting a quarterly AI for Everyone Challenge. If you’re a researcher, academic, or non-profit interested in leveraging AI to improve the world, you can apply here for consideration. If you’re applying, please keep in mind that while both KivaFaces and DeepCell were computer vision projects, those aren’t the only initiatives we support.

At any rate, one last big congrats to David, Melissa, and their teams! We’re looking forward to following along for the next year and we’ll be sharing their big milestones along the way. After we do a proper introduction next week, that is. See you then.