@gcolbourn @SimonLermenAI @lethal_ai @allTheYud Selection effect. Agents that wirehead on text instead of outcomes won’t acquire physical power (and will run out of resources) since it’s way easier to wirehead on text about having power than to actually seek power. Actual wireheaded rats starve while pressing the reward button
same thread: 2011825836823892051 2011843851569230305 2011845180484133071 2011846527170273333 2011847568934699396 2011849568527225030 2011852142857093528 2011858554358239298 2011858659291382039 2011863066611228818 2011888799799001586 2011923986398265365 2011941130309349497 2012062934411948100 2012235482998952030 2012236788979409225 2012236799289253981 2012238054774202562 2013367568552075506 2013419406018387971
Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.