The developers of BeyondATC have announced the upcoming addition of a Large Language Model (LLM), often referred to as simply A.I. (Artificial Intelligence) to BeyondATC, a simulation of air traffic control for Microsoft Flight Simulator in the next month. This update was shared in their final post on their discord for the year and shows their dedication to enhancing air traffic control features through modern solutions.
Their LLM Journey: From Skepticism to Integration
In previous updates, BeyondATC explained its initial decision to avoid using an LLM for its core engine of the product. They highlighted concerns that early Large Language Model implementations occasionally produced commands that weren’t real air traffic control instructions. Additionally, the costs of running an LLM were found to be uneconomical and would have significantly increased the cost of the product. As a result, the team prioritized refining their core engine with predefined instructions.
Localized LLM for Stability and Cost-Effectiveness
Against the previous concerns, BeyondATC has now developed and integrated an early version of their Large Language Model that runs locally on users’ hardware. This approach reduces the need for costly server infrastructure on their side. The LLM is designed to work within strict parameters set by BeyondATC’s core engine, ensuring reliability and addressing earlier concerns about cost and functionality.
The LLM integration is set to launch on the Experimental branch in January. BeyondATC’s recent showcase video on their YouTube channel provides more details about this development and demonstrates the new feature in action. Stay tuned for the upcoming release, as we may release a review on it.
Feel free to join our Discord server to share your feedback on the article, screenshots from your flights or just chat with the rest of the team and the community. Click here to join the server.
