This is a guest post on a topic that fascinates FlightChic by someone far more qualified to address it. Learn more About the Author.
The aviation industry just celebrated National Aviation Day. It’s a day meant to celebrate our pilots and all the other important players in the aviation ecosystem.
At the same time, avionics developers are working hard to engineer the technologies needed for fully automated flight through artificial intelligence, which some say will put pilots out of a job someday. Why are they saying that? And what does it mean for the average passenger?
AI Autopilot vs. Human Pilots: Why replace pilots?
Currently, two main factors contribute to any plans to eventually replace pilots with AI-powered autopilot systems. Number one is that there’s a pilot shortage. Training new pilots is complex and costly, and it doesn’t seem like the shortage will go away any time soon, so it seems natural to look for other options. And dramatically increasing flightdeck automation means aircraft complexity is increasing so that pilot training is even more important while also being more rigorous.
The second reason is that the aviation industry is headed for a new trend called urban air mobility (UAM). UAM would effectively replace traditional cars and taxis with “flying taxis.” To make that grand vision a reality, manufacturers would have to produce these aircraft in much greater quantities than they do today, and there likely won’t be enough pilots to keep up with the plan. So, one way forward involves automated aircraft such as Wisk’s planned autonomous eVTOL.
The day when UAM becomes a daily reality is still years away due to the technological constraints and infrastructure limitations involved. But autopilots are already commonplace, though usually coupled with human pilots for safety reasons.
Can an AI-enabled Autopilot Help Passengers Feel Safe In-flight?
On the topic of safety, I like to make a distinction between “feeling” safe and “being” safe. For example, if you enter an elevator that makes that awful shuddering sound as you head up to the next floor, you might not feel safe. The reality is that you almost certainly are, yet no one likes the experience of being in a misbehaving elevator. In the same way, not all passengers feel completely secure with the idea of a totally AI-piloted plane, but that feeling may not always line up with reality.
We like the idea of a human having control should anything go wrong. Many people have read the stories of crashes that may have involved some failure in an onboard automated system. On the other hand, some tests have shown that autopilot systems can usually perform better than human pilots in most conditions, and the technology is advancing daily. I will say, however, that the technology has a long way to go before it can completely guarantee safe flights for passengers and adhere to mandatory aviation standards.
Guaranteeing the safety of air travel
Actually, no one can really “guarantee” safety. That said, we can come pretty close with the right testing and development principles and compliance frameworks. After all, airplanes are a well-established technology. We know the right processes and development principles to make things safe. However, things get more complicated when AI-based autopilot systems enter the mix.
Of course, planes are already relying on autopilot systems to fly safely daily, though landing is usually left up to a human pilot. However, the autopilot systems of today depend on human oversight in case of an emergency rather than functioning in a completely automated fashion. So, while an autopilot can automatically perform certain pre-set tasks, it needs a pilot to remain in the cockpit and ensure the system completes those tasks correctly and safely. The pilot can regain control at any time and must remain awake and alert throughout any flight involving an autopilot system.
An AI Autopilot is Not All-Knowing, Yet
By contrast, you have many technological and regulatory difficulties if you talk about an AI autopilot entirely replacing pilots. This is because true AI, by definition, can learn and adapt, meaning it can change in ways that are hard to predict. Things that are hard to predict don’t sit well with the aviation industry.
For example, DO-178C, the primary compliance standard covering the safe development and deployment of airborne software, requires a high level of predictability, or determinism, in any airborne software. Determinism means that the software responds to tests and outside factors in a predictable way every time. So a true artificially intelligent autopilot – one that can make independent decisions based on weather, the proximity of other aircraft, and a million other factors – can’t offer the level of certainty DO-178C requires.
In aviation, safety and certainty go hand in hand. So either the regulations will need to change, or the AI integrated into your plane will have to be much more reliable and deterministic than we’re used to seeing from AI systems.
We’ve seen, for example, how many generative AI systems like ChatGPT simply make unexplainable errors, often called “hallucinations.” Experts question whether that can ever really be fixed. Autopilot is considerably more complex than a chatbot, and the risk of similar unexplainable errors remains. And that’s not a possibility we like to consider in the aviation industry. So, aviation is steadily progressing toward adopting artificial intelligence and machine learning. But, this author believes such a system is a decade away from a formal definition.
Final Thoughts on AI Autopilot vs. Human Pilots
I don’t see a reason to replace pilots at the moment. One day, perhaps the pressing needs of UAM will necessitate small aircraft piloted only by AI, but that isn’t the case today. Today, pilots offer an essential safeguard that compliance standards can’t guarantee. And any time in the near future, that will likely continue to be the case.
About the Author
BIO: Mr. Vance Hilderman is the principal founder/CTO of three of the world’s most significant aviation development/certification companies, including TekSci, HighRely, and AFuzion. Hilderman has trained over 31,000 engineers in over 700 aviation companies and 30+ countries. His intellectual property is in use by 70% of the world’s top 300 aviation and systems developers worldwide, and he has employed and personally presided over 500 of the world’s foremost aviation engineers on 300+ projects over the past thirty-five years. AFuzion’s solutions are on 90% of the aircraft developed over the past three decades. His latest book, Aviation Development Ecosystem, debuted at #1 on the Aviation category best-seller list.
Do you have thoughts to share on aviation?
Expert commentary is always welcome at FlightChic. Feel free to contact me to share your own vision of the future in the skies.