Skip to content

Introducing LLM Integration in BeyondATC!

We are thrilled to announce the integration of a Large Language Model (LLM) into BeyondATC! This groundbreaking update represents a significant leap forward in our mission to provide the most immersive and realistic air traffic control experience for flight simulation.

This update will be available in the experimental branch for supporter users starting in early January 2025. After thorough testing, it will roll out to the main branch for all users at no additional cost. This LLM integration also paves the way for future features, including VFR functionality — a highly anticipated addition to our roadmap.

From the very beginning, we envisioned BeyondATC as a fully AI-powered tool. While earlier versions relied on carefully structured systems to ensure accuracy and functionality, the rapid advancement of AI technology has made it possible to reintroduce LLMs into the core of BeyondATC. Here's why now is the perfect time:

Improved structure

We have built a robust framework that gives the LLM a comprehensive understanding of the simulated flight environment. This framework ensures that every response aligns with the rules and logic of ATC operations. By maintaining strict "rails", the LLM avoids hallucinating instructions or responding inaccurately.

Reduced costs

Our proprietary, homegrown LLM is specifically designed for BeyondATC’s unique requirements. By running directly a distributed model instead of relying on expensive server infrastructure, we have significantly reduced operational costs. This enables us to offer this update as a free enhancement to all users.

Key features

  • Context-aware responses: The LLM processes situational data directly from BeyondATC’s 3D engine, ensuring it understands and reacts appropriately to the current environment.
  • Flexible communication options: Users can interact via voice, text, or a mix of both. An on-screen keyboard is now available for those who prefer typing.
  • Enhanced realism: The LLM’s responses are informed by the same rigorous structure that governs BeyondATC. The LLM is not involved in any decision making, it is primarily used for its speaking versatility. You will now be able to receive a response to any questions you ask and that is not falling into the flow of what it has been implemented.
  • No additional costs: This update will be included with your existing purchase of BeyondATC. There are no ongoing subscription fees. It will be first released to the experimental branch for our Supporters' users, but will be rolled out to everybody once it is stable, at no extra cost.

FAQ

What is a Large Language Model (LLM)?

An LLM is an advanced AI system trained to understand and generate human-like text. In BeyondATC, the LLM interprets user inputs, contextualizes them within the simulation, and provides accurate and logical responses.

How does the LLM enhance BeyondATC?

The LLM brings open-ended communication to BeyondATC, allowing users to interact naturally with controllers. It’s capable of answering specific questions about flights, airport operations, and traffic situations, significantly enhancing immersion.

Will this update cost extra?

No, the LLM integration is a free update for all BeyondATC users. It will be first available in the experimental branch available to all our Supporters. This update will then be included in the early access version at no extra cost.

Is the LLM always accurate?

While the LLM is built on a solid framework to ensure reliability, minor bugs and inconsistencies may occur, especially during the experimental phase. We’ll continuously improve the system based on user feedback.

Can I use this feature without a microphone?

Absolutely! BeyondATC will include an on-screen keyboard, allowing users to type questions and commands. This provides flexibility for those who prefer or need an alternative to voice input.

Does this mean that ATC will become less reliable than it was before?

This does not affect how BeyondATC operates. Core ATC interactions remain handled outside the LLM, ensuring accurate and reliable instructions without the risk of hallucinations. The LLM primarily serves to provide information based on the data it receives, enhancing the overall experience.

Why is the controller not saying the fix name/numbers correctly?

This is an experimental feature that is still being refined. As with any experimental functionality, adjustments are needed to ensure consistent performance. These details will be addressed when the development team focuses on finalizing the feature.

So now I will be able to declare emergencies? Or another approach?

This is not possible. The core ATC interactions are still handled outside the LLM to ensure consistency, reliability, and avoid any inaccuracies. The LLM enhances the realism of interactions by allowing controllers to understand anything you say on the frequency. However, it does not introduce new functionalities beyond what the development team has already implemented.

Does this mean no more "request not understood" or "did not copy"?

Not anymore! The LLM now handles all questions that fall outside the scope of the current ATC actions, ensuring you'll always receive a response! But remember that it will not be able to make any decision making. All the LLM will do is purely informational.

When will this update be available?

The experimental branch release is scheduled for early January 2025 for all our supporters. After testing, it will be made available to all users as part of a free update.