Alexei Grinbaum

Des robots domestiques se font délateurs, des agents conversationnels injurient leurs interlocuteurs. Pire encore : des systèmes informatiques participent aux conflits humains et parfois même les provoquent. Le 18 mars 2018, un véhicule autonome de la société Uber a tué une femme qui traversait la rue dans une ville de l'Arizona. Ce fut la première mort d'un piéton provoquée par un algorithme.

Qui est responsable ? La réponse à cette question compte parmi les défis les plus urgents à relever dans notre rapport aux technologies numériques. Mais il ne s'agit pas de savoir comment rendre l'intelligence artificielle bienveillante. Il s'agit de faire en sorte qu'elle ne se substitue pas à l'homme en tant qu'agent moral. Seul le recours au hasard, et ceci dès sa conception, peut libérer la machine de la responsabilité qu'on veut lui faire porter.

« Nous avons appris récemment qu'une large partie de la spécificité de la physique quantique réside dans la manière dont elle décrit, non un objet isolé, mais deux, trois ou plusieurs objets, ainsi que la règle de leur composition. Autrement dit, que signifie prendre deux entités et les traiter comme une seule ? Au travers de récits anciens et de métaphores contemporaines, ce livre relate trois moments clefs de la réflexion sur cette question. »

A recent report on ethical questions of artificial intelligence (AI) stipulates that ‘the potential misuses of AI technology suggest a downside to openly sharing all new capabilities and algorithms by default’. I could not agree more. However, I do not believe that the need for opacity only stems from the fear of malicious use. Opacity plays a greater role in ethics, most prominently in achieving or maintaining trust. Trust in the procedures used for acquiring information relies on their non-epistemic, non-transparent aspects. Chance becomes an instrument for obtaining it. The following is a review of the central argument of my book published in Russian under the title Machina Delatrix and soon available in French as Satanas Ex Machina, in which I make an extended case for chance as a value for AI systems.

If physics is a science that unveils the fundamental laws of nature, then the appearance of mathematical concepts in its language can be surprising or even mysterious. This was Eugene Wigner’s argument in 1960. I show that another approach tophysical theory accommodates mathematics in a perfectly reasonable way. To explore unknown processes or phenomena, one builds a theory from fundamental principles, employing them as constraints within a general mathematical framework. The rise ofsuch theories of the unknown, which I call blackbox models, drives home the unsurprising effectiveness of mathematics. I illustrate it on the examples of Einstein’s principle theories, the S-matrix approach in quantum field theory, effective field theories, and device-independent approaches in quantum information.

Quantum technologies can be presented to the public with or without introducing a strange trait of quantum theory responsible for their non-classical efficiency. Traditionally the message was centered on the superposition principle, while entanglement and properties such as contextuality have been gaining ground recently. A less theoretical approach is focused on simple protocols that enable technological applications. It results in a pragmatic narrative built with the help of the resource paradigm and principle-based reconstructions. I discuss the advantages and weaknesses of these methods. To illustrate the importance of new metaphors beyond the Schr\"odinger cat, I briefly describe a non-mathematical narrative about entanglement that conveys an idea of some of its unusual properties. If quantum technologists are to succeed in building trust in their work, they ought to provoke an aesthetic perception in the public commensurable with the mathematical beauty of quantum theory experienced by the physicist. The power of the narrative method lies in its capacity to do so.

Les agents conversationnels (chatbots en anglais) ont fait parler d’eux récemment, notamment le Tay de Microsoft. Ils nous ramènent à une tradition ancienne qui, des golems à la science fiction, de Turing aux services clientèle robotisés, a généré toutes les inquiétudes et tous les fantasmes. Les chatbots dialoguent avec les humains dans leur langue naturelle. Alexei Grinbaum nous explique que ce dialogue touche à l’essence de l’homme.

À supposer que, sur le plan politique, le type de gouvernement qu’exige le code ne soit pas totalement nouveau, il s’agit d’identifier, parmi les instruments et les méthodes de pouvoir, ceux adéquats et adaptés au code. L’ubiquité du code produit une structuration socio-anthropologique nouvelle. Elle joue ainsi un rôle analogue à celui des armes ou de l’argent. Toutefois, les moyens les plus pertinents pour gouverner le code se rapprochent plutôt des instruments de pouvoir relevant de la sphère religieuse. Foucault définissait le pouvoir pastoral comme une gouvernementalité qui ne s’exerce pas sur un territoire, mais sur des individus qu’elle vise, à la fois tous et chacun. L’individualisation et la bienfaisance forment un type de rapport fondamental entre les hommes et un pouvoir transcendant. Ce dernier, pour mériter d’être appelé pastoral, doit se manifester par son zèle et son application indéfinie à la vie des hommes sur lesquels il veille. Foucault pensait, certes, à la religion et à la puissance divine, mais son schéma reste valide au-delà de cet exemple. Omniprésent dans la vie de chacun, le code constitue dans la société de l’information une transcendance par rapport au niveau des utilisateurs. Contrairement à la divinité qui fait objet de croyance non empirique, le code est matériellement écrit par des milliers d'informaticiens, pour qui la signification de chaque ligne est supposée être claire. Cependant, même s’il est fabriqué par des experts, le code agit sur l’utilisateur comme un pouvoir opaque quoique directement présent. Il se caractérise par l’ubiquité, l’opacité, l’individualisation, la promesse du bien et la veille continue sur l’utilisateur ainsi que la non-territorialité. Or, ce sont des propriétés permettant d’établir une analogie entre le pouvoir du code et le pouvoir pastoral. Limitée mais instructive, cette analogie permet de se demander si, et comment, l'homme imite les valeurs du code.

Dirac sought an interpretation of mathematical formalism in terms of physical entities and Einstein insisted that physics should describe "the real states of the real systems". While Bell inequalities put into question the reality of states, modern device-independent approaches do away with the idea of entities: physics is not built of physical systems. Focusing on the correlations between operationally defined inputs and outputs, device-independent methods promote a view more distant from conventional theory than Einstein's 'principle theories' were from 'constructive theories'. On the examples of indefinite causal orders and almost quantum correlations, we ask a puzzling question: if physical theory is not about systems, then what is it about? The answer given by the device-independent models is that physics is about languages. In moving away from the information-theoretic reconstructions of quantum theory, this answer marks a new conceptual development in the foundations of physics.

Some mathematical theories in physics justify their explanatory superiority over earlier formalisms by the clarity of their postulates. In particular, axiomatic reconstructions drive home the importance of the composition rule and the continuity assumption as two pillars of quantum theory. Our approach sits on these pillars and combines new mathematics with a testable prediction. If the observer is defined by a limit on string complexity, information dynamics leads to an emergent continuous model in the critical regime. Restricting it to a family of binary codes describing ‘bipartite systems,’ we find strong evidence of an upper bound on bipartite correlations equal to 2.82537. This is measurably different from the Tsirelson bound. The Hilbert space formalism emerges from this mathematical investigation as an effective description of a fundamental discrete theory in the critical regime.

If the technological situation is unique, the ways to understand the contemporary moral condition are not. We link it to age-old questions: in fact, the power promised by technol-ogy only establishes a new form of human finitude. In the face of this continuity of the fundamental moral condition, we examine a number of alternative ways of thinking about the basis of responsible innovation, exploring the metaphors of quasi-parental and political responsibilities, as well as the place of virtue in innovation and the role of cul-tural narratives in helping us understand the limits of responsible innovation.

Parmi les questions d’ordres social, économique, politique, éthique et métaphysique liées au développement des biotechnologies, la biologie de synthèse suscite certes quelques interrogations nouvelles. Mais surtout, elle en réactualise d’anciennes. La confiance que le citoyen place dans le scientifique n’est plus aujourd’hui acquise par défaut. Il en va de même pour les technologues, les ingénieurs et les industriels.
Dans ce dialogue renouvelé entre la science et la société s’impose la nécessité d’élaborer sans attendre un cadre normatif qui devra, autant que faire se peut, anticiper les problématiques futures que soulèvera une diffusion à grande échelle des produits issus de la biologie de synthèse.

Quantum logical axiomatic systems for quantum theory usually include a postulate that a lattice under consideration is orthomodular. We propose a derivation of orthomodularity from an information-theoretic axiom. This provides conceptual clarity and removes a long-standing puzzle about the meaning of orthomodularity.

... Observer of 22 February 2004, The New York Times of 26 February 2004) makes use of the first and the second sources of uncertainty. ... The catastrophic change in question is precisely the change due to the passage of a tipping point, to an abrupt fall of the complex ecosystem ...

The argument from naturalness is widely employed in contemporary quantum field theory. Essentially a formalized aesthetic criterion, it received a meaning in the debate on the Higgs mechanism, which goes beyond aesthetics. We follow the history of technical definitions of fine tuning at the scale of electroweak symmetry breaking. It is argued that they give rise to a special interpretation of probability, which we call Gedankenfrequency. By extension of its original meaning, the argument from naturalness is used to compare different models beyond the Standard Model. We show that in this case naturalness cannot be defined objectively. Rather, it functions as socio-historical heuristics in particle physics and it contributes to the advent of a probabilistic version of Popper's falsificationism.

We give reasons for the importance of old narratives, including myths, in ethical thinking about science and technology. On the example of a legend about creating artificial men we explore the side effects of having too much success and the problem of intermediate social status of bioengineered artefacts.

Different observers do not have to agree on how they identify a quantum system. We explore a condition based on algorithmic complexity that allows a system to be described as an objective "element of reality". We also suggest an experimental test of the hypothesis that any system, even much smaller than a human being, can be a quantum mechanical observer.

Information-theoretic derivations of the formalism of quantum theory have recently attracted much attention. We analyze the axioms underlying a few such derivations and propose a conceptual framework in which, by combining several approaches, one can retrieve more of the conventional quantum formalism.

Modern microscopes create a capacity to see and act at the scale where unassisted human senses are powerless. Images of nanoscale phenomena represent a world that effectively intervenes in human life while remaining distant and ineffable. This combination of an unbridgeable distance between man and technology with a real power of the latter over the human condition is characteristic, not only of nanotechnology, but also of the theology of sacred icons that mediate in the knowledge of divine reality. We draw an anthropological analogy between nanotechnological images and sacred icons, and we employ it to analyze the functioning of such images in society. The questions that we study include: a) the art of mediation in representing the ineffable and its necessity; b) the effects provoked by such images; c) primitive comprehension of nanoscale images. This work has repercussions for a wide variety of disciplinary accounts and practices, from the anthropologist and the philosopher to the laboratory scientist, the art critic, and the engineer who conceives and designs nanoscale instruments.

This article is concerned with predictions of future events, such as technological achievements and changes in the human condition that they will bring about. Cognitive barriers arise when human agents are either asked or forced to make judgments and decisions with respect to unknown singular events. This article argues that barriers such as an aversion to not knowing and the impossibility to believe trump expert and ordinary human reasoning. These barriers apply to nanotechnology. To avoid undesired societal effects arising from them, this essay proposes a set of steps designed to foster responsible public dialogue.

Different observers do not have to agree on how they identify a quantum system. We explore a condition based on algorithmic complexity that allows a system to be described as an objective "element of reality". We also suggest an experimental test of the hypothesis that any system, even much smaller than a human being, can be a quantum mechanical observer.

What belongs to quantum theory is no more than what is needed for its derivation. We argue for an approach focusing on reconstruction rather than interpretation of quantum mechanics and analyze several examples of reconstruction. We submit that reconstruction advances our understanding of quantum theory irrespective of one's ontological stance.

Every user, but not the entire society, gains assurance through initiation, better: gets a new voice. This voice arranges labels and numbers into meaningful sentences. It announces a new mystery of meaning: our participation in the truth of the computer. In code we trust.