4 ms·
This is an AGI-hard problem that's roughly equivalent to "convert arbitrary free-form natural language text into a program that queries over arbitrary databases
by throwawaygh 5y ago
This is an AGI-hard problem that's roughly equivalent to "convert arbitrary free-form natural language text into a program that queries over arbitrary databases, email servers, or even physical records".
Actually, even worse than that, because FOIA requests are often written by non-lawyers trying to sound like they're lawyers. So the free form natural language text is horribly tortured pseudo-legalese.
- edoceo 5y agoAGI = Artificial General Intelligence
- Schiendelman 5y agoI don’t think so. I think it’s having all the comms in the open, and ensuring people can access that data and sort through it themselves.
- throwawaygh 5y ago1. FOIA/sunshine requests cover a lot of ground. Tons of stuff that isn't digitized or easily dumped to the internet is none-the-less subject to FOIA. 2. Still more is subject to FOIA but requires by-hand PII filtering before release. 3. TONS of stuff that's not subject to FOIA/sunshine passes through .gov Exchange email servers.
- Schiendelman 5y ago1) Yep, difficult to start digitizing everything but that’s well understood. 2) Not if you build the system in (1) to ensure that happens. 3) I am saying we should start making basically everything subject to FOIA unless it’s part of an established process (like bid discussions) that would delay that. Not saying it’s easy. I am saying it’s not the set of problems in the grandparent comment if you build your systems to fail open.