Flying is totally safe: airlines show AI how it’s done!
Article reprinted under a Creative Commons license. See the original here.
Over the past century, about 185,000 people have died. However, in the last five years, the. In fact, you are more likely to win the lottery than die as a passenger on a US airline. How did flying become so safe? Can we apply the hard-earned safety lessons to artificial intelligence?
When humanity introduces a new paradigm-changing technology that is rapidly adopted globally, the future consequences are unknown and often feared. In 1903, the Wright brothers introduced the, and this new technology was no exception. It sparked controversy, objections, and concerns of political, religious, and technical nature.
It was on the day of the first plane crash. On that same day, in the. Five years later, in 1908, during a plane crash. And since then, there have been 89,000 plane crashes worldwide.
As a researcher focusing on, I see today that the AI industry resembles those early years of the aviation industry, and it is decidedly less safe.
From studying accidents to predicting them
Although tragic, each accident and fatality represented a moment for reflection and learning. Accident investigators tried to recreate each accident and identify the precursors and causes in each case. When they managed to identify what had caused each accident, aircraft manufacturers and operators implemented safety measures in hopes of preventing additional accidents.
For example, if a pilot in that early era of flight forgot to lower the landing gear, there would likely be an accident upon landing. So the industry installed warning systems to alert pilots of the landing gear status. A lesson that was only learned after accidents occurred. It is a reactive and necessary process, albeit at a high cost for us to learn to improve safety.
Throughout the 20th century, the aviation world organized and standardized operations, processes, and procedures. In 1938, President Franklin Roosevelt signed the Civil Aeronautics Act, creating the, precursor to the Federal Aviation Administration, and an Air Safety Board.
The completely reactive safety paradigm began to change to become proactive and then predictive. In 1997, a group of industry organizations, labor, and government formed the. They began analyzing data to identify trends and analyzed reports to identify risks before they turned into accidents.
It’s all about the data
The Commercial Aviation Safety Team helped the industry transition from reactive to predictive by adopting a systemic and data-based perspective to solve safety problems. They generated this data using reports from people and aircraft data.
There are millions of flights worldwide every day, and each flight records thousands of data points. Today, aviation safety professionals use flight data recorders – which were always used after accidents for investigation – and analyze the data from each flight. With this thorough analysis, safety analysts can. For example, a trained safety scientist can analyze the data and detect if there are riskier approaches to the landing strip due to factors like excessive speed or poor alignment, before an accident can occur upon landing.
To enhance proactive and predictive capabilities, those operating within the aviation system can submit anonymous, non-punitive safety reports. Without this guarantee of anonymity, some might hesitate to report a problem, and that would cause the aviation industry to miss out on crucial safety-related information.
All this data is stored, combined, and analyzed. Safety scientists study the entire system trying to find accident precursors before they can cause accidents. The risk of dying as a passenger on a US airline today is less than. You are more likely to die than in a plane crash. Over 100 years since the advent of powered flight, the aviation industry, after the most painful lessons, has become extremely safe.
A model for AI
AI is rapidly permeating many aspects of life, from autonomous vehicles to criminal justice actions and hiring or mortgage decisions. The technology is not yet entirely safe, however, and there have been AI-attributed errors that in some cases had.
Almost all AI companies now try to implement. However, they seem to be making individual efforts, as it used to happen in aviation. Their efforts are also mostly reactive. That is, the error is expected and then action is taken.
What if there was a group like the Commercial Aviation Safety Team, where all AI companies, regulators, academics, and stakeholders in AI come together to start proactive and predictive processes to ensure that AI does not lead us to disasters?
Technical Analysis: Ensuring AI Safety
From the perspective of reports, imagine if each AI interface had a report button that the user could activate to report unsafe or potentially hallucinated results, providing information to the company and an AI organization like the Aviation Security Team? The data generated by AI systems, as in aviation, could be recorded, aggregated, and analyzed to detect security risks.
While this may not be the ideal solution to prevent harm from the use of AI, if big tech companies adopt the lessons learned from other high-risk industries like aviation, they may be able to learn how to regulate, control, and make AI safer for everyone.
– Aviation Professor, North Dakota.
