Essays

Matching pennies essays

Rewards vs goals	-	The wirehead problem and forecasting first;
Command line	-	The Matching Pennies Command Line Protocol;
Probabilities	-	Revealing probability estimates;
Stream virtues	-	The virtues of streaming;
Generality	-	Maintaining generality while matching pennies;
Forecasting as compression	-	The link between forecasting and compression;
Inductive inference is like lift	-	Comparing machine intelligence to powered flight;
The broad scope of inductive inference	-	The implications and significance of inductive inference;
Manipulative forecasting	-	Forecasters may try to make the world simple;
Cybernetic diagrams	-	Diagrams of a simple forecasting agents;
Wirehead analysis	-	Will forecasting agents wirehead?
Collective intelligence	-	Strategies can join forces;
Code squeezing	-	On squeezing the model;
The virtues of generality	-	Why be a "jack of all trades"?;

Machine intelligence essays

(2023)

Universal instrumental values	-	Some instrumental values are widespread;
Expected utility maximisers	-	Basic optimisation architecture;
Self-improving systems	-	Systems thet feed on themselves;
Solomonoff induction	-	On near-optimal inductive inference;
Universal agents	-	Universal agents, sensation and action;
Oracles, sages, genies	-	A pantheon of superintelligences;
Utility counterfeiting	-	The problems of fake utility;
The one true razor	-	The optimality of Occam's razor;
On AIXI	-	On AIXI: a model of synthetic intelligent agents;
Reversibility and machine intelligence	-	Advanced machine intelligence and reversible computation;
Open source	-	On open source machine intelligence;
Beyond Occam's razor	-	Occam's razor has been superseded;
Machine faith	-	Self-fulfilling prophesies;
Cryptocurrencies	-	Cryptocurrencies and machine intelligence
Turing Test Tournaments	-	Designing Turing tournaments
Architectural limits to computation	-	A roadmap for overcoming computational bottlenecks
Deep prediction	-	Using neural networks for forecasting
Intelligence augmentation	-	Machine intelligence on a human foundation;
Angelic foundations	-	Building angels;
Intelligent design vs random mutations	-	How does intelligent design stack up against GAs;
The second superintelligence	-	One superintelligence - or many?;
The awakening marketplace	-	Stockmarket AI;
Superintelligence - the driving problem	-	The path to superintelligent machines;
The wirehead problem	-	Superintelligence vs addiction;
On parallelism	-	Computer architecture: the path to a parallel world;
Ubiquitous digitalisation	-	On the progressive digitalisation of the world;
Evaluation under simulation	-	How evolution performs evaluations under simulation;
Self improving systems are here already	-	Companies and governments as SISs;
Optimization	-	A brief introduction;
The number of insights before superintelligence	-	Number of insights required to produce superintelligence;
Against strong AI	-	The term "Strong AI" - and why you should probably avoid using it;
How quickly will machine intelligence take off?	-	How quickly will machine intelligence "take off";
Will machines take our jobs?	-	Will the advent of intelligent machines make humans redundant;
Handicapped superintelligence	-	How utility functions of intelligent agents may affect their survival;
Stopping superintelligence	-	Superintelligent agents that don't object to being switched off;
Ethical Artificial Intelligence	-	A review of Bill Hibbard's book;
Belief aggregation	-	A collective intelligence technique;
Far future humans	-	Instrumental reasons for keeping humans around;
Let's Merge!	-	Merging with machines;
p(doom)	-	The end of the world as we know it; (2024)
Agent Orchestration	-	Combining agents using collective intelligence; (2024)
Fellowship of AIs	-	Combining agents into a general purpose model; (2024)
Automating Intelligence Tests	-	Testing LLMs by using LLMs; (2024)
Bayesianism	-	Probability theory basics;

Tim Tyler | Contact | http://matchingpennies.com/