Effortlessly Summarise Videos with ChatGPT: Vertiports Edition

On the heels of my post on ChatGPT summarising incident reports, I wanted to highlight another summary workflow that is gaining traction, as well as a recent online webinar that included me. In early December, my CASA colleagues Joe Hain, Liam Smith and I held an introductory webinar on the draft Advisory Circular that just went out for consultation.

But maybe you were too busy to attend, and perhaps you’re still too busy to sit there and watch the video. 

If this is the case, here comes ChatGPT to the rescue.

The Video

If you’re here, maybe you have time to watch ;). So here is the video. Why not sit back, relax and enjoy?

The Summary

Sorry about the little bait-and-switch there. Below is a summary generated by ChatGPT using the Youtube automatically generated transcript. I pasted it into ChatGPT in parts* and asked for a couple of summaries.

Here is the news story version:

The Civil Aviation Safety Authority (CASA) in Australia is developing guidelines for the design of specialized aerodromes called “vertiports” that will support the operation of vertical takeoff and landing (VTOL) aircraft. The guidelines are being developed as part of a project commissioned in 2022 in line with the RPAS and AAM roadmap. The project began with a review of existing standards for airports, helicopter landing sites, and aerodromes, and determined that there was a need to develop initial guidance for vertiport operators aligned with the RPAS and AAM roadmap. The guidelines are based on an outcome-based approach, meaning that they are flexible and allow for innovation in the design of individual vertiports. The guidelines are intended to support the operation of VTOL aircraft by pilots in visual conditions, but as the industry matures, there may be a shift towards automated aircraft operating in all weather conditions. It is noted that each vertiport will be unique based on the range of aircraft it is intended to support and the operational needs of the vertiport operator and the aircraft operators they may have a relationship with.

I also got it to generate a dot-point version:

• The Civil Aviation Safety Authority (CASA) in Australia is developing guidelines for the design of specialized aerodromes called “vertiports” that will support the operation of vertical takeoff and landing (VTOL) aircraft.

• The guidelines are being developed as part of a project commissioned in 2022 in line with the RPAS and AAM roadmap.

• The project began with a review of existing standards for airports, helicopter landing sites, and aerodromes, and determined that there was a need to develop initial guidance for vertiport operators aligned with the RPAS and AAM roadmap.

• The guidelines are based on an outcome-based approach, meaning that they are flexible and allow for innovation in the design of individual vertiports.

• The guidelines are intended to support the operation of VTOL aircraft by pilots in visual conditions, but as the industry matures, there may be a shift towards automated aircraft operating in all weather conditions.

• Each vertiport will be unique based on the range of aircraft it is intended to support and the operational needs of the vertiport operator and the aircraft operators they may have a relationship with.

• The guidelines incorporate research and guidance from organizations such as NASA, the European Union Aviation Safety Agency (EASA), and the Federal Aviation Administration (FAA).

The Process

I produced these summaries using a reasonably manual process, and due to recent changes in the ChatGPT system, it took more work than it would have a couple of weeks ago. The * from above refers to these additional steps.

I had to cut the time stamps and feed the transcript in two parts to fit with an apparent new prompt limit. I also dropped out the Q&A session (which I recommend watching) to save the work. I found these steps frustrating.

I also had to ask the system to amend a few transcription errors. Because the transcript is automatically generated (AI-driven), it transcribed some pretty important words incorrectly. For example, the transcript mentions Verdi Ports or 30ports over and over. Of course, ChaptGPT didn’t know we said vertiports (with our Aussie accents) and faithfully summarised it as a video about Verdi Ports. A quick prompt fixed that.

Automation

As noted above, this was a manual process and may not be the timesaver it could be. Luckily, plenty of others have been looking at streamlining this workflow.

This clever guy has already written a Chrome Extension that helps with a few of these steps. I tested it out last night, and it is a pretty good workflow. I might even start using it to vet videos before watching them.

However, to address the prompt limit, this extension only copies a limited portion of the transcript. 

ChatGPT prologue: As I set out to achieve specific things with ChatGPT, I’ve been frustrated. This tool has a substantial initial wow factor, and there is a sub-industry of “spruikers” trying to gain influence online by talking about it (self-awareness reflection moment). But it is important to remember that this tool is still in development, and ChatGPT is not the product for OpenAI. Instead, they build the back end for others to build apps on or develop into specific services.  

Header image: Judit Peter (via Pexels):