This Week in AI, March 15th, 2018
Building musical instruments with machine learning, opening the black box of neural networks, training models to discover new planets, and more!
Musical Instruments Built with Machine Learning
The Magenta team at Google has been working on the intersection of machine learning and art. Last year they released an AI that plays piano with you. The team just released an open source project for building a synthesizer that combines instrument sounds — the tones and timbre — using machine learning. These instruments are completely novel and greatly expand the sounds musicians can use in their art.
Understanding the Inner Mind of Neural Networks
A major issue with deep neural networks is the opaqueness of their inner workings. In many cases these networks are black boxes, it’s almost impossible to interpret how they make decisions. To address this, the team at Distill developed a framework for understanding how a neural network views the world which you can explore yourself in this amazing interactive article.
Planet Hunting with Machine Learning
Machine learning is continuing to be used in amazing ways across all disciplines—this time, in astronomy. Using data collected from hundreds of thousands of stars by the Kepler project, researchers trained a machine learning model to detect extrasolar planets. The model searched 670 stars and found two previously undetected planets.
DeepMind Advocates for Fairness in AI
It’s becoming apparent that AI systems are often biased, resulting from biased data, incomplete testing, and other issues. As these systems become more prevalent in our daily lives, we’ll need to understand where these biases come from, and how to avoid them. Recognizing this need, DeepMind has created a new research team devoted to fairness in AI.
Applications Are Open for OpenAI Scholars
One way to tackle the “Bias in AI” problem is to ensure that the teams building the AI systems are diverse. Having a variety of backgrounds and experiences in the team helps eliminate blindspots that can lead to biased AI. OpenAI is addressing this through its scholars program, providing stipends for individuals from underrepresented groups to study deep learning.
~
Stay tuned for new updates as we continue to review all that’s new in the world of AI! And if you’re interested in mastering these transformational skills, and building a rewarding career in this amazing space, consider one of our Nanodegree programs:
