Papers by Federico Benitez
arXiv (Cornell University), May 23, 2024
Brains can process sensory information from different modalities at astonishing speed; this is su... more Brains can process sensory information from different modalities at astonishing speed; this is surprising as already the integration of inputs through the membrane causes a delayed response. Neuronal recordings in vitro reveal a possible explanation for the fast processing through an advancement of the output firing rates of individual neurons with respect to the input, a concept which we refer to as prospective coding. The underlying mechanisms of prospective coding, however, is not completely understood. We propose a mechanistic explanation for individual neurons advancing their output on the level of single action potentials and instantaneous firing rates. Using the Hodgkin-Huxley model, we show that the spike generation mechanism can be the source for prospective (advanced) or retrospective (delayed) responses with respect the underlying somatic voltage. A simplified Hodgkin-Huxley model identifies the sodium inactivation as a source for the prospective firing, controlling the timing of the neuron's output as a function the voltage and its derivative. We also consider a slower spike-frequency adaptation as a mechanisms that generates prospective firings to inputs that undergo slow temporal modulations. In general, we show that adaptation processes at different time scales can cause advanced neuronal responses to time-varying inputs that are modulated on the corresponding time scales.
arXiv (Cornell University), Mar 25, 2024
How physical networks of neurons, bound by spatio-temporal locality constraints, can perform effi... more How physical networks of neurons, bound by spatio-temporal locality constraints, can perform efficient credit assignment, remains, to a large extent, an open question. In machine learning, the answer is almost universally given by the error backpropagation algorithm, through both space and time. However, this algorithm is well-known to rely on biologically implausible assumptions, in particular with respect to spatio-temporal (non-)locality. Alternative forwardpropagation models such as real-time recurrent learning only partially solve the locality problem, but only at the cost of scaling, due to prohibitive storage requirements. We introduce Generalized Latent Equilibrium (GLE), a computational framework for fully local spatio-temporal credit assignment in physical, dynamical networks of neurons. We start by defining an energy based on neuron-local mismatches, from which we derive both neuronal dynamics via stationarity and parameter dynamics via gradient descent. The resulting dynamics can be interpreted as a real-time, biologically plausible approximation of backpropagation through space and time in deep cortical networks with continuous-time neuronal dynamics and continuously active, local synaptic plasticity. In particular, GLE exploits the morphology of dendritic trees to enable more complex information storage and processing in single neurons, as well as the ability of biological neurons to phase-shift their output rate with respect to their membrane potential, which is essential in both directions of information propagation. For the forward computation, it enables the mapping of time-continuous inputs to neuronal space, effectively performing a spatio-temporal convolution. For the backward computation, it permits the temporal inversion of feedback signals, which consequently approximate the adjoint variables necessary for useful parameter updates.
Critics of Artificial Intelligence posit that artificial agents cannot achieve consciousness even... more Critics of Artificial Intelligence posit that artificial agents cannot achieve consciousness even in principle, because they lack certain necessary conditions for consciousness present in biological agents. Here we highlight arguments from a neuroscientific and neuromorphic engineering perspective as to why such a strict denial of consciousness in artificial agents is not compelling. We argue that the differences between biological and artificial brains are not fundamental and are vanishing with progress in neuromorphic architecture designs mimicking the human blueprint. To characterise this blueprint, we propose the conductor model of consciousness (CMoC) that builds on neuronal implementations of an external and internal world model while gating and labelling information flows. An extended Turing test (eTT) lists criteria on how to separate the information flow for learning an internal world model, both for biological and artificial agents. While the classic Turing test only assesses external observables (i.e., behaviour), the eTT also evaluates internal variables of artificial brains and tests for the presence of neuronal circuitries necessary to act on representations of the self, the internal and the external world, and potentially, some neural correlates of consciousness. Finally, we address ethical issues for the design of such artificial agents, formulated as an alignment dilemma: if artificial agents share aspects of consciousness, while they (partially) overtake human intelligence, how can humans justify their own rights against growing claims of their artificial counterpart? We suggest a tentative human-AI deal according to which artificial agents are designed not to suffer negative affective states but in exchange are not granted equal rights to humans.
Disputatio
This work explores issues with the eliminativist formulation of ontic structural realism. An onto... more This work explores issues with the eliminativist formulation of ontic structural realism. An ontology that totally eliminates objects is found lacking by arguing, first, that the theoretical frameworks used to support the best arguments against an object-oriented ontology (quantum mechanics, relativity theory, quantum field theory) can be seen in every case as physical models of empty worlds, and therefore do not represent all the information that comes from science, and in particular from fundamental physics, which also includes information about local interactions between objects. Secondly, by giving a critical assessment of the role of symmetries in these fundamental physical theories; and, lastly, by warning about unfounded metaphysical assumptions. An argument is made for a moderate form of structural realism instead, one in which objects play the fundamental role of representing symmetries and bearing their conserved charges, and of participating in the network of interactions...
Disputatio
This work deals with obstacles hindering a metaphysics of laws of nature in terms of dispositions... more This work deals with obstacles hindering a metaphysics of laws of nature in terms of dispositions, i.e., of fundamental properties that are causal powers. A recent analysis of the principle of least action has put into question the viability of dispositionalism in the case of classical mechanics, generally seen as the physical theory most easily amenable to a dispositional ontology. Here, a proper consideration of the framework role played by the least action principle within the classical image of the world allows us to build a consistent metaphysics of dispositions as charges of interactions. In doing so we develop a general approach that opens the way towards an ontology of dispositions for fundamental physics also beyond classical mechanics.
International Studies in the Philosophy of Science
Le réalisme scientifique est la position métaphysique qui soutient qu’il y a quelque chose dans n... more Le réalisme scientifique est la position métaphysique qui soutient qu’il y a quelque chose dans nos meilleures théories scientifiques qui correspond à la réalité indépen- dante de l’esprit. Par exemple, si ma théorie des phénomènes électriques explique les choses dans le monde en supposant l’existence d’électrons, alors il doit y avoir des objets dans le monde jouant le rôle que jouent les électrons dans ma théorie. étant donné l’immense succès de certaines de nos théories scientifiques en termes de prédiction, d’explication et de contrôle des phénomènes, il semblerait erroné de ne pas croire qu’elles décrivent des aspects du monde réel. Cependant, bien qu’intuitif à première vue, le réalisme scientifique peut être con- testé sur plusieurs fronts, notamment par la question du changement des théories: nous avons plusieurs exemples historiques de théories qui pendant un certain temps ont été considérées comme faisant partie de la meilleure description de la réalité, mais cela s’est av...
Un grand nombre de problemes ouverts parmi les plus importants en mecanique statistique sont lies... more Un grand nombre de problemes ouverts parmi les plus importants en mecanique statistique sont lies aux systemes hors de l'equilibre thermique. Dans ce travail, nous utilisons des methodes de theorie de champs pour etudier certains de ces systemes. Pour ce faire, nous introduisons une representation de type theorie de champs pour les systemes d'interet, ainsi que le formalisme specifique utilise partout dans ce travail, le groupe de renormalisation non perturbatif (NPRG). Ce formalisme a emerge dans les dernieres annees comme un moyen tres efficace pour etudier les systemes fortement correles, et il a ete applique avec succes a certains problemes dans et hors de l'equilibre thermique. Avant de traiter les systemes qui nous interessent, nous developpons de nouveaux outils et methodes dans le cadre du NPRG, et nous les testons dans le cas relativement simple d'une theorie de champ scalaire, appartenant a la classe d'universalite d'Ising. Celui ci nous permet d...
Synthese
In this work we discuss issues of ontological commitment towards one of the most important exampl... more In this work we discuss issues of ontological commitment towards one of the most important examples of contemporary fundamental science: the standard model of particle physics. We present a new form of selective structural realism, which uses as its basis the distinction between what have been called framework and interaction theories. This allows us to advance the ongoing debate about the ontological status of (quasi-)particles and quantum fields, by emphasising the distinction between quantum field theory serving as a framework, and the standard model itself, which we argue is an interaction theory embedded within this framework. Following a discussion of what ontological commitments corresponds to each of these two classes, we argue that some of the previous proposals in the literature might have been misguided by the blending of quantum field theory and the standard model into an undifferentiated unity, and defend a moderate form of object realism with respect to particle-like e...
Foundations of Physics, 2021
Science & Education
Scientific research is a human endeavour, performed by communities of people. Disproportionate fo... more Scientific research is a human endeavour, performed by communities of people. Disproportionate focus on only some of the features related to this obvious fact has been used to discredit the reliability of scientific knowledge and to relativize its value when compared with knowledge stemming from other sources. This epistemic relativism is widespread nowadays and is arguably dangerous for our collective future, as the threat of climate change and its denialism clearly shows. In this work, we argue that even though the social character of science is indeed real, it does not entail epistemic relativism with respect to scientific knowledge, but quite the opposite, as there are several characteristic behaviours of this specific human community that were built to increase the reliability of scientific outputs. Crucially, we believe that present-day scientific education is lacking in the description and analysis of these particularities of the scientific community as a social group and that further investing in this area could greatly improve the possibilities of critical analysis of the often very technical issues that the citizens and future citizens of our modern societies have to confront.
The Astrophysical Journal
ulating discussions. We acknowledge support from the Swiss National Science Foundation via the NC... more ulating discussions. We acknowledge support from the Swiss National Science Foundation via the NCCR PlanetS. Contributions H.D. conceived the idea of linking the moon formation impact to the Earth mantle heterogeneity and planned the project. C.R. built the equation of state (EOS) library and F.B. prepared the EOS lookup table. H.D. and C.R. incorporated the EOS library into the hydrodynamical code and prepared the initial conditions. H.D. ran the simulations and did the visualisation and interpretation. C.R., L.M., J.S. also helped in the interpretation. M.B. and M.M. contributed to the geodynamic and cosmochemistry argument, respectively. H.D., M.B. and M.M. prepared the manuscript. L.M. reviewed the manuscript and all authors commented on it.
Research in Science Education
The Astrophysical Journal
Foundations of Science
We argue that the distinction between framework and interaction the-ories should be taken careful... more We argue that the distinction between framework and interaction the-ories should be taken carefully into consideration when dealing with the philo-sophical implications of fundamental theories in physics. In particular, conclusions concerning the nature of reality can only be consistently derived from assessing the ontological and epistemic purport of both types of theories. We put forward an epistemic form of realism regarding framework theories, such as Quantum Field Theory. The latter, indeed, informs us about the general properties of quantum fields, laying the groundwork for interaction theories. Yet, concerning interaction theories, we recommend a robust form of ontological realism regarding the enti-ties whose existence is assumed by these theories. As an application, we refer to the case of the Standard Model, so long as it has proved to successfully inform us about the nature of various sorts of fundamental particles making up reality. In short, although we acknowledge that both framework and interaction theories partake in shaping our science-based view of reality, and that neither would do by itself the work we expect them to accomplish together, our proposal for a coher-ent ontology of fundamental entities advances a compromise between two forms of realism about theories in each case.
Physical Review E
We consider the Diffusive Epidemic Process (DEP), a two-species reaction-diffusion process origin... more We consider the Diffusive Epidemic Process (DEP), a two-species reaction-diffusion process originally proposed to model disease spread within a population. This model exhibits a phase transition from an active epidemic to an absorbing state without sick individuals. Field-theoretic analyses suggest that this transition belongs to the universality class of Directed Percolation with a Conserved quantity (DP-C, not to be confused with conserved-directed percolation C-DP, appearing in the study of stochastic sandpiles). However, some exact predictions derived from the symmetries of DP-C seem to be in contradiction with lattice simulations. Here we revisit the field theory of both DP-C and DEP. We discuss in detail the symmetries present in the various formulations of both models. We then investigate the DP-C model using the derivative expansion of the non-perturbative renormalization group formalism. We recover previous results for DP-C near its upper critical dimension dc = 4, but show how the corresponding fixed point seems to no longer exist below d 3. Consequences for the DEP universality class are considered.
Physical Review Letters, 2016
We present some exact results on the behavior of Branching and Annihilating Random Walks, both in... more We present some exact results on the behavior of Branching and Annihilating Random Walks, both in the Directed Percolation and Parity Conserving universality classes. Contrary to usual perturbation theory, we perform an expansion in the branching rate around the non trivial Pure Annihilation model, whose correlation and response function we compute exactly. With this, the non-universal threshold value for having a phase transition in the simplest system belonging to the Directed Percolation universality class is found to coincide with previous Non Perturbative Renormalization Group approximate results. We also show that the Parity Conserving universality class has an unexpected RG fixed point structure, with a PA fixed point which is unstable in all dimensions of physical interest.
Uploads
Papers by Federico Benitez