NTSB pulls docket after AI recreates dead pilots' voices (opens in new tab)

Before you advocate changing an extremely successful safety culture because you want to apply abstract principles, you might want to do some Chesterton's Fence thinking. Aviation safety depends on fearless analysis of objective data and blameless reporting, which is a very unnatural and sometimes counterintuitive framework for humans to operate in.

The NTSB releases transcripts of cockpit voice recordings, just not the literal voices. This is a human consideration that doesn't affect the quality or transparency of the analysis.

The NTSB's publishing of the transcript but not the recording is a pretty standard means for providing full privacy while increasing safety. A recording allows an incident analysis which is extremely useful in updating safety procedures to prevent incidents and plan for the ones that do occur. Publishing the raw sound recording reduces privacy with no increase in safety, but publishing an analysis of the recordings does not harm privacy, while getting the entire safety benefit.

This is different than a privacy vs liability conflict, where a recording isn't going to provide a safety benefit, it'll just move liability around, where there's far more controversy over publishing any analysis of the recording, or even creating one in the first place.

The NTSB should never have published the unredacted spectrograph, as it is effectively a raw sound recording.

On what basis should the recording of voices of identifiable flight crew (often the last words of people soon-to-perish, perhaps unknowingly) be public information?

You seem fairly sure that the public has a right to hear them and that view is not universal and I'm not even sure is a majority view.

I'm only an ardent viewer of crash investigation stuff, not a pro, but it seems to be a good way to show specifics of warning noises, engine sounds, unusual cabin noises (if relevant) and sometimes even structural failures happening over time in a more direct way from the cockpit voice recorder without sharing the actual "audio".

A distressed airplane makes a lot of noises which are very difficult for the human ear to pick out and identify. For example, multiple closely-spaced bangs or rumbling noises will appear distinctly on a spectrogram but will be very hard to hear.

Basically they were just trying to track down weird noises on the recording. They also went and recorded audio on a similar UPS MD-11 to try to identify the source. According to the NTSB:

"This high pitch ringing sound, primarily noted at approximately 6.35 kHz, occurred at 17:13:05.5 EST, shortly after the aircraft rotated for takeoff, and continued with varying amplitude throughout the remainder of the recording. Additionally, a tone at about 2.1 kHz was present along with the ringing that could not be identified."

Where else would they get sounds from inside the cockpit that weren’t transmitted on the radio?

> Even a trained eye can read one, especially if context is provided.

I'd hope a trained eye could read one, that's the point of the training.

The article quotes the creator saying he used AI

FFTs are found in every nook and cranny of modern communications and computing.

It also works with time domain video files like audio visualizers: https://www.youtube.com/watch?v=E3gf88rSzqo

Nothing extremely surprising though.

It says in the article that the creator used OpenAI Codex, presumably because the spectrogram image wouldn’t have enough resolution by itself.

you are correct - I coded this in the late 1980s with digital sound domain experts