
Managing your downloads and pack in ForeFlight
Use ForeFlight’s Downloads and Pack tools to ensure you have the right charts and data—even when flying offline. Stay flight-ready!
Today we launched the first feature in ForeFlight Mobile leveraging artificial intelligence – though technically the second our team has shipped, following ForeFlight Voyager’s AI Aviation Historian feature released earlier this month. For the entire team, these features mark the beginning of a new era of computing in aviation and one that I think will have a profound impact over the coming years. So, I thought it would be a good idea to share my thoughts about the direction we’re taking with ForeFlight and artificial intelligence in aviation.
Artificial intelligence (AI) is a group of technologies that is fundamentally transforming the capabilities of computers today. The result is that they appear to ‘behave’ in ways that are more akin to human intelligence. AI has lived in pop culture for a very long time – often the antagonist in a science fiction movie (and on occasion an actual airplane). In reality though, the multiple technologies that are part of AI are delivering some incredibly positive things that only a few years ago were unthinkable. AI is also surrounded by some healthy skepticism and concerns over its application, which is often rooted in these pop culture portrayals.
To be precise here, AI or the use of AI in applications isn’t entirely new. AI technology has been in use in many products for a long time, including ForeFlight. When you get recommendations for alternates in the Flights view of ForeFlight Mobile, or when you get a recommended route from any of our products, there are AI components helping make those recommendations. However, over the past couple of years a wave of AI innovation, broadly known as Generative AI, has been rolling over the software landscape.
The models that have received the most attention in the past few years are Large Language Models (LLMs) such as ChatGPT. You’ve likely used ChatGPT, which is a text to text model. You prompt the model with text and it gives you a textual answer – and it’s pretty incredible what it can do when those text outputs are used in clever ways. As an example, those textual outputs can be used as input to other LLMs and the models then work together to achieve some outcome, an approach called multi-agent systems. Beyond textual output, AI models are also rapidly evolving to better interpret visual and audio input, enabled by a revolution in how the AI models are trained through an architecture known as transformer models. I won’t go into more detail, but you can find one of my favorite overviews on AI here.
There is absolutely no shortage of challenges that AI can help within aviation. But we all recognize this won’t happen overnight and it’s impossible to predict when we’ll feel the impact. Today, the discussion around AI will seem very familiar to anyone who has been in technology for a while. All new game-changing technologies go through a similar cycle of promise/hype, then disappointment, and eventually useful adoption. This notion was captured in the hype-cycle made famous by Gartner. AI is very much in the early stages of this cycle, but I actually think it’s going to go through this process faster than any technology that preceded it as the innovations are happening more rapidly than I’ve ever seen before.
Specific to aviation, I’m confident it will change our industry for the better, it’s just a matter of when. When considering the pace of AI adoption, I think about it in terms of three factors: optimization, regulation, and trust.
First, the systems being built tend to be very broad in their application. Take ChatGPT for example. It’s a large language model that can do a lot of different things and is being used by huge numbers of people. But it’s not optimized around one specific industry or topic. If you ask it some deep questions on aviation you’ll see that it comes up short. These models will need to be trained and optimized for our industry to understand our terminology and data. This will likely happen relatively quickly
Secondly, the regulatory nature of our industry will ensure adoption will be slow. By design, this conservative posture ensures safety and predictability. In areas where AI can be applied that are outside the purview of regulation (e.g. our airport comment summaries) things will move much faster.
Lastly, our ability to trust these models – on a very human level – will define our customers’ willingness to accept the use of AI. The adoption of AI driven solutions will often be a very personal one and ultimately comes down to their ability to trust the companies and AI models that are available to them.
As with any new technology – and perhaps more that is typical – AI is generating a lot of uncertainty and apprehension. When we look at today’s AI technology and how we’re bringing it into ForeFlight, we look at it in terms of its current strengths and weaknesses.
Our first step into AI leveraged the technology of LLMs to save our users time in their flight planning (if you’ve never read our airport comments or left the community a comment, please do as they are wonderfully useful). We used a trained LLM to analyze these comments and summarize them for quick reference. This is done off of the device and periodically, so it doesn’t impact the performance of the application at all. Eventually, more and more capabilities will be integrated into the app and device itself and you’ll see more incredible developments.
Our mission has always been to make flying easier and more enjoyable. We found that we can positively impact aviation safety too – and this is now a major binding element of our company culture. We’re now the EFB standard for cockpits around the world and we have a responsibility to continue to innovate and deliver on our mission. New technologies – especially AI – give us an opportunity to do this in ways we never thought possible.