Then there is the question of whether SMART could worsen the chronic shortage of air traffic controllers. The FAA hopes technology will ease the shortage by requiring fewer workers; However, other sectors have yet to see the efficiencies generated by AI materialize. Even if the software doesn’t directly replace humans, its effectiveness could be used to justify downsizing. In May, the FAA reduced its active air traffic controller staffing goal from the 2024 target of 14,633 to 12,563, citing “modern staffing models and planning tools.”
Some are more optimistic than others. “I think the SMART program is a significant improvement,” says Schiavo. “No one wants today’s aviation environment with midair flights, runway incursions, myriad near misses, delays, groundings, loss of separation and long wait times for taxiing or clearance. »
Michael McCormick, a professor of air traffic management at Embry-Riddle Aeronautical University, also sees the benefits of a predictive tool. “When you have the computing capability to look at the entire trajectory of the plane, you can see where the conflicts are going to be,” he says. “And the general axiom of air traffic control is that the sooner you are informed of a potential conflict, the less effort it will take to resolve it.”
This sounds great, but building machines that do a good job most of the time is one thing; machines capable of operating at the level of reliability required by aviation are another. Currently, researchers are focusing on so-called deterministic neural networks, that is, models trained offline whose weights are frozen before being deployed in the real world. These cannot hallucinate, but they may harbor other risky failure modes. In a 2021 article, Darren Cofer, a researcher at Collins Aerospace, pointed out that when such systems find themselves in situations they have never encountered before in training, they can generate “unexpected (and potentially dangerous) results.”
The stakes will be even higher when AI software begins to make real-time air traffic control decisions, even in the airspace reserved for autonomous aircraft. According to McCormick, these will most likely be select corridors near drone hubs and eVTOL vertiports. With swarms of robot planes coming and going, things will be too complicated for human pilots and controllers. “The goal is not to use traditional air traffic control,” he says, “because that would overwhelm the national airspace system if it tried to manage that volume and complexity of moving cells.”
With two systems working side by side – one with human decision-makers, the other with AI – we will have a real-time basis for comparing the performance of the two approaches. Any computer system can potentially be hacked; any AI system creative enough to imagine new solutions could also hallucinate crucial details. Without a human in the know, seemingly obvious errors could go unnoticed until it’s too late.
Let us know what you think about this article. Send a letter to the editor to (email protected).
Gn bussni
