3 Ways To Use AI In Your Pilot Training

I’m a private pilot, author, and speaker, who is working on his instrument rating. I’m a big believer in using all tech tools at our disposal to fly safer and train better. Usually, when we think of tech helping our flying, we think of things in the cockpit. But with the advent of AI, new ways to become a better pilot are available to us while feet are firmly planted on the ground. I’ve been experimenting a lot with AI to see how it can help me train more effectively. Below, I share three use cases that you might find helpful in your own training.

Organizing disjointed information – who else finds ATOMATOFLAMES pretty useless? If so, ChatGPT (the tool I used in all the use cases here) is very good at suggesting alternative ways to think about the laundry list of things we need to commit to memory during training. As I was trying to memorize not only ATOMATOFLAMES (VFR equipment requirements) but also FLAPS (night equipment requirements) and GRABCARD (IFR equipment requirements) I found the list of 26 letters difficult to remember.

That shouldn’t be surprising. About the only 26 letter string you know by heart is the alphabet itself! It doesn’t help that so many letters are repeated. The letter “A” makes an appearance six times in the three mnemonics.  So I asked ChatGPT to reimagine these mnemonics and suggest a better way to recall the requirements. In response, the AI created a table that grouped the information by systems and instruments, lighting & electrical, safety, and comms/navs. These items made up the rows of the table and the type of flight (VFR, night, and IFR) made up the columns. I found this representation much easier to remember; it made me think more logically about why certain items are needed for certain types of flight. Another example: I used ChatGPT to make a table that put in one place the four types of GPS approaches I’ll encounter: LPV, LNAV, LNAV/VNAV, and LP. Such a table was hard to find online, and it significantly increased my understanding of the differences in these types of approaches.

Scripting ATC communications – While its first effort wasn’t perfect, ChatGPT wasn’t too bad at creating a script of radio communications that I could use when flying a home simulator to help me think through the sorts of things that I’d hear while shooting practice approaches. Take this prompt for example:

I want to use a flight simulator to practice flying instrument approaches. Assume I'm flying from 6B6 to KASH and I'm going to fly the ILS 14. Can you provide a script that I can just repeat to myself for realism’s sake that includes what I am likely to say to approach and what they are likely to say to me. Also simulate the same back-and-forth once I am handed over to the tower. My call sign is NXXXX and I'm flying a Cessna. Assume I'm not flying on a flight plan but am with a safety pilot practicing the approach. The approach should terminate with me flying the published missed approach. 

The AI did an amazing job of developing the script. It accurately depicted me making my request to practice the ILS runway 14 to Boston Approach, had approach respond with a squawk code, vectored me, scripted the hand-off to tower, etc.

So, why wasn’t it perfect? First, it didn’t initially give me realistic vectors. KASH is north of 6B6 so an initial on-course heading should have been in the range of 350 to 010. The first vector it gave me was 090, followed by 040. While it can’t be perfect since the vector you fly in any given approach depends on how long your stay on another vector, your speed, and other factors, having me point away from KASH right off the bat didn’t make sense. Two other issues were evident in its first attempt too: it made up the names of fixes on the approach, and it made up the frequencies for approach and the tower.

But here’s the cool thing … I told ChatGPT about these errors and it fixed them on the next go-around, producing a flawless script for how communications might go on an actual flight. The lesson – don’t throw out results that seem strange at the outset; ask the tool to correct its mistakes.

Clarifying regulations – the large language models that power AI were made for clarifying “legalese.” If you want to understand in plain language what a section of the FAR/AIM is saying, paste it directly into ChatGPT and you’ll get a very nice summary that is much easier to understand. And if you’re not sure what section and paragraph of the FAR/AIM covers a certain topic, just ask. Here’s an example prompt with the response:

Me:

Can you tell me the part of the FAR/AIM that deals with instrument currency?

ChatGPT:

In the FAR/AIM, instrument currency requirements are primarily laid out in:

14 CFR § 61.57(c) — Instrument Experience

This is the key regulation you’re looking for.

What it covers:

§ 61.57(c) specifies the requirements to act as PIC under IFR or in weather less than VFR minimums.

The core rule (the “6-6-6”):

Within the preceding 6 calendar months, you must have performed and logged:

• 6 instrument approaches

• Holding procedures and tasks

• Intercepting and tracking courses through navigational electronic systems

These can be done:

• In actual IMC, or

• Simulated (with a safety pilot), or

• In an approved simulator/FTD

I’m sure there are hundreds of other ways to use AI tools to enhance training. Let me know if you’ve found some unique approaches (no pun intended). And, as I learn more, I’ll do the same here!