AI is being used to resurrect the voices of dead pilots

In a striking development reflecting the increasing role of artificial intelligence, the National Transportation Safety Board (NTSB) recently suspended access to its docket system. This action was taken after it was revealed that the voices of pilots who tragically lost their lives in a UPS plane crash last year had been reconstructed using AI and were being shared online. Federal regulations prevent the NTSB from including cockpit audio recordings in its publicly accessible docket system, which typically houses extensive data from investigations. However, the docket for this specific flight contained a spectrogram file derived from the cockpit voice recorder. This spectrogram, a visual representation created through a mathematical process, transforms sound signals into imagery, capturing a range of frequencies. Scott Manley, a well-known YouTuber who explores topics in physics, astronomy, and gaming, highlighted on social media the possibility of extracting audio from the spectrogram data. This led to individuals utilizing the spectrogram, alongside an available transcript, to approximate the cockpit audio from UPS flight 2976, which crashed in Louisville, Kentucky. Tools like Codex were reportedly employed in this reconstruction effort. While the NTSB reinstated public access to its docket system on Friday, it has kept 42 investigations, including that of Flight 2976, sealed for further review, raising questions about the implications of AI technology in sensitive cases.