Prior scholarly approaches to meaningful work have largely fallen into two camps. One focuses on identifying how work can contribute to a meaningful life. The other studies the antecedents and outcomes of workers experiencing their work as meaningful. Neither of these approaches, however, captures what people look for when they seek meaningful work—or so I argue. In this paper, I give a new, commitment-based account of meaningful work by focusing on the reasons people have to choose meaningful work over other (...) options. I draw on philosopher Ruth Chang’s account of voluntarist reasons (reasons that arise from an act of the will) to argue that commitments can create distinctive reasons to pursue certain work. It is the presence of these distinctive reasons that makes work meaningful. (shrink)

We study the structure of families of theories in the language of arithmetic extended to allow these families to refer to one another and to themselves. If a theory contains schemata expressing its own truth and expressing a specific Turing index for itself, and contains some other mild axioms, then that theory is untrue. We exhibit some families of true self-referential theories that barely avoid this forbidden pattern.

We construct a machine that knows its own code, at the price of not knowing its own factivity.

Elementary patterns of resemblance notate ordinals up to the ordinal of Pi^1_1-CA_0. We provide ordinal multiplication and exponentiation algorithms using these notations.

We argue that C. Darwin and more recently W. Hennig worked at times under the simplifying assumption of an eternal biosphere. So motivated, we explicitly consider the consequences which follow mathematically from this assumption, and the infinite graphs it leads to. This assumption admits certain clusters of organisms which have some ideal theoretical properties of species, shining some light onto the species problem. We prove a dualization of a law of T.A. Knight and C. Darwin, and sketch a decomposition result (...) involving the internodons of D. Kornet, J. Metz and H. Schellinx. A further goal of this paper is to respond to B. Sturmfels’ question, “Can biology lead to new theorems?”. (shrink)

Reinhardt’s conjecture, a formalization of the statement that a truthful knowing machine can know its own truthfulness and mechanicalness, was proved by Carlson using sophisticated structural results about the ordinals and transfinite induction just beyond the first epsilon number. We prove a weaker version of the conjecture, by elementary methods and transfinite induction up to a smaller ordinal.

Decision theory has had a long-standing history in the behavioural and social sciences as a tool for constructing good approximations of human behaviour. Yet as artificially intelligent systems (AIs) grow in intellectual capacity and eventually outpace humans, decision theory becomes evermore important as a model of AI behaviour. What sort of decision procedure might an AI employ? In this work, I propose that policy-based causal decision theory (PCDT), which places a primacy on the decision-relevance of predictors and simulations of agent (...) behaviour, may be such a procedure. I compare this account to the recently-developed functional decision theory (FDT), which is motivated by similar concerns. I also address potentially counterintuitive features of PCDT, such as its refusal to condition on observations made at certain times. (shrink)

The Gamer’s Dilemma consists of three intuitively plausible but conflicting assertions: (i) Virtual murder is morally permissible. (ii) Virtual child molestation is morally forbidden. (iii) There is no relevant moral difference between virtual murder and virtual child molestation in computer games. Numerous attempts to resolve (or dissolve) the Gamer’s Dilemma line the field of computer game ethics. Mostly, the phenomenon is approached using expressivist argumentation: Reprehensible virtual actions express something immoral in their performance but are not immoral by themselves. Consequentialists, (...) on the other hand, claim that the immorality of virtual actions arises from their harmful consequences. I argue that both approaches have serious difficulties meeting the moral challenge posed by the Gamer’s Dilemma. They tend to confuse the morality of in-game actions either with the morality of their real-world counterparts or with the morality of games as objects. Following this critical analysis, I will develop a Kantian argument and defend it against two objections. So far, deontological responses to the Gamer’s Dilemma have been sought in vain. Yet, with Kant, its moral challenge can be met by looking at the gamer’s reasons. From this perspective, the Gamer’s Dilemma is based on a false assumption: the moral status of gaming acts does not derive from a normative equation with their real-world counterparts but only from their justifications. (shrink)

Fragmentalism was originally introduced as a new A-theory of time. It was further refined and discussed, and different developments of the original insight have been proposed. In a celebrated paper, Jonathan Simon contends that fragmentalism delivers a new realist account of the quantum state—which he calls conservative realism—according to which: the quantum state is a complete description of a physical system, the quantum state is grounded in its terms, and the superposition terms are themselves grounded in local goings-on about the (...) system in question. We will argue that fragmentalism, at least along the lines proposed by Simon, does not offer a new, satisfactory realistic account of the quantum state. This raises the question about whether there are some other viable forms of quantum fragmentalism. (shrink)

We present a game mechanic called pseudo-visibility for games inhabited by non-player characters (NPCs) driven by reinforcement learning (RL). NPCs are incentivized to pretend they cannot see pseudo-visible players: the training environment simulates an NPC to determine how the NPC would act if the pseudo-visible player were invisible, and penalizes the NPC for acting differently. NPCs are thereby trained to selectively ignore pseudo-visible players, except when they judge that the reaction penalty is an acceptable tradeoff (e.g., a guard might accept (...) the penalty in order to protect a treasure because losing the treasure would hurt even more). We describe an RL agent transformation which allows RL agents that would not otherwise do so to perform some limited self-reflection to learn the training environments in question. (shrink)

Do we live in a computer simulation? I will present an argument that the results of a certain experiment constitute empirical evidence that we do not live in, at least, one type of simulation. The type of simulation ruled out is very specific. Perhaps that is the price one must pay to make any kind of Popperian progress.

The verb ‘to know’ can be used both in ascriptions of propositional knowledge and ascriptions of knowledge of acquaintance. In the formal epistemology literature, the former use of ‘know’ has attracted considerable attention, while the latter is typically regarded as derivative. This attitude may be unsatisfactory for those philosophers who, like Russell, are not willing to think of knowledge of acquaintance as a subsidiary or dependent kind of knowledge. In this paper we outline a logic of knowledge of acquaintance in (...) which ascriptions like ‘Mary knows Smith’ are regarded as formally interesting in their own right, remaining neutral on their relation to ascriptions of propositional knowledge. The resulting logical framework, which is based on Hintikka’s modal approach to epistemic logic, provides a fresh perspective on various issues and notions at play in the philosophical debate on acquaintance. (shrink)

In recent years, some have claimed that a Darwinian perspective will revolutionize the study of human society and culture. This project is viewed with disdain and suspicion, on the other hand, by many practicing social scientists. This article seeks to clear the air in this heated debate by dissociating two claims that are too often assumed to be inseparable. The first is the ‘ontological’ claim that Darwinian principles apply, at some level of abstraction, to human society and culture. The second (...) is the more ‘pragmatic’ claim that this observation necessitates substantial changes in the practices of social scientists. Even if some version of the first claim is true, I argue – which I believe is quite likely – the second does not follow. This observation ought to chasten the most overzealous advocates of Darwinian social science, as well as softening the instinctive resistance of many social scientists and historians to the genuine insights enabled by a Darwinian approach. The conclusion discusses these insights, the most important of which is a methodological prescription for normative theory. (shrink)

In response to the spread of COVID-19, governments across the world have, with very few exceptions, enacted sweeping restrictive lockdown policies that impede citizens’ freedom to move, work, and assemble. This paper critically responds to the central arguments for restrictive lockdown legislation. We build our critique on the following assumption: public policy that enjoys virtually unanimous support worldwide should be justified by uncontroversial moral principles. We argue that that the virtually unanimous support in favor of restrictive lockdowns is not adequately (...) justified by the arguments given in favor of them. Importantly, this is not to say that states ought not impose restrictive lockdown measures, but rather that the extent of the acceptance of these measures is not proportionate to the strength of the arguments for lockdowns. -/- We begin by exploring the case for restrictive lockdowns. We first argue that several of the principles that are used to justify the lockdowns yield unexpectedly revisionary implications for other political problems that many would be unwilling to accept. We then outline what we consider the strongest argument for a lockdown—namely, that its net welfare benefits are great enough to defeat the moral presumption against restricting citizens’ civil liberties to move, work, and assemble. However, we give a number of reasons for doubting that the lockdown’s net welfare benefits are, in fact, sufficiently high to defeat the presumption against it. (shrink)

In recent years, the number of studies examining mind wandering has increased considerably, and research on the topic has spread widely across various domains of psychological research. Although the term “mind wandering” has been used to refer to various cognitive states, researchers typically operationalize mind wandering in terms of “task-unrelated thought” (TUT). Research on TUT has shed light on the various task features that require people’s attention, and on the consequences of task inattention. Important methodological and conceptual complications do persist, (...) however, in current investigations of TUT. As we argue, these complications may be dampening the development of a more nuanced scientific account of TUT. Here, we outline three of the more prominent methodological and conceptual complications in the literature on TUT, and we discuss potential directions for researchers to take as they move forward in their investigations of TUT. (shrink)

The Generality Problem is widely recognized to be a serious problem for reliabilist theories of justification. James R. Beebe's Statistical Solution is one of only a handful of attempted solutions that has garnered serious attention in the literature. In their recent response to Beebe, Julien Dutant and Erik J. Olsson successfully refute Beebe's Statistical Solution. This paper presents a New Statistical Solution that countenances Dutant and Olsson's objections, dodges the serious problems that trouble rival solutions, and retains the theoretical virtues (...) that made Beebe's solution so attractive in the first place. There indeed exists a principled, rigorous, conceptually sparse, and plausible solution to the Generality Problem: it is the New Statistical Solution. (shrink)

The third Meditation is typically understood to contain two a posteriori arguments for the existence of God. The author focuses on the second argument, where Descartes proves the existence of God partly in virtue of proving that Descartes cannot be the cause of himself. To establish this, Descartes argues that if he were the cause of himself, then he would endow himself with any conceivable perfection. The justification for this claim is that bringing about a substance is more difficult than (...) creating an attribute, so anything that can do the former can do the latter. While current explanations of this justification are either implausible or inadequate, the author argues that this principle derives support from a scholastic distinction between being-as-such and determinate being. With this distinction in view, we can make sense of Descartes's argument without appealing to ambiguous or inadequate notions. (shrink)

In this paper, I have argued that whatever might be said about his attack on other German philosophers, Santayana’s attack on Kant, despite its subtlety, its force and its intelligence, is fundamentally misguided. Teasing out where Santayana’s attack rests on misunderstandings of Kant’s philosophy is a useful exercise: it is useful for Kantians, for it gives us a chance to show Kant at his best; it is useful for Santayana scholars, for it reminds us that Santayana, for all his brilliance, (...) was not infallible; and it is useful more generally, for the mistakes Santayana makes about Kant are, perhaps in part because of Santayana’s well-deservedly wide influence, still prevalent today. (shrink)

The development of the science of poverty has largely been driven by the need to define more precisely what poverty is, as well as to provide theoretical and empirical criteria for identifying those who suffer from it. This thesis focuses on a notable response to these and related questions: the conception and measure of poverty by the British sociologist Peter Townsend. Townsend defines poverty as relative deprivation caused by lack of resources. This conception, along with his corresponding cut-off measure, constitutes (...) some of the key components of his theory of poverty. The theory discusses what poverty is, its causes, and how it can be eradicated, and is detailed in Poverty in the United Kingdom: A Survey of Household Resources and Standards of Living (1979), his magnum opus. The primary aim of this thesis is to interpret the theory from Townsend’s adherence to the value-free ideal. He pursued this ideal in response to what he perceived as the pervasive and deleterious influence of moral and political values on measures, theories, and policies on poverty throughout its scientific and political history. I argue that to be free from such influence, Townsend aimed to be as guided by epistemic values as possible, which resulted in a systematic theory of poverty. -/- The thesis is structured as follows: In Chapter 1, I situate Townsend’s conception and measurement of poverty within the context of both the recent history of scientific poverty lines and the most important approaches to measuring the phenomenon. Chapter 2 discusses the systematic nature of the theory, including its role as a middle-range theory that bridges broad-range theories with empirical data. This connection is achieved through several elements: a conception of poverty; hypotheses and anthropological observations related to the elements of the conception; measures to test these hypotheses; an explanation of poverty; and reports on how the research was conducted. Furthermore, Townsend’s theory aligns with Bradburn et al.’s (2017) measurement theory, which asserts that a good measure must include a conception of the relevant phenomenon, a representation, and procedures to gather data. Additionally, all three components must be supported by arguments showing that they fit together properly. -/- In the remaining chapters, I present and discuss the main elements of the theory: his conception of poverty as relative deprivation caused by lack of resources, along with the related hypotheses and observations (Chapter 3); the representations of this conception and the procedures for data collection, processing, as well as ensuring their reliability (Chapter 4); and the explanation of poverty (Chapter 5). I conclude by presenting what Townsend considered the “policy prescriptions” of his theory. Despite its potential shortcomings, I also argue that Townsend’s theory fulfills an important epistemic value: fruitfulness. (shrink)

I present an argument that for any computer-simulated civilization we design, the mathematical knowledge recorded by that civilization has one of two limitations. It is untrustworthy, or it is weaker than our own mathematical knowledge. This is paradoxical because it seems that nothing prevents us from building in all sorts of advantages for the inhabitants of said simulation.

We provide an intuitive motivation for the hyperreal numbers via electoral axioms. We do so in the form of a Socratic dialogue, in which Protagoras suggests replacing big-oh complexity classes by real numbers, and Socrates asks some troubling questions about what would happen if one tried to do that. The dialogue is followed by an appendix containing additional commentary and a more formal proof.

Jennifer Saul argues that the evidence from the literature on implicit biases entails a form of skepticism. In this paper, I argue that Saul faces a dilemma: her argument is either self-defeating, or it does not yield a skeptical conclusion. For Saul, both results are unacceptable; thus, her argument fails.

Several decades of work in both philosophy and psychology acutely highlights our limitations as individual inquirers. One way to recognize these limitations is to defer to experts: roughly, to form one’s beliefs on the basis of expert testimony. Yet, as has become salient in the age of Brexit, Trumpist politics, and climate change denial, people are often mistrustful of experts, and unwilling to defer to them. It’s a trope of highbrow public discourse that this unwillingness is a serious pathology. But (...) to what extent is this trope accurate? Answering this requires us to settle both a normative question—under exactly what conditions ought we to defer to experts?—and an empirical question—under what conditions are people willing to defer to experts? The first question has been investigated primarily by philosophers; the second, primarily by psychologists. Yet there is little work integrating these literatures and putting together their results. The aim of this review article is to begin this task, enabling us to begin reaching conclusions about how much real practices of deference diverge from the ideal. We present an opinionated guide to relevant work from both philosophy and psychology, and note places where the literature has important gaps. (shrink)

There is an old meta-philosophical worry: very roughly, metaphysical theories have no observational consequences and so the study of metaphysics has no value. The worry has been around in some form since the rise of logical positivism in the early twentieth century but has seen a bit of a renaissance recently. In this paper, I provide an apology for metaphysics in the face of this kind of concern. The core of the argument is this: pure mathematics detaches from science in (...) much the same manner as metaphysics and yet it is valuable nonetheless. The source of value enjoyed by pure mathematics extends to metaphysics as well. Accordingly, if one denies that metaphysics has value, then one is forced to deny that pure mathematics has value. The argument places an added burden on the sceptic of metaphysics. If one truly believes that metaphysics is worthless (as some philosophers do), then one must give up on pure mathematics as well. (shrink)

In this paper I defend the traditional interpretations of Kant’s Formula of a Law of Nature from recent attacks leveled by Faviola Rivera-Castro, James Furner, Ido Geiger, Pauline Kleingeld and Sven Nyholm. After a short introduction, the paper is divided into four main sections. In the first, I set out the basics of the three traditional interpretations, the Logical Contradiction Interpretation, the Practical Contradiction Interpretation and the Teleological Contradiction Interpretation. In the second, I examine the work of Geiger, Kleingeld and (...) Nyholm: these three commentators reject the traditional interpretations entirely, but I argue that this rejection is ill-founded. In the third and fourth, I take a detailed look at Furner’s work, work in which he seeks to revise (rather than reject) the traditional interpretations. I argue that, despite his more modest aims, Furner’s revision is also ill-founded. (shrink)

Loyalty is considered central to people’s moral life, yet little is known about how people think about what it means to be loyal. We used a prototype approach to understand how loyalty is represented in Colombia and the United States and how these representations mediate attributions of loyalty and moral judgments of loyalty violations. Across 7 studies (N = 1,984), we found cross-cultural similarities in the associative meaning of loyalty (Study 1) but found differences in the centrality of features associated (...) with loyalty (Study 2) and the latent structure of loyalty representations (Study 3). Colombians represent loyalty in terms of more general moral characteristics, while US participants represent loyalty in terms of interpersonal commitment, both in contrast with current approaches to loyalty. By comparing representations of loyalty and honesty, we establish that difference in loyalty conceptualizations reflect a different way of thinking about loyalty rather than morality more generally (Study 4). Further, Colombians attributed greater loyalty to individuals with general moral characteristics compared to participants from the United States sample (Study 5) and were more likely to classify non-loyal values as loyalty-related (Study 6). While the centrality of prototypical features predicts categorizing norm violations as loyalty-related, differences in prototypical structure account for differences in the severity of moral judgment for such violations (Study 7), which suggests that loyalty representations have similar functions even though these representations differ in structural characteristics. (shrink)

Are the real numbers rich enough to measure intelligence? We generalize a result of Alexander and Hutter about the so-called Legg-Hutter intelligence measures of reinforcement learning agents. Using the generalized result, we exhibit a paradox: in one particular version of the Legg-Hutter intelligence measure, certain agents all have intelligence 0, even though in a certain sense some of them outperform others. We show that this paradox disappears if we vary the Legg-Hutter intelligence measure to be hyperreal-valued rather than real-valued.

In recent years evolutionary psychologists have developed and defended the Massive Modularity Hypothesis, which maintains that our cognitive architecture—including the part that subserves ‘central processing’ —is largely or perhaps even entirely composed of innate, domain-specific computational mechanisms or ‘modules’. In this paper I argue for two claims. First, I show that the two main arguments that evolutionary psychologists have offered for this general architectural thesis fail to provide us with any reason to prefer it to a competing picture of the (...) mind which I call the Library Model of Cognition. Second, I argue that this alternative model is compatible with the central theoretical and methodological commitments of evolutionary psychology. Thus I argue that, at present, the endorsement of the Massive Modularity Hypothesis by evolutionary psychologists is both unwarranted and unmotivated. (shrink)

Iris Murdoch’s The Sovereignty of Good—especially the first essay, “The Idea of Perfection”—is often associated with a critique of a certain picture of agency and its proper place in ethical thought. There is implicit in this critique, however, an alternative, much richer one. I propose a reading of Murdochian agency in terms of the continuous activity of cultivating and refining a distinctive practical standpoint, and I apply this reading to her account of moral progress. For Murdoch moral progress depends on (...) transcending egoism and achieving clear perception of a normatively-saturated reality, but it would be a mistake to think of egoism in terms of selfishness, or clarity in terms of altruism. Rather, I argue, Murdochian moral progress requires overcoming socially-conditioned and often ideological forms of alienation, and making the social conditions that inform our practical standpoints self-conscious. (shrink)

One shortcoming of the chain rule is that it does not iterate: it gives the derivative of f(g(x)), but not (directly) the second or higher-order derivatives. We present iterated differentials and a version of the multivariable chain rule which iterates to any desired level of derivative. We first present this material informally, and later discuss how to make it rigorous (a discussion which touches on formal foundations of calculus). We also suggest a finite calculus chain rule (contrary to Graham, Knuth (...) and Patashnik's claim that "there's no corresponding chain rule of finite calculus"). (shrink)

Inspired by recent progress in multi-agent Reinforcement Learning (RL), in this work we examine the collective intelligent behaviour of theoretical universal agents by introducing a weighted mixture operation. Given a weighted set of agents, their weighted mixture is a new agent whose expected total reward in any environment is the corresponding weighted average of the original agents' expected total rewards in that environment. Thus, if RL agent intelligence is quantified in terms of performance across environments, the weighted mixture's intelligence is (...) the weighted average of the original agents' intelligences. This operation enables various interesting new theorems that shed light on the geometry of RL agent intelligence, namely: results about symmetries, convex agent-sets, and local extrema. We also show that any RL agent intelligence measure based on average performance across environments, subject to certain weak technical conditions, is identical (up to a constant factor) to performance within a single environment dependent on said intelligence measure. (shrink)

Scholars have often thought that a monistic reading of Aristotle’s definition of the human good – in particular, one on which “best and most teleios virtue” refers to theoretical wisdom – cannot follow from the premises of the ergon argument. I explain how a monistic reading can follow from the premises, and I argue that this interpretation gives the correct rationale for Aristotle’s definition. I then explain that even though the best and most teleios virtue must be a single virtue, (...) that virtue could in principle be a whole virtue that arises from the combination of all the others. I also clarify that the definition of the human good aims at capturing the nature of human eudaimonia only in its primary case. (shrink)

Dasgupta poses a serious challenge to realism about natural properties. He argues that there is no acceptable explanation of why natural properties deserve the value realists assign to them and are consequently absent of value. In response, this paper defines and defends an alternative non-explanatory account of normativity compatible with realism. Unlike Lewis and Sider, who believe it is sufficient to defend realism solely on realist terms, I engage with the challenge on unfriendly grounds by revealing a tu quoque. Dasgupta (...) and anti-realists face a similar challenge to that directed against realism: one that not only undermines the objection to realism by legitimising non-explanatory normativity but leaves them facing a significant dilemma. (shrink)

Jaakko Hintikka proposed treating objectual perception sentences, such as “Alice sees Bob,” as de re propositional perception sentences. Esa Saarinen extended Hintikka’s idea to eventive perception sentences, such as “Alice sees Bob smile.” These approaches, elegant as they may be, are not philosophically neutral, for they presuppose, controversially, that the content of all perceptual experiences is propositional in nature. The aim of this paper is to propose a formal treatment of objectual and eventive perception sentences that builds on Hintikka’s modal (...) approach to propositional attitude ascriptions while avoiding controversial assumptions on the nature of perceptual experiences. Despite being simple and theoretically frugal, our approach is powerful enough to express a variety of interesting philosophical views about propositional, objectual, and eventive perception sentences, thus enabling the study of their inferential relationships. (shrink)

The conventional wisdom regarding the aims and shortcomings of Kantian constructivism is mistaken. The aim of metaethical constructivism is not to provide a naturalistic account of the objectivity of normative facts by deriving substantive morality from a conception of agency so thin as to be uncontroversial (a task at which it is generally regarded to have failed). Its aim is to explain the “grip” that normative facts have on us—to avoid what I call the problem of normative alienation. So understood, (...) Kantian constructivism faces two problems: that determinate normative facts cannot be derived from agency and that its individualistic conception of agency cannot account for the sociality of morality. I propose and elaborate a social conception of agency that is better able to address the latter problem while still avoiding normative alienation, and evaluate two different strategies for responding to the former problem. (shrink)

In their thought-provoking paper, Legg and Hutter consider a certain abstrac- tion of an intelligent agent, and define a universal intelligence measure, which assigns every such agent a numerical intelligence rating. We will briefly summarize Legg and Hutter’s paper, and then give a tongue-in-cheek argument that if one’s goal is to become more intelligent by cultivating music appreciation, then it is bet- ter to use classical music (such as Bach, Mozart, and Beethoven) than to use more recent pop music. The (...) same argument could be adapted to other media: books beat films, card games beat first-person shooters, parables beat dissertations, etc. We leave it to the reader to decide whether this argument tells us something about classical music, something about Legg-Hutter intelligence, or something about both. (shrink)

Samuel Kerstein’s recent (2013) How To Treat Persons is an ambitious attempt to develop a new, broadly Kantian account of what it is to treat others as mere means and what it means to act in accordance with others’ dignity. His project is explicitly nonfoundationalist: his interpretation stands or falls on its ability to accommodate our pretheoretic intuitions, and he does an admirable job of handling carefully a range of well fleshed out and sometimes subtle examples. In what follows, (...) I shall give a quick summary of the chapters and then say two good things about the book and one critical thing. (shrink)

Aristotle analyses a large range of objects as composites of matter and form. But how exactly should we understand the relation between the matter and form of a composite? Some commentators have argued that forms themselves are somehow material, that is, forms are impure. Others have denied that claim and argued for the purity of forms. In this paper, I develop a new purist interpretation of Metaphysics Z.10-11, a text central to the debate, which I call 'hierarchical purism'. I argue (...) that hierarchical purism can overcome the difficulties faced by previous versions of purism as well as by impurism. Roughly, on hierarchical purism, each composite can be considered and defined in two different ways: From the perspective of metaphysics, composites are considered only insofar as they have forms and defined purely formally. From the perspective of physics, composites are considered insofar as they have forms and matter and defined with reference to both. Moreover, while the metaphysical definition is a definition in the strict sense of 'definition', the physical definition is a definition in a loose sense. Analogous points hold for intelligible composites and geometry. Finally, neither sort of definitional practice implies that, for Aristotle, forms are impure. (shrink)

Can an AGI create a more intelligent AGI? Under idealized assumptions, for a certain theoretical type of intelligence, our answer is: “Not without outside help”. This is a paper on the mathematical structure of AGI populations when parent AGIs create child AGIs. We argue that such populations satisfy a certain biological law. Motivated by observations of sexual reproduction in seemingly-asexual species, the Knight-Darwin Law states that it is impossible for one organism to asexually produce another, which asexually produces another, and (...) so on forever: that any sequence of organisms (each one a child of the previous) must contain occasional multi-parent organisms, or must terminate. By proving that a certain measure (arguably an intelligence measure) decreases when an idealized parent AGI single-handedly creates a child AGI, we argue that a similar Law holds for AGIs. (shrink)

We define a notion of the intelligence level of an idealized mechanical knowing agent. This is motivated by efforts within artificial intelligence research to define real-number intelligence levels of compli- cated intelligent systems. Our agents are more idealized, which allows us to define a much simpler measure of intelligence level for them. In short, we define the intelligence level of a mechanical knowing agent to be the supremum of the computable ordinals that have codes the agent knows to be codes (...) of computable ordinals. We prove that if one agent knows certain things about another agent, then the former necessarily has a higher intelligence level than the latter. This allows our intelligence no- tion to serve as a stepping stone to obtain results which, by themselves, are not stated in terms of our intelligence notion (results of potential in- terest even to readers totally skeptical that our notion correctly captures intelligence). As an application, we argue that these results comprise evidence against the possibility of intelligence explosion (that is, the no- tion that sufficiently intelligent machines will eventually be capable of designing even more intelligent machines, which can then design even more intelligent machines, and so on). (shrink)

Legg and Hutter, as well as subsequent authors, considered intelligent agents through the lens of interaction with reward-giving environments, attempting to assign numeric intelligence measures to such agents, with the guiding principle that a more intelligent agent should gain higher rewards from environments in some aggregate sense. In this paper, we consider a related question: rather than measure numeric intelligence of one Legg- Hutter agent, how can we compare the relative intelligence of two Legg-Hutter agents? We propose an elegant answer (...) based on the following insight: we can view Legg-Hutter agents as candidates in an election, whose voters are environments, letting each environment vote (via its rewards) which agent (if either) is more intelligent. This leads to an abstract family of comparators simple enough that we can prove some structural theorems about them. It is an open question whether these structural theorems apply to more practical intelligence measures. (shrink)

In this paper I provide a novel argument for scientific realism. In contrast to most recent defenses of SR, my defense of SR does not rely on the no-miracles argument. Instead, I take a more unconventional approach: I focus on the different kinds of justification available to different individuals in relation to different kinds of propositions. I maintain that this alternative focus shows that most people are warranted in believing many propositions about unobservables. The paper is divided into three main (...) sections. In the first, I rehearse the main moves in the recent debate about SR. In the second, I argue that the discussion in section one enables us to see that most of the arguments in the recent debate about SR mistake their target: instead of being about SR, they are about meta-SR. I argue that what I call the JJ-principle should be rejected and, further, that if the JJ-principle is rejected, then meta-SR may be cleaved from SR. This enables me to advance to a position I call thin realism in the third and final section of the paper. (shrink)

We sometimes fail unwittingly to do things that we ought to do. And we are, from time to time, culpable for these unwitting omissions. We provide an outline of a theory of responsibility for unwitting omissions. We emphasize two distinctive ideas: (i) many unwitting omissions can be understood as failures of appropriate vigilance, and; (ii) the sort of self-control implicated in these failures of appropriate vigilance is valuable. We argue that the norms that govern vigilance and the value of self-control (...) explain culpability for unwitting omissions. (shrink)

According to many philosophers, rationality is, at least in part, a matter of one’s attitudes cohering with one another. Theorists who endorse this idea have devoted much attention to formulating various coherence requirements. Surprisingly, they have said very little about what it takes for a set of attitudes to be coherent in general. We articulate and defend a general account on which a set of attitudes is coherent just in case and because it is logically possible for the attitudes to (...) be jointly satisfied in the sense of jointly fitting the world. In addition, we show how the account can help adjudicate debates about how to formulate various rational requirements. (shrink)

I provide a theory of the metaphysical foundations of identity: an account what grounds facts of the form a=b. In particular, I defend the claim that indiscernibility grounds identity. This is typically rejected because it is viciously circular; plausible assumptions about the logic of ground entail that the fact that a=b partially grounds itself. The theory I defend is immune to this circularity.

Perdurantists think of continuants as mereological sums of stages from different times. This view of persistence would force us to drop the idea that there is genuine change in the world. By exploiting a presentist metaphysics, Brogaard proposed a theory, called presentist four-dimensionalism, that aims to reconcile perdurantism with the idea that things undergo real change. However, her proposal commits us to reject the idea that stages must exist in their entirety. Giving up the tenet that all the stages are (...) equally real could be a price that perdurantists are unwilling to pay. I argue that Kit Fine ’s fragmentalism provides us with the tools to combine a presentist metaphysics with a perdurantist theory of persistence without giving up the idea that reality is constituted by more than purely present stages. (shrink)

In a recent book and an article, Carl Craver construes the relations between different levels of a mechanism, which he also refers to as constitutive relations, in terms of mutual manipulability (MM). Interpreted metaphysically, MM implies that inter-level relations are symmetrical. MM thus violates one of the main desiderata of scientific explanation, namely explanatory asymmetry. Parts of Craver’s writings suggest a metaphysical interpretation of MM, and Craver explicitly commits to constitutive relationships being symmetrical. The paper furthermore explores the option of (...) interpreting MM epistemologically, as a means for individuating mechanisms. It is argued that MM then is redundant. MM should therefore better be abandoned. (shrink)

After generalizing the Archimedean property of real numbers in such a way as to make it adaptable to non-numeric structures, we demonstrate that the real numbers cannot be used to accurately measure non-Archimedean structures. We argue that, since an agent with Artificial General Intelligence (AGI) should have no problem engaging in tasks that inherently involve non-Archimedean rewards, and since traditional reinforcement learning rewards are real numbers, therefore traditional reinforcement learning probably will not lead to AGI. We indicate two possible ways (...) traditional reinforcement learning could be altered to remove this roadblock. (shrink)