RSSAmplifier

Blog

Mark Xu

Writing by Mark Xu

markxu.com ↗RSS feed ↗54 posts

Latest posts

A model of commitments

Preference utilitarians should care about the preferences of past people

How to do theoretical research, a personal perspective

Facts from Essential Cell Biology CH5: DNA and chromosomes

On London

Comparing utilitarianism to deontology is a type error

Facts from Essential Cell Biology CH4: protein structure and function

Advice I commonly give people new to AI alignment

Facts from Essential Cell Biology CH3: energy, catalysis, and biosynthesis

Facts from Essential Cell Biology CH2: chemical components of cells

Facts from Essential Cell Biology CH1: cells, the fundamental units of life

Insights from Better: A Surgeon's Notes on Performance

Notes on SMILE eye surgery

Coase's theorem means the customer doesn't have to always be right

Your time might be more valuable than you think

Review: A Really Short History of Nearly Everything

Quirks with the Solomonoff prior

Insights from Complications

The simulation hypothesis undercuts the SIA/Great Filter doomsday argument

Answering questions honestly in the Game of Life

Fractional progress estimates for AI timelines and implied resource requirements

Intermittent distillations #4

Anthropic effects in estimating evolution difficulty

A rough perspective on strategy stealing

The wild world of policy debate

Rogue AGI embodies valuable intellectual property

An intuitive guide to Garrabrant induction

Intermittent distillations #3

Pre-Training + fine-tuning favors deception

Agents over Cartesian world models

Intermittent distillations #2

Transparency trichotomy

Intermittent distillations #1

Strong evidence is common

Open problems in myopia

Towards a mechanistic understanding of goal-directedness

How simulacra levels increase

Coincidences are improbable

Definitions and examples of simulacra levels

Money can't (easily) buy talent

Chain breaking

Defusing AGI danger

The first sample gives the most information

A math student's guide to options

Does SGD produce deceptive alignment?

The Solomonoff prior is malign

Fixed point theorems

Flavors of choice

Löb's theorem

Review: Against the Grain