As ML practitioners build, evaluate, and deploy machine learning models, they should keep fairness considerations (such as how different demographics of people will be affected by a model’s predictions) in the forefront of their minds. Additionally, they should proactively develop strategies to identify and ameliorate the effects of algorithmic bias.The self-paced ML learning course helps you identify the types of human bias that are found in machine learning models, and teaches you how to identify potential areas of bias, as well as how to evaluate ML performance while accounting for bias. Google’s course naturally focuses on training models, which is where most errors or biases occur. 'Fairness' fits in the middle of Google’s ‘crash course’ on Machine Learning. The first part, ‘Concepts,’ is a more general overview of how machine learning works, and best practices for training a model; it’s the nuts and bolts of ML. Fairness slots into ‘Engineering,’ which focuses on interference and dependencies (and now, eliminating bias). The third part of Google’s course dives into real-world examples of how machine learning is making a difference (such as cancer detection). If you’re not sure how bias in ML may actually affect models, look no further than Amazon. Sources say the e-commerce giant's experimental ML tool for recruiting began to accidentally prioritize men for jobs. The tool’s algorithms analyzed 10 years’ worth of résumés and assumed that, because the majority of applicants were male, males were preferable hires. As a result, it began de-emphasizing any mention of “women” or “women’s” in résumé submissions. Amazon later shut that tool down. Google’s TensorFlow, an open-source machine-learning framework, is the third-most-popular repo on GitHub, and the most popular dedicated machine-learning repo by a country mile. Dice data shows TensorFlow is one of the most in-demand skills for machine-learning developers and engineers, and the most useful unique skill (one relating specifically to a framework) for ML devs. Google also rolled out on-device machine learning for Android via ML Kit earlier this year.
Google loves machine learning (ML). Now, it’s launched a new course module that aims to help you, a human, recognize your own bias before training ML models. Named ‘Fairness,’ the course is 70 minutes on how humans are compromising machine learning models. From Google: