There is a mantra in technology that if a program is fed bad data, it will produce bad results. The same is true of AI. The quality of a machine learning application is directly related to the quality of data it is fed.
Sébastien Boria, mechatronics and robotics technology leader at Airbus, says an AI system that takes a sample of all the people who work on a particular manual process, to understand how they work, may well come up with a machine learning model based on the average of the people sampled. He says: “Just because 50% of the population does something, does not make it the right solution.”
Statistically speaking, population samples generally follow the so-called Normal distribution, where the majority of people are at the top of the bell curve, and 5% at the tail ends of the curve. This may be fine for discarding anomalous results and one-offs, but when applied to machine learning, Boria believes the top performer’s results “may be seen by the machine as an anomaly.”
Miracle on the Hudson
A decade ago captain Chesley Burnett Sullenberger landed US Airways flight 1549 on the Hudson river in New York after both engines on his Airbus A320 failed. In the FAA transcript of the cockpit recording, air traffic control guided Sullenberger to make an emergency landing at Teterboro airport. He replies: “We can’t do it….we’re gonna be in the Hudson.” There is no flight operations manual on how to land an A320 on a river, but Sullenberger did it, saving 155 passengers and crew.
Industry 4.0 needs quality data
The heroic actions of Sullenberger is just one example demonstrating how humans can think outside the box. It is often necessary to work around normal procedures in order to achieve extraordinary results.
AI and machine learning were among the hot topics discussed at the recent World Economic Forum in Davos. Many believe AI technology will be the fuel that powers Industry 4.0. This is not necessarily about fully automated manufacturing. Instead, relatively small changes can be scaled up, significantly improving operational efficiency. But, as Airbus’ Boria has noted, basing machine learning on the average of how people work, results in average AI. Industry 4.0 needs extraordinary AI, and this means capturing exceptional data.