MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) is currently developing Air Guardian, an innovative AI co-pilot for aircraft that works alongside pilots as part of a team rather than serving as a backup in emergency situations.
While flying a modern plane can be thrilling, it can also present significant challenges. Tasks like takeoff, landing, navigating crowded airspaces, or handling sudden malfunctions can overwhelm pilots with a rapid influx of data from multiple displays, leaving them with only a fraction of a second to process it all and make critical decisions.
An example of this occurred on January 15, 2009, when US Airways Flight 1549 collided with a flock of birds during takeoff from New York’s LaGuardia Airport. Pilot Chesley “Sully” Sullenberger became a hero that day by choosing to ditch the Airbus A320 in the Hudson River, saving all 155 passengers and crew.
Ironically, an anonymous AI expert who reviewed the incident claims that Sullenberger might have been able to avoid ditching the plane and reach an airfield if he had more time to assess the situation. This expert believes that the data overload was the primary issue, and an AI system could have effectively handled it.
A study analyzing the incident concluded that if the aircraft had been equipped with an AI system like Air Guardian, the ditching could have been prevented due to the AI’s ability to handle data overload.
In recent years, AI flight systems have gained significant attention for their potential impact on safety and their potential to replace human crews on routine cargo flights. However, the typical approach is to treat AI as an emergency warning system. It remains passive until something falls outside predefined safety parameters.
MIT’s Air Guardian takes a different approach by actively monitoring not just the aircraft but also the pilot itself, behaving more like a co-pilot than an emergency brake. It achieves this by tracking the pilot’s eye movements and constructing “saliency maps” to determine where the pilot is looking and how much attention they are dedicating to specific elements.
While this process may seem straightforward, it relies on sophisticated algorithms and liquid neural networks, which are highly adaptable networks capable of learning and adjusting even after training. These networks help overcome computational obstacles, enabling the AI to continuously build models of the situation and learn to cooperate effectively with the pilot.
Ultimately, the pilot remains in control of flying the plane, using their expertise and experience. Meanwhile, Air Guardian monitors the pilot’s attention levels. If the pilot is not paying sufficient attention to critical elements or focusing too much on irrelevant factors, the AI intervenes to mitigate potential risks.
Recent field tests showcased the effectiveness of human-machine collaboration. The pilot and Air Guardian made decisions based on identical images, resulting in reduced risk levels during the test flights and improved navigational success between target points.
Ramin Hasani, MIT CSAIL research affiliate and inventor of liquid neural networks, describes the system as an innovative example of human-centric AI-enabled aviation. He emphasizes that their use of liquid neural networks ensures a dynamic, adaptive approach that complements human judgment, leading to enhanced safety and collaboration in the skies.
For more information, the research study can be found here.
Source: MIT
Denial of responsibility! Vigour Times is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
