neuwirth/levineuwirth.org
neuwirth
/
levineuwirth.org
1
1
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

auto: 2026-08-29T19:06:07Z [skip ci]

This commit is contained in:
Levi Neuwirth 2026-08-29 21:06:07 +02:00
parent b837ac2813
commit e74b87a537
No known key found for this signature in database
1 changed files with 8 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
content/work.md
View File

@ -151,14 +151,15 @@ against 0.655 for the Charlson index. The calculator is deployed.
## The question
::: {.work-thesis}
How do you establish trustworthy claims about an AI system when the system, the
operator, and the evaluator may each be untrusted?
Those four are one problem approached from different sides. Cryptography works
from below, certifying properties of a computation without trusting the party
that ran it. Evaluations work from above, measuring what a model actually does
under conditions you control. Formal methods supply the machinery for checking a
claim without trusting its author.
Cryptography attacks this from below, certifying properties of a computation
without trusting the party who ran it. Evaluations attack it from above,
measuring what a model actually does under conditions you control. Formal
methods supply the machinery for checking a claim without trusting its author.
The four projects above are attempts on different faces of the same problem.
The question underneath all of it: how do you establish trustworthy claims about
an AI system when the system, the operator, and the evaluator may each be
untrusted?
:::
## More work