Google’s introduction to machine learning (ML) course begins with a brief video with Peter Norvig describing what he learned about ML since joining Google in 2001 as their Director of Machine Learning.
Norvig describes the important benefits of ML and its ability to simplify tasks for programmers. No longer do they need to think of every conceivable rule to ensure the correct outcome – the machine can learn how to derive the correct outcome on its own.
“…There’s a philosophical reason machine learning changes the way you think about a problem.”
Whereas software programmers are trained to use logic and mathematics to prescribe what a program will do, ML allows the focus to shift to experimentation and observation, allowing the programmer to think like a scientist.
To effectively task ML, we will need to be creative and critical thinkers. We will need to know the right questions to ask it and the right data to feed it.
Machine learning can be used to mix and match design elements. As an experiment, Norman Di Palo, an artificial intelligence and robotics student, constructed a small database of chairs. The machine was able to learn to sort the chairs into similar sets and then combine the notable elements of two chairs into an “experimental” chair that could inspire a new design.
Using the same process, Di Palo was able to have the computer take colour and design elements from two patterns and combine them into a third pattern. Imagine an interior designer attempting to find a couch pattern that is different from but still congruent with the patterns already in the carpet and drapes. The designer could have the machine present candidate patterns automatically.
The movie Morgan is about an artificially created humanoid. IBM used Watson to help assemble a trailer for the movie, by using machine learning to identify the pivotal moments of the movie. Interestingly, it still took a human editor to put the actual trailer together.
All machines are intended to increase human productivity. Effectively harnessing the abilities of machines (especially learning machines) is a skill – one we owe it to ourselves to learn.