The archive is too big for anyone to have read even a fraction. Fortunately, a group of software engineers has turned the PDFs into a format that is easier to analyse. Using Reducto, an AI tool, they have identified which files contained emails; extracted the listed senders, recipients, dates, subjects and message bodies; and posted them on a website called Jmail.world. In total, the group processed 1.4m emails, finishing its work on February 11th. The Economist has collaborated with it to assign each message to unique individuals regardless of spellings or email addresses, and researched the backgrounds of the 500 people who appear most often. We then used a large language model (LLM) to score each email chain on how disturbing its content would be to a typical reader, creating an “alarm index”.
https://archive.is/VgvU7
The archive is too big for anyone to have read even a fraction. Fortunately, a group of software engineers has turned the PDFs into a format that is easier to analyse. Using Reducto, an AI tool, they have identified which files contained emails; extracted the listed senders, recipients, dates, subjects and message bodies; and posted them on a website called Jmail.world. In total, the group processed 1.4m emails, finishing its work on February 11th. The Economist has collaborated with it to assign each message to unique individuals regardless of spellings or email addresses, and researched the backgrounds of the 500 people who appear most often. We then used a large language model (LLM) to score each email chain on how disturbing its content would be to a typical reader, creating an “alarm index”.