Thane Ruthenis

A Crisper Ex­pla­na­tion of Si­mu­lacrum Levels

Ideal­ized Agents Are Ap­prox­i­mate Causal Mir­rors (+ Rad­i­cal Op­ti­mism on Agent Foun­da­tions)

Most Peo­ple Don’t Real­ize We Have No Idea How Our AIs Work

How Would an Utopia-Max­i­mizer Look Like?

Don’t Share In­for­ma­tion Exfo­haz­ardous on Others’ AI-Risk Models

The Short­est Path Between Scylla and Charybdis

A Com­mon-Sense Case For Mu­tu­ally-Misal­igned AGIs Ally­ing Against Humans

“Hu­man­ity vs. AGI” Will Never Look Like “Hu­man­ity vs. AGI” to Humanity

Cur­rent AIs Provide Nearly No Data Rele­vant to AGI Alignment

Hands-On Ex­pe­rience Is Not Magic

A Case for the Least For­giv­ing Take On Alignment

World-Model In­ter­pretabil­ity Is All We Need

In­ter­nal In­ter­faces Are a High-Pri­or­ity In­ter­pretabil­ity Target

In Defense of Wrap­per-Minds

Ac­cu­rate Models of AI Risk Are Hyper­ex­is­ten­tial Exfohazards

Cor­rigi­bil­ity Via Thought-Pro­cess Deference

Value For­ma­tion: An Over­ar­ch­ing Model

Greed Is the Root of This Evil

Are Gen­er­a­tive World Models a Mesa-Op­ti­miza­tion Risk?

AI Risk in Terms of Un­sta­ble Nu­clear Software