AI and Deep Learning: A Primer for Aviation Professionals
What AI and deep learning actually mean for helicopter operations, and what VoxVision does with them.
Updated September 10, 2026. This primer was first published in March 2024. We have clarified the distinction between training and deployed models, program-specific data rights, and the conditions for software releases. See VoxVision and Missions for current capabilities and roadmap priorities.
The Basics of AI in Aviation
What is Artificial Intelligence (AI), and how does it apply to aviation? AI and deep learning are in the headlines a lot these days, and much of the focus is on the amazing work being done with Large Language Models like ChatGPT and other “generative” AI.
AI in aviation is different. As aviation professionals we’re focused on doing jobs in the real world, to unparalleled safety and reliability standards. So how can AI and deep learning help? Answering that starts with what we actually mean when we talk about AI.
At some level, an AI or deep learning system can be likened to the sophisticated systems and automation that pilots and operators are already familiar with, but taken to a significantly more advanced level. At its core, AI is about creating systems that can perform tasks requiring human intelligence, such as recognizing objects and conditions, interpreting data, making decisions, and learning from experiences.
Deep Learning: The Brain Behind AI
Deep learning uses neural networks to learn patterns from examples. In an airborne application, a trained model can help interpret imagery, such as identifying a feature for an operator to review. Running that model on the aircraft is different from training it: a deployed model does not necessarily learn from each flight.
More data does not automatically mean a better model. Relevant examples, data quality, evaluation, and validation against the intended task determine whether a change is useful. A model developed for one mission should not be assumed to work for every other mission.
This concept applies to almost every relevant area in which pilot experience matters: flight path planning, understanding and predicting weather conditions, recognizing incipient hazards, and extracting the most performance from a given aircraft type, to name a few. It also applies to any task that could be performed by a crew member onboard: looking for objects, assessing terrain, or capturing data.
The Potential of AI for Helicopter Operators and Users
Potential applications include crew decision support and mission-specific observation. These are different engineering tasks, with different data, evaluation, integration, and operational requirements. A list of possible AI applications is not a list of available Voxelis features.
VoxVision: Enabling Advanced Capabilities
VoxVision combines visible and thermal imaging, onboard computing, and a tablet interface in an aircraft-mounted system. Current wildfire functions include automatic fire detections, mapped fire extent, temperature estimates, and exports. VoxNet adds live viewing and ground control of the gimbal, subject to connectivity and the access granted for the operation.
The Role of Data
Useful airborne models need relevant data and a way to test performance against the mission. Voxelis develops proprietary datasets alongside its hardware and software. Whether collected material may be used for training depends on the applicable data rights, permissions, and program arrangements.
Looking Forward: The Future of AI in Helicopter Aviation
Future priorities include mission-specific models, crew awareness tools, suppression aircraft assistance, and sharing observations between aircraft. These are roadmap directions, not a commitment that every function is released or available for every aircraft.
Software and model releases depend on validation, compatible sensors and computing, aircraft integration, connectivity, and applicable approvals. The release scope and supported configurations are confirmed for each deployment. See the VoxVision upgrade path for the current overview.
Join Us
We’re building this with operators and users, not for them. Contact us if you want to follow the work.