Pramod Dibble's Blog


Smart technologies can’t replace humans - Plane crash caused by overreliance on autopilot

24 Jun 2014 | by Pramod Dibble
Share this:

                Slow news days seem to be characterized by the blogosphere reaching into its bag of social-injustice-topics and gleefully ranting about whichever it blindly selects. On every fifth iteration or so of this charming cycle, in between politicians breaking the economy and the top ten blah-blah-blah, we get to hear about how the machines are taking all of our jobs. While I suppose it succeeds in generating clicks for those who live and die by such things, it contributes very little to our understanding of smart technologies and how they are used in our world, even adding to the misinformation inherent to cognitive computing, robotics, and smart tech.

                On July 6th 2013, Asiana Airlines flight 214 crashed into a seawall on landing at San Francisco International. The plane approached the runway too low and slowly, causing the accident which resulted in the deaths of three of the 303 people on board (one death was caused by a rescue vehicle running over an injured person). Of all the possible scenarios involving the crash of an airplane, a survival rate of over 99% is a testament to the engineering of the vehicle and the skill and training of the crew and first responders.

                The insufficient speed and altitude of landing was set by the flight crew, believing that an automated landing program would engage if the speed fell too low. Autopilot software is able to fly planes from runway to runway, but there is a very good reason to have a human being in the cockpit nevertheless; in a classic-role-reversal, the human being has become the failsafe mechanism for the autopilot. But this plan breaks down if the human being does not have an expert understanding of how the flight software works, and what specifically they, as the pilot, need to watch for to ensure the safety of everyone onboard. Automation software is not a person, and the ways computers learn and execute commands is fundamentally different from what we as humans can compute, a paradigm which will remain for the foreseeable future.

                The machines are not taking our jobs, but they are changing them. The responsibilities of machine operators need to change, so they act more as trainers and supervisors rather than directly controlling functions. This includes an intimate understanding of how and why machines and programs work, and being able to identify scenarios that are difficult for those programs to understand and compensate for sub-optimal performance.

                Autopilot programs have unequivocally improved air travel safety, but the pilot is not likely to become superfluous in my lifetime. Having a trained expert to watch over a program’s operations and correct for errors will remain vital to the airline industry, as well as a slew of others. Training and certification programs for people interacting with these programs needs to reflect this change to their role. Plane crashes may never fully disappear, but through sophisticated programming coupled with experts who understand its function can reduce them to statistical insignificance.

Comments: