Rauno Arike

Rauno’s Shortform

A Dialogue on De­cep­tive Align­ment Risks

[In­terim re­search re­port] Eval­u­at­ing the Goal-Direct­ed­ness of Lan­guage Models

Early Ex­per­i­ments in Re­ward Model In­ter­pre­ta­tion Us­ing Sparse Autoencoders

Ex­plor­ing the Lot­tery Ticket Hypothesis

[Question] Re­quest for Align­ment Re­search Pro­ject Recommendations

Coun­ter­ing ar­gu­ments against work­ing on AI safety

Clar­ify­ing the con­fu­sion around in­ner alignment