Eliezer Yudkowsky

Created by a_soulless_automaton at 2y
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of cognitive biases – systematic deviations from rationality – influencing our judgment of global catastrophic risks. These are defined as "where an adverse outcome would either annihilate Earth-originating intelligent life or permanently and drastically curtail its potential." Examples include volcanic eruptions, pandemic infections, nuclear war, worldwide tyrannies, out-of-control scientific experiments, or cosmic hazards). Yudkowsky's chapter specifically examines how cognitive biases impact thinking about global catastrophic risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Another chapter in the compilation "Global Catastrophic Risks", it analyses possible philosophical and technical failures in the construction of a Friendly AI, which could lead to an Unfriendly AI posing a enormous global risk. He also discusses how a Friendly AI could help decrease some global risks discussed in the book. Finally, because a powerful Friendly AI could reduce global risk, he argues that researching Friendly AI is extremely important for the future of humanity.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent — "Friendly" — Artificial Intelligence . It also gives one of the first precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): Analyzes AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. Also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It argues that making an AGI simply do what we tell it to could be dangerous, since we don't know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of what humanity would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless decision theory, ”an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory or Evidential Decision Theory don't have a plausible solution: Newcomb's problem, Solomon's Problems and Prisoner's dilemma.
  • "Complex Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of value: we can’t come up with a simple rule or description that sums up all human values. It analyzes how this problem makes it difficult to build a valuable future.

Eliezer Yudkowsky is a research fellow of the SingularityMachine Intelligence Research Institute for Artificial Intelligence, which he co-founded in 2001. He is mainly concerned with the obstacles and importance of developing a Friendly AI, such as a reflective decision theory that would lay a foundation for describing fully recursive self modifying agents that retain stable preferences while rewriting their source code. He also co-founded Less Wrong, writing the Sequences, long sequences of posts dealing with epistemology, AGI, metaethics, rationality and so on.

Eliezer Yudkowsky is a research fellow of the Singularity Institute for Artificial Intelligence -, which he co-founded in 2001. He is mainly concerned with the obstacles and importance of developing a Friendly AI, such as a reflective decision theory that would lay a foundation for describing fully recursive self modifying agents that retain stable preferences while rewriting their source code. He also co-founded Less Wrong,Wrong, writing most part of Thethe Sequences, long sequences of posts dealing with epistemology, AGI, metaethics, rationality and so on.

  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biasescognitive biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks.global catastrophic risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threatening more than millions of human lives(e.g.:"where an adverse outcome would either annihilate Earth-originating intelligent life or permanently and drastically curtail its potential." Examples include volcanic eruptions, pandemic infections, nuclear accidents,war, worldwide tyrannies, out-of-control scientific experiments, climatic changes,or cosmic hazards and economic collapse)hazards). It is a bookYudkowsky's chapter from a larger tome analyzing thosespecifically examines how cognitive biases impact thinking about global catastrophic risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter ofAnother chapter in the same book of the previous paper,compilation "Global Catastrophic Risks", it analyses possible philosophical and technical failures in the construction of a Friendly AI, which could lead to an Unfriendly AI posing a enormous global risk. He also discusses how a Friendly AI could help decrease some Global Risksglobal risks discussed in the book. Finally, because a powerful Friendly AI could go from been a Global Risk to help reduce some of them,global risk, he argues that researching such topicFriendly AI is extremely important.important for the future of humanity.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): AnalysesAnalyzes AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It alsoAlso discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposesargues that making an AGI doingsimply do what we tell it to could be dangerous, since we don`don't know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of uswhat humanity would want, "if"if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless decision theory,”an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory or Evidential Decision Theory don't have a plausible solution: Newcomb's problem, Solomon's Problems and Prisoner's dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of value: we can’t come up with a simple rule or description that sums up all human values. It analysesanalyzes how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threatening more than millions of human lives(e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter of the same book of the previous paper, it analyses possible philosophical and technical failures in the construction of a Friendly AI, which could lead to an Unfriendly AI posing a enormous global risk. He also discusses how a Friendly AI could help decrease some Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce some of them, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): Analyses AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless decision theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory or Evidential Decision Theory don't have a plausible solution: Newcomb's problem, Solomon's Problems and Prisoner's dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of valuesvalue: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threatening more than millions of human lives(e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter of the same book of the previous paper, it analyses possible philosophical and technical failures in the construction of a Friendly AI, which could lead to an Unfriendly AI posing a enormous global risk. He also discusses how a Friendly AI could help decrease some Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce some of them, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): Analyses AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless decision theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory or [[Evidential Deicision Theory]Theory don't have a plausible solution: Newcomb’s Problem, Solomon’s Problems and Prisoner’s Dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of values: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threatening more than millions of human lives(e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter of the same book of the previous paper, it analyses possible philosophical and technical failures in the construction of a Friendly AI, which could lead to an Unfriendly AI posing a enormous global risk. He also discusses how a Friendly AI could help decrease some Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce some of them, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): Analyses AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless decision theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory doesn`or [[Evidential Deicision Theory] don't have a plausible solution: Newcomb’s Problem, Solomon’s Problems and Prisoner’s Dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of values: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threatening more than millions of human lives(e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter of the same book of the previous paper, it analyses possible philosophical and technical failures in the construction of a Friendly AI, which could lead to an UnfriendilyUnfriendly AI posing a enormous global risk. He also discusses how a Friendly AI could help decrease some Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce some of them, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): Analyses AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless decision theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory doesn`t have a plausible solution: Newcomb’s Problem, Solomon’s Problems and Prisoner’s Dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of values: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threatening more than millions of human lives(e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter of the same book of the previous paper, it analysisanalyses possible philosophical and technical failures in the construction of a Friendly AI leading, which could lead to an UnfrendilyUnfriendily AI posing a enormous global risk. He also discusses how a Friendly AI could help decreasing the othersdecrease some Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce others risks,some of them, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first more precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): AnalysisAnalyses AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless Decision Theorydecision theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory doesn`t have a plausible solution: Newcomb’s Problem, Solomon’s Problems and Prisoner’s Dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of values: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008): A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of Global Catastrophic Risks. These are defined as risks with the potential to inflict serious damage to human well-being on a global scale, threaten could killthreatening more than millions of human lives called Global Catastrophic Risks (e.lives(e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”: Chapter of the same book of the previous paper, it analysis possible philosophical and technical failures in the construction of a Friendly AI leading to an Unfrendily AI posing a enormous global risk. He also discusses how a Friendly AI could help decreasing the others Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce others risks, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001): One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first more precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002): Analysis AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004): Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless Decision Theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory doesn`t have a plausible solution: Newcomb’s Problem, Solomon’s Problems and Prisoner’s Dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011): Discusses the Complexity of values: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.
  • “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008), : A pioneer compilation of Cognitive Biases – systematic deviations from rationality – influencing our judgment of risks with the potential to inflict serious damage to human well-being on a global scale, threaten could kill more than millions of human lives called Global Catastrophic Risks (e.g.: volcanic eruptions, pandemic infections, nuclear accidents, worldwide tyrannies, out-of-control scientific experiments, climatic changes, cosmic hazards and economic collapse). It is a book chapter from a larger tome analyzing those risks.
  • “AI as a Positive and Negative Factor in Global Risk. (2008)”, : Chapter of the same book of the previous paper, it analysis possible philosophical and technical failures in the construction of a Friendly AI leading to an Unfrendily AI posing a enormous global risk. He also discusses how a Friendly AI could help decreasing the others Global Risks discussed in the book. Finally, because a powerful AI could go from been a Global Risk to help reduce others risks, he argues that researching such topic is extremely important.
  • "Creating Friendly AI"(2001), : One of the first articles to address the challenges in designing the features and cognitive architecture required to produce a benevolent - "Friendly" - Artificial Intelligence . It also gives one of the first more precise definitions of terms such as Friendly AI and Seed AI.
  • "Levels of Organization in General Intelligence" (2002), : Analysis AGI through its decomposition in five subsystems, successive levels of functional organization: Code, sensory modalities, concepts, thoughts, and deliberation. It also discusses some advantages artificial minds would have, such as the possibility of Recursive self-improvement.
  • "Coherent Extrapolated Volition"(2004), : Presents the difficulties and possible solutions for incorporating friendliness into an AGI. It proposes that making an AGI doing what we tell it to could be dangerous, since we don`t know what we want. Instead we should program the AGI to do what we want, predicting what the vectorial sum of an idealized version of us would want, "if we knew more, thought faster, were more the people we wished we were, had grown up farther together”. He calls this the coherent extrapolated volition of humankind, or CEV.
  • "Timeless Decision Theory" (2010): Describes Timeless Decision Theory,” an extension of causal decision networks that compactly represents uncertainty about correlated computational processes and represents the decision maker as such a process”. It solves many problems which Causal Decision Theory doesn`t have a plausible solution: Newcomb’s Problem, Solomon’s Problems and Prisoner’s Dilemma.
  • Value Systems are Required to Realize Valuable Futures" (2011).

    : Discusses the Complexity of values: we can’t come up with a simple rule or description that sums up all human values. It analyses how this problem makes it difficult to build a valuable future.

Eliezer Yudkowsky is a research fellow of the Singularity Institute for Artificial Intelligence - which he co-founded in 2001. He is mainly concernconcerned with the obstacles and importance of developing a Friendly AI and, such as a reflective decision theory -that would lay a foundation for describing fully recursive self modifying agents that retain stable preferences while rewriting their source code. He also co-founded Less Wrong, writing most part of The Sequences, long sequences of posts dealing with epistemology, AGI, metaethics, rationality and so on.

Eliezer Yudkowsky is a research fellow of the Singularity Institute for Artificial Intelligence - which he co-founded in 2001. He is mainly concern with the obstacles and importance of developing a Friendly AI and a reflective decision theory - a foundation for describing fully recursive self modifying agents that retain stable preferences while rewriting their source code. He also co-founded Less Wrong, writing most part of The Sequences, long sequences of posts dealing with on epistemology, AGI, metaethics, rationality and so on.

Eliezer Yudkowsky is onea research fellow of the foundersSingularity Institute for Artificial Intelligence - which he co-founded in 2001. He is mainly concern with the obstacles and importance of developing a Overcoming BiasFriendly AI and a reflective decision theory - a foundation for describing fully recursive self modifying agents that retain stable preferences while rewriting their source code. He also co-founded Less Wrong, writing most part of The Sequences, long sequences of posts dealing with on epistemology, AGI, metaethics, rationality and so on.

He has published several articles, including: “Cognitive Biases Potentially Affecting Judgment of Global Risks” (2008), “AI as a Positive and Negative Factor in Global Risk. (2008)”, "Creating Friendly AI"(2001), "Levels of Organization in General Intelligence" (2002), "Coherent Extrapolated Volition"(2004), "Timeless Decision Theory" (2010) and Less Wrong. His official webpage is hereValue Systems are Required to Realize Valuable Futures" (2011).