Edimah SYNESIUS SONGO
Data & statistics freelancer with an applied maths degree learning about AI safety.
Recent writing
-
How to evaluate AI progress when the evidence is confidential
Thoughts about an independent evaluator's access
-
Detecting evidence of secret loyalty in a model using only its answers requires careful design
An exploration of statistical methods for secret loyalties detection.
-
The impact of a base-level AI risk course on an aspiring technical AI safety researcher
Why keep trying in the age of AI