Modern companies collect tons of data about retail transactions, marketing touchpoints, consumer feedback, and more as an organic part of their day-to-day operations. It’s easy to gloss over the challenges of procuring enough data for machine learning, but in reality insufficient data is a common setback in data modeling and developing artificial intelligence. In many cases — like the ones we’ll explain below — outsourcing data collection to the masses has been the answer. That’s why crowd-sourced data and artificial intelligence make such a perfect pair.
Waze users train the app’s algorithm to find the fastest path to your heart
Widespread adoption of smartphones with GPS created an opportunity to change navigation forever by giving drivers access to real-time traffic information through platforms that could continuously optimize routes along the way. Adapting to such dynamic change requires tons of data and user feedback, which 50 million Waze users contribute as they use the app. Submitting a report of an accident is an obvious example of this kind of feedback, but even actions like failing to make a recommended turn are considered possible indicators that something’s amiss and a route should be reconfigured.
What’s more, the company leverages its network of feedback to give back to the communities of its users. Information from Waze has been used across the United States to improve crisis responses during severe inclement weather and to address threats to traffic safety internationally. Whether ensuring your timely arrival for Valentine’s day reservations or informing local governments of high-priority streets for maintenance services, Waze’s uses of crowd-sourced data to support its technology are easy to love.
Everyday food critics make Yelp an expert on date night
Yelp is an obvious case study for the benefits of crowdsourcing opinions to help inform user decisions, but the company’s use of AI to make reviews more reliable is considerably less apparent. For this initiative, Yelp relies on the millions user-inputted photos of locations and the food or products they offer to power image recognition models that can tag key attributes of each establishment.
Ideally, training models based on images rather than the content of the reviews can help reduce the impact of individuals’ biases in Yelp’s recommendations. It works by analyzing images to make determinations about a restaurant’s ambiance, the kind of food served there, or whether it’s accommodating for children or pets. That means Yelp can help you impress your date with a restaurant choice that even an artificially intelligent system knows is the epitome of class.
Listeners teach Spotify how to set the mood
Spotify’s recommendation engines actually use a combination of methods to suggest music to its listeners. Crowd-sourced data plays an especially important role in a method called collaborative filtering, which analyzes a listener’s behavior and compares it to everyone else’s to make recommendations based on the preferences of people with similar taste.
Spotify listeners contribute data both actively and passively to make collaborative filtering systems more and more effective. When the app recommends music, users can tag whether they like or dislike Spotify’s choice, and even specify whether it’s the artist or the song that didn’t resonate with them in the event of a flop. However, relying on the listener to note their preferences limits how much data Spotify can use to keep improving its suggestions, so passive listening data is an even more important way you help Spotify share music you’ll love in your Discover Weekly playlist every time.
Ultimately, the best evidence that crowd-sourced data and AI are such a great match is that they never stop improving each other the more they communicate back and forth. To learn more about how humans can help AI-powered bots get better at the services they provide, check out the chatbot example in this article featured on VentureBeat.