Before you go, check out these stories!

Hackernoon logoWe're collecting AI problem statements to crowdsource solutions to data scientists by@FrederikBussler

We're collecting AI problem statements to crowdsource solutions to data scientists

Author profile picture

@FrederikBusslerFrederik Bussler

Democratizing data science.

As technology penetrates every facet of life, and continues to grow exponentially, the solution potential becomes enormous. At the same time, we're in a world where billions live in poverty, and millions are on the brink of famine. In order to support an ever-growing populace, we need to leave no stone un-turned in the search for solutions. AI provides many potential solutions to humanity's greatest challenges.

"AI" is a vague, even confusing term. If you hear the phrase "artificial intelligence," you might wonder why there aren't sentient robots walking around, or why everyone isn't in self-driving cars already. The reality is that "AI" is just a marketing term for a set of computational statistical tools, or more simply, algorithms.

However, as versatile as mathematics is, so is AI. AI is limited by (primarily) a couple things: data and computational power. Both the data and the compute power we have available are growing exponentially, so AI is becoming more and more powerful.

With this increase in data and computational ability, AI is now being used in a wide variety of applications.

For example, bitgrit (disclaimer: I'm CEO), collects meaningful AI problem statements to crowd-source solutions to data scientists. Some problem statements include saving animals’ lives, increasing agricultural yield, and speeding up healthcare claims processing.

Michael Suttles, CEO at Save All The Pets, explains how data and AI can be used to save shelter animals:

“I'm planning to use data science to save animals' lives. Long story short: I’m going to collect data about which dogs are more likely to be adopted in which locations, and then relocate dogs to where they’re more likely to be adopted. For example, if Terriers are more likely to be adopted in Dallas than Houston--or older dogs are more likely to be adopted in Chicago than Raleigh--we’ll move them there.”

More specifically, AI models such as NLP can be used to analyze the text description of an animal in a shelter, and Convolutional Neural Nets can analyze images of the animals and help determine the probabilities of animals being killed or adopted in different shelters.

Dog breeds are a genetic mechanism - not visual, so you can't know for certain what a dog's breed is just by looking at it. Therefore, creating a Convolutional Neural Network to predict a dog's breed is harder than it sounds.

And CNN's are data hungry, as are most neural networks. Given that there are literally hundreds of dog breeds, and we need data for each breed, we'll need tens of thousands of labelled images. Fortunately, bitgrit was given such a data-set. We gave users roughly 80,000 images of dogs, with their breeds attached as labels. While a CNN would not strictly be necessary for classifying dog breeds, we'd recommend it to get started very quickly.

What about increasing agricultural yield?

As with the dog breed challenge, the agricultural challenge is highly complex, so it's not as if an accurate model means that we have an end-to-end solution for the real-world, but it is a step in the right direction.

Suvrajit Saha, Founder at Klimazone, explains how AI can be used in agriculture:

“Agriculture is shifting to precision farming methodologies, allowing you to use less water and chemicals and reducing its environmental impact… The first step in gathering information about a specific crop is to identify the field from satellite imagery.”

In this problem, we attempt to use AI to determine contours of agricultural fields, kicking off the pre-requisite steps for precision farming. To add some difficulty into the challenge (as contour detection could be done without AI, for instance with canny edge detection filters), we include crop detection in the task.

Although seemingly completely different, crop detection and dog breed detection are surprisingly similar. Just as users were given dog images with breed labels for the dog challenge, users will be given field images with crop labels for the farming challenge. As before, one potential method would be using a CNN, but this challenge is a lot trickier, as you'll be looking at aerial satellite views (and a lot fewer of them), instead of portrait pictures.

Don't buy into the hype that says AI will have real intelligence tomorrow, or the naysayers who say AI is just a fad. As always, the truth lies somewhere in the middle - AI is just a tool, and it's up to us to use it properly!


Join Hacker Noon

Create your free account to unlock your custom reading experience.