Ask HN: What data have the frontier AI companies not ingested yet?
3 comments
The unredacted Epstein files.
Not being glib. Without them, a whole lot of incriminating facts and allegations are being whitewashed from the world's knowledge base. I believe that the distinction between the two is made in the documents released so far.
Not being glib. Without them, a whole lot of incriminating facts and allegations are being whitewashed from the world's knowledge base. I believe that the distinction between the two is made in the documents released so far.
time between tokens, composite data but that would require more data centers.
Corporate data belonging to non-AI companies.
Data that exists only in filing cabinets or on microfiche.
Whatever’s on that collection of CDs and floppy disks that you still have sitting around.
The home videos nobody’s bothered to digitize.
Private art collections.
Photo prints from film cameras that haven’t been digitized.
Airplane black box data.
Personal handwritten journals.
Plenty of private dead tree correspondence.
The vast majority of grocery store receipts.
Properly stored passwords.
Proprietary code produced by companies that don’t use AI or use only self-hosted models.
Scientific data that hasn’t been published or analyzed with AI, including plenty of older data.
Annual reviews predating LLMs.
Information exchanged during phone and video calls that aren’t recorded or haven’t been submitted for transcript generation / summaries.
Most board meeting minutes.
Anything handwritten that hasn’t been scanned.
The contents of many private online communities.
A lot of classified information.
I think that’s enough for now.