Artificial Neural Networks have reached “grandmaster” and even “super-human” performance across a variety of games, from those involving perfect information, such as Go, to those involving imperfect information, such as “Starcraft”. Such technological developments from artificial intelligence (AI) labs have ushered concomitant applications across the world of business, where an “AI” brand-tag is quickly becoming ubiquitous. A corollary of such widespread commercial deployment is that when AI gets things wrong—an autonomous vehicle crashes, a chatbot exhibits “racist” behavior, automated credit-scoring processes (...) “discriminate” on gender, etc.—there are often significant financial, legal, and brand consequences, and the incident becomes major news. As Judea Pearl sees it, the underlying reason for such mistakes is that “... all the impressive achievements of deep learning amount to just curve fitting.” The key, as Pearl suggests, is to replace “reasoning by association” with “causal reasoning” —the ability to infer causes from observed phenomena. It is a point that was echoed by Gary Marcus and Ernest Davis in a recent piece for theNew York Times: “we need to stop building computer systems that merely get better and better at detecting statistical patterns in data sets—often using an approach known as ‘Deep Learning’—and start building computer systems that from the moment of their assembly innately grasp three basic concepts: time, space, and causality.” In this paper, foregrounding what in 1949 Gilbert Ryle termed “a category mistake”, I will offer an alternative explanation for AI errors; it is not so much that AI machinery cannot “grasp” causality, but that AI machinery (qua computation) cannot understand anything at all. (shrink)

Is the mathematical function being computed by a given physical system determined by the system’s dynamics? This question is at the heart of the indeterminacy of computation phenomenon (Fresco et al. [unpublished]). A paradigmatic example is a conventional electrical AND-gate that is often said to compute conjunction, but it can just as well be used to compute disjunction. Despite the pervasiveness of this phenomenon in physical computational systems, it has been discussed in the philosophical literature only indirectly, mostly with reference (...) to the debate over realism about physical computation and computationalism. A welcome exception is Dewhurst’s ([2018]) recent analysis of computational individuation under the mechanistic framework. He rejects the idea of appealing to semantic properties for determining the computational identity of a physical system. But Dewhurst seems to be too quick to pay the price of giving up the notion of computational equivalence. We aim to show that the mechanist need not pay this price. The mechanistic framework can, in principle, preserve the idea of computational equivalence even between two different enough kinds of physical systems, say, electrical and hydraulic ones. (shrink)

Very plausibly, nothing can be a genuine computing system unless it meets an input-sensitivity requirement. Otherwise all sorts of objects, such as rocks or pails of water, can count as performing computations, even such as might suffice for mentality—thus threatening computationalism about the mind with panpsychism. Maudlin in J Philos 86:407–432, ( 1989 ) and Bishop ( 2002a , b ) have argued, however, that such a requirement creates difficulties for computationalism about conscious experience, putting it in conflict with the (...) very intuitive thesis that conscious experience supervenes on physical activity. Klein in Synthese 165:141–153, ( 2008 ) proposes a way for computationalists about experience to avoid panpsychism while still respecting the supervenience of experience on activity. I argue that his attempt to save computational theories of experience from Maudlin’s and Bishop’s critique fails. (shrink)

A single physical process may often be described equally well as computing several different mathematical functions—none of which is explanatorily privileged. How, then, should the computational identity of a physical system be determined? Some computational mechanists hold that computation is individuated only by either narrow physical or functional properties. Even if some individuative role is attributed to environmental factors, it is rather limited. The computational semanticist holds that computation is individuated, at least in part, by semantic properties. She claims that (...) the mechanistic account lacks the resources to individuate the computations performed by some systems, thereby leaving interesting cases of computational indeterminacy unaddressed. This article examines some of these views, and claims that more cases of computational indeterminacy can be addressed, if the system-environment interaction plays a greater role in individuating computations. A new, long-arm functional strategy for individuating computation is advanced. (shrink)

Is the mathematical function being computed by a given physical system determined by the system’s dynamics? This question is at the heart of the indeterminacy of computation phenomenon (Fresco et al. [unpublished]). A paradigmatic example is a conventional electrical AND-gate that is often said to compute conjunction, but it can just as well be used to compute disjunction. Despite the pervasiveness of this phenomenon in physical computational systems, it has been discussed in the philosophical literature only indirectly, mostly with reference (...) to the debate over realism about physical computation and computationalism. A welcome exception is Dewhurst’s ([2018]) recent analysis of computational individuation under the mechanistic framework. He rejects the idea of appealing to semantic properties for determining the computational identity of a physical system. But Dewhurst seems to be too quick to pay the price of giving up the notion of computational equivalence. We aim to show that the mechanist need not pay this price. The mechanistic framework can, in principle, preserve the idea of computational equivalence even between two different enough kinds of physical systems, say, electrical and hydraulic ones. (shrink)

The question ‘What is computation?’ might seem a trivial one to many, but this is far from being in consensus in philosophy of mind, cognitive science and even in physics. The lack of consensus leads to some interesting, yet contentious, claims, such as that cognition or even the universe is computational. Some have argued, though, that computation is a subjective phenomenon: whether or not a physical system is computational, and if so, which computation it performs, is entirely a matter of (...) an observer choosing to view it as such. According to one view, which we dub bold anti-realist pancomputationalism, every physical object (can be said to) computes every computer program. According to another, more modest view, some computational systems can be ascribed multiple computational descriptions. We argue that the first view is misguided, and that the second view need not entail observer-relativity of computation. At least to a large extent, computation is an objective phenomenon. Construed as a form of information processing, we argue that information-processing considerations determine what type of computation takes place in physical systems. (shrink)

This paper seeks to understand machine cognition. The nature of machine cognition has been shrouded in incomprehensibility. We have often encountered familiar arguments in cognitive science that human cognition is still faintly understood. This paper will argue that machine cognition is far less understood than even human cognition despite the fact that a lot about computer architecture and computational operations is known. Even if there have been putative claims about the transparency of the notion of machine computations, these claims do (...) not hold out in unraveling machine cognition, let alone machine consciousness (if there is any such thing). The nature and form of machine cognition remains further confused also because of attempts to explain human cognition in terms of computation and to model/simulate (aspects of) human cognitive processing in machines. Given that these problems in characterizing machine cognition persist, a view of machine cognition that aims to avoid these problems is outlined. The argument that is advanced is that something becomes a computation in machines only when a human interprets it, which is a kind of semiotic causation. From this it follows that a computing machine is not engaged in a computation unless a human interprets what it is doing; instead, it is engaged in machine cognition, which is defined as a member or subset of the set of all possible mappings of inputs to outputs. The human interpretation, which is a semiotic process, gives meaning to what a machine does, and then what it does becomes a computation. (shrink)

Research is starting to identify correlations between consciousness and some of the spatiotemporal patterns in the physical brain. For theoretical and practical reasons, the results of experiments on the correlates of consciousness have ambiguous interpretations. At any point in time a number of hypotheses co-exist about and the correlates of consciousness in the brain, which are all compatible with the current experimental results. This paper argues that consciousness should be attributed to any system that exhibits spatiotemporal physical patterns that match (...) the hypotheses about the correlates of consciousness that are compatible with the current experimental results. Some computers running some programs should be attributed consciousness because they produce spatiotemporal patterns in the physical world that match those that are potentially linked with consciousness in the human brain. (shrink)

John Searle’s Chinese Room Argument purports to demonstrate that syntax is not sufficient for semantics, and, hence, because computation cannot yield understanding, the computational theory of mind, which equates the mind to an information processing system based on formal computations, fails. In this paper, we use the CRA, and the debate that emerged from it, to develop a philosophical critique of recent advances in robotics and neuroscience. We describe results from a body of work that contributes to blurring the divide (...) between biological and artificial systems; so-called animats, autonomous robots that are controlled by biological neural tissue and what may be described as remote-controlled rodents, living animals endowed with augmented abilities provided by external controllers. We argue that, even though at first sight, these chimeric systems may seem to escape the CRA, on closer analysis, they do not. We conclude by discussing the role of the body–brain dynamics in the processes that give rise to genuine understanding of the world, in line with recent proposals from enactive cognitive science. (shrink)

Do the dynamics of a physical system determine what function the system computes? Except in special cases, the answer is no: it is often indeterminate what function a given physical system computes. Accordingly, care should be taken when the question ‘What does a particular neuronal system do?’ is answered by hypothesising that the system computes a particular function. The phenomenon of the indeterminacy of computation has important implications for the development of computational explanations of biological systems. Additionally, the phenomenon lends (...) some support to the idea that a single neuronal structure may perform multiple cognitive functions, each subserved by a different computation. We provide an overarching conceptual framework in order to further the philosophical debate on the nature of computational indeterminacy and computational explanation. (shrink)