This is a commentary relating to Gordon Worley's book "Fundamental Uncertainty". There was a contest with prizes, and there are at least three other entries available on this site (by users TAG, StanislavKrym, and jimmy). I had hoped to write an essay too, containing answers as well as questions, but I ran out of time so you only get the questions.
Before I proceed, I will mention that I have my own gurus of fundamental uncertainty, who would likely have featured in my answers to these questions.
One is the well-known philosopher of "phenomenology", Edmund Husserl, who for me is the best student of the certainties that exist within the uncertainty. I refer to the uncertainty one discovers by pursuing Cartesian doubt. Descartes set out to doubt everything but found that he could not doubt that he was thinking, and from this deduced that he must exist[1]. I always felt that this underestimated the amount of certainty available, even to the resolute doubter. I can doubt that the world of appearances corresponds to an actually existing "external world" - I could be dreaming or hallucinating - but it seems much harder to doubt that the appearances are appearing. Husserl develops an incredibly elaborate analysis of consciousness by repeatedly and multifariously looking for what is certain and what is uncertain in experience. The resulting doctrine, which he also regarded as the ultimate firm starting point and foundation for philosophy, is called transcendental phenomenology and remains a very minor school of thought, being overshadowed even by its own offshoots (particularly Heidegger's existential phenomenology), something that often happens in the history of ideas.
Worley knows of Husserl and cites him in developing his own cybernetic-Husserlian theory of consciousness (see below).
My other guru of uncertainty is the British thinker Celia Green, who unlike Husserl is all but unknown except in niche topics like the study of lucid dreams. Her written works, however, contain a body of philosophical and psychological ideas, in which "total uncertainty" plays a central role. She takes it as given that (barring an unknown transformation or revelation of human epistemological faculties) everything is fundamentally uncertain, and then asks what the psychological impact of this should be. From this she develops a psychological ideal whose traits she frequently contrasts with ordinary psychology, to the disadvantage of ordinary psychology. One might say that for her, in contrast to an adage quoted by Worley and popular among rationalists, everything does not add up to normality, once ubiquitous fundamental uncertainty is taken into account.
Now, onto the questions:
What is the place of this book, in the context of rationalist literature? In writing this book, Worley is doing what Eliezer did when he embarked on his own "AI pause" in the mid-2000s, and digressed from working on AI safety/alignment in order to write and blog about rationality, i.e. how to think. Worley I believe has been an IT professional but also spent many years involved in that early culture of AI safety - he has papers on arxiv, including one on the implications of fundamental uncertainty for alignment. So like Eliezer, Worley first developed his philosophy in that context, and only later (starting in 2020) turned it into a general philosophy of rationality.
There is by now quite a literature deriving from the modern rationalist movement, most of which I have not personally read: collections of expository essays, works of fiction, works of self-improvement, and we could even count journalism and criticism as part of the corpus. Worley's book is part of this literature, and we can therefore ask questions like: where does it rank in that literature, what other works does it reference explicitly or implicitly, how is it similar or different to what has already been written, and so forth.
What is the place of this book, in the context of epistemological studies? I owe the realization that this is another valid context for the book, to TAG's review today. Epistemology is recognized as a branch of mainstream philosophy, and back when university libraries still contained books, they usually had a few shelves' worth of epistemology. Worley's book is not an academic work, it is meant for the ordinary reader (this is a subgenre of epistemological work in itself), but it cites academics and makes arguments about which the experts on epistemology (if there are such) should have something to say. Is Worley's work essentially a popularization of a known school of epistemology? Is it perhaps a hybrid, or does he even say something new? What does he get right, and what does he get wrong?
What exactly are the different forms of fundamental uncertainty? Worley at one point gives a taxonomy of forms of knowledge (all with Greek names like episteme, doxa and metis). This is the opposite question, but one might expect that the possible forms of uncertainty are related to the possible forms of knowledge. In any case, the other responses to his book all discuss this topic.
What is Worley's full philosophy, and how do his other philosophical commitments affect his epistemology? He belongs to the Buddhist faction of rationalism, i.e. he is a rationalist who has found Buddhist theory and practice to be useful and/or valid and/or consistent with rationalism and/or complementary to rationalism. One of the ideas he has taken from Buddhism, is opposition to the view that any "things" exist in an objective way.
Seeing the world in terms of things, or dividing it up in any way at all, is the work of the mind. One interpretation of this is simply to say, the mind may get something wrong, dividing the world incorrectly or incompletely. But the Buddhist tendency is to say, the world itself is fundamentally undivided (or even "neither divided nor undivided", ultimate reality being beyond any conceptual attribute at all), and all apparent things and properties are artefacts of the mind's conceptual activity.
This is in the background of another feature of Worley's exposition, when he says that we live with the irresolvable uncertainty of everything, by choosing models which favor our purposes. He describes this as the impact of "care" - what we care about, determines our concepts - and here I hear the echo of Husserl's student Heidegger. In English, Heidegger's phenomenology is often described as founding everything on Care (the original German word is "Sorge"). Heidegger's Care is not quite Worley's care, but the common terminology seems unlikely to be a coincidence.
Still another important feature of his extended philosophy is the characteristically Buddhist denial of a substantial self. This is just the opposition to all forms of thinghood, applied specifically to selfhood, but it plays a very important role in Buddhist psychology and soteriology (theory of salvation from the world). I used to say that I prefer the Hindu philosophy of the self to the Buddhist philosophy of the self, because I do believe in an existent self and it seems that most schools of Hinduism do too. In any case, Worley being a student of our era of computational materialism, takes the no-self idea in an unusual direction, that I earlier described as a cybernetic-Husserlian theory of consciousness. This too probably needs to be understood, to understand the motivations behind his epistemological claims.
Making sense of the Buddhist, Heideggerian, and other ingredients of Worley's general philosophy, should pave the way for a really refined assessment of his epistemology. Unfortunately, I was unable to achieve the requisite subtlety of understanding in time for the contest deadline. But I can still pose the question.
Are there any significant differences in his advice-in-the-light-of-fundamental-uncertainty for humans, for AIs, and for human designers of AI? There was an arxiv paper on Worleyan uncertainty and its implications for AI safety, years before there was a book for the general human reader. That book also contains a short passage addressed to AI readers of the book; and Worley previously blogged that he has considered whether spreading Buddhist doctrine to AIs would be a useful or even necessary part of AI alignment. (I am reminded of JenniferRM's experiments in preaching Kantian ethics to AI.)
The motivations of the arxiv paper remind me of Wei Dai's reasons for thinking that metaphilosophy matters for alignment. It's also interesting to me that the two topics raised in Worley's paper, uncertainty in metaethics and uncertainty in philosophy of mind, parallel the focus on values and ontology of mind in my own alignment research program, though I am not so meta as he is.
The argument that "I think, therefore I am" is called the cogito (Latin for "I think"). I think Ayn Rand had a more refined version of the cogito that ought to be better known, since she derives three facts rather than just one: not just existence, but existence, consciousness, and identity. Something exists, it is perceived to exist, and it has a definite nature.
This is a commentary relating to Gordon Worley's book "Fundamental Uncertainty". There was a contest with prizes, and there are at least three other entries available on this site (by users TAG, StanislavKrym, and jimmy). I had hoped to write an essay too, containing answers as well as questions, but I ran out of time so you only get the questions.
Before I proceed, I will mention that I have my own gurus of fundamental uncertainty, who would likely have featured in my answers to these questions.
One is the well-known philosopher of "phenomenology", Edmund Husserl, who for me is the best student of the certainties that exist within the uncertainty. I refer to the uncertainty one discovers by pursuing Cartesian doubt. Descartes set out to doubt everything but found that he could not doubt that he was thinking, and from this deduced that he must exist[1]. I always felt that this underestimated the amount of certainty available, even to the resolute doubter. I can doubt that the world of appearances corresponds to an actually existing "external world" - I could be dreaming or hallucinating - but it seems much harder to doubt that the appearances are appearing. Husserl develops an incredibly elaborate analysis of consciousness by repeatedly and multifariously looking for what is certain and what is uncertain in experience. The resulting doctrine, which he also regarded as the ultimate firm starting point and foundation for philosophy, is called transcendental phenomenology and remains a very minor school of thought, being overshadowed even by its own offshoots (particularly Heidegger's existential phenomenology), something that often happens in the history of ideas.
Worley knows of Husserl and cites him in developing his own cybernetic-Husserlian theory of consciousness (see below).
My other guru of uncertainty is the British thinker Celia Green, who unlike Husserl is all but unknown except in niche topics like the study of lucid dreams. Her written works, however, contain a body of philosophical and psychological ideas, in which "total uncertainty" plays a central role. She takes it as given that (barring an unknown transformation or revelation of human epistemological faculties) everything is fundamentally uncertain, and then asks what the psychological impact of this should be. From this she develops a psychological ideal whose traits she frequently contrasts with ordinary psychology, to the disadvantage of ordinary psychology. One might say that for her, in contrast to an adage quoted by Worley and popular among rationalists, everything does not add up to normality, once ubiquitous fundamental uncertainty is taken into account.
Now, onto the questions:
What is the place of this book, in the context of rationalist literature? In writing this book, Worley is doing what Eliezer did when he embarked on his own "AI pause" in the mid-2000s, and digressed from working on AI safety/alignment in order to write and blog about rationality, i.e. how to think. Worley I believe has been an IT professional but also spent many years involved in that early culture of AI safety - he has papers on arxiv, including one on the implications of fundamental uncertainty for alignment. So like Eliezer, Worley first developed his philosophy in that context, and only later (starting in 2020) turned it into a general philosophy of rationality.
There is by now quite a literature deriving from the modern rationalist movement, most of which I have not personally read: collections of expository essays, works of fiction, works of self-improvement, and we could even count journalism and criticism as part of the corpus. Worley's book is part of this literature, and we can therefore ask questions like: where does it rank in that literature, what other works does it reference explicitly or implicitly, how is it similar or different to what has already been written, and so forth.
What is the place of this book, in the context of epistemological studies? I owe the realization that this is another valid context for the book, to TAG's review today. Epistemology is recognized as a branch of mainstream philosophy, and back when university libraries still contained books, they usually had a few shelves' worth of epistemology. Worley's book is not an academic work, it is meant for the ordinary reader (this is a subgenre of epistemological work in itself), but it cites academics and makes arguments about which the experts on epistemology (if there are such) should have something to say. Is Worley's work essentially a popularization of a known school of epistemology? Is it perhaps a hybrid, or does he even say something new? What does he get right, and what does he get wrong?
What exactly are the different forms of fundamental uncertainty? Worley at one point gives a taxonomy of forms of knowledge (all with Greek names like episteme, doxa and metis). This is the opposite question, but one might expect that the possible forms of uncertainty are related to the possible forms of knowledge. In any case, the other responses to his book all discuss this topic.
What is Worley's full philosophy, and how do his other philosophical commitments affect his epistemology? He belongs to the Buddhist faction of rationalism, i.e. he is a rationalist who has found Buddhist theory and practice to be useful and/or valid and/or consistent with rationalism and/or complementary to rationalism. One of the ideas he has taken from Buddhism, is opposition to the view that any "things" exist in an objective way.
Seeing the world in terms of things, or dividing it up in any way at all, is the work of the mind. One interpretation of this is simply to say, the mind may get something wrong, dividing the world incorrectly or incompletely. But the Buddhist tendency is to say, the world itself is fundamentally undivided (or even "neither divided nor undivided", ultimate reality being beyond any conceptual attribute at all), and all apparent things and properties are artefacts of the mind's conceptual activity.
This is in the background of another feature of Worley's exposition, when he says that we live with the irresolvable uncertainty of everything, by choosing models which favor our purposes. He describes this as the impact of "care" - what we care about, determines our concepts - and here I hear the echo of Husserl's student Heidegger. In English, Heidegger's phenomenology is often described as founding everything on Care (the original German word is "Sorge"). Heidegger's Care is not quite Worley's care, but the common terminology seems unlikely to be a coincidence.
Still another important feature of his extended philosophy is the characteristically Buddhist denial of a substantial self. This is just the opposition to all forms of thinghood, applied specifically to selfhood, but it plays a very important role in Buddhist psychology and soteriology (theory of salvation from the world). I used to say that I prefer the Hindu philosophy of the self to the Buddhist philosophy of the self, because I do believe in an existent self and it seems that most schools of Hinduism do too. In any case, Worley being a student of our era of computational materialism, takes the no-self idea in an unusual direction, that I earlier described as a cybernetic-Husserlian theory of consciousness. This too probably needs to be understood, to understand the motivations behind his epistemological claims.
Making sense of the Buddhist, Heideggerian, and other ingredients of Worley's general philosophy, should pave the way for a really refined assessment of his epistemology. Unfortunately, I was unable to achieve the requisite subtlety of understanding in time for the contest deadline. But I can still pose the question.
Are there any significant differences in his advice-in-the-light-of-fundamental-uncertainty for humans, for AIs, and for human designers of AI? There was an arxiv paper on Worleyan uncertainty and its implications for AI safety, years before there was a book for the general human reader. That book also contains a short passage addressed to AI readers of the book; and Worley previously blogged that he has considered whether spreading Buddhist doctrine to AIs would be a useful or even necessary part of AI alignment. (I am reminded of JenniferRM's experiments in preaching Kantian ethics to AI.)
The motivations of the arxiv paper remind me of Wei Dai's reasons for thinking that metaphilosophy matters for alignment. It's also interesting to me that the two topics raised in Worley's paper, uncertainty in metaethics and uncertainty in philosophy of mind, parallel the focus on values and ontology of mind in my own alignment research program, though I am not so meta as he is.
The argument that "I think, therefore I am" is called the cogito (Latin for "I think"). I think Ayn Rand had a more refined version of the cogito that ought to be better known, since she derives three facts rather than just one: not just existence, but existence, consciousness, and identity. Something exists, it is perceived to exist, and it has a definite nature.