The intelligence behind artificial "intelligence" (AI) is suspect. Admittedly, we are less and less doubtful of its remarkable technical capabilities. On the other hand, we increasingly doubt its... intelligence. For example, we hesitate to talk about genuine intelligence - in the strongest sense of the word - even though Google's AI amazes us by making a hair appointment without us. Of course, it does us a favour, but is it really intelligence? I won't venture to define intelligence. This is a slippery slope. However, I suggest you read an elegant article on the subject, published by Margarida Romero. (1). Yet the current lack of consensus, among technophobes and technophiles alike - not to mention those in between - reveals one thing: intelligence, whether artificial or natural, invites us to think critically.
Who's mimicking who? At first glance, one would be tempted to answer that AI imitates the human mind, its gait, language and reasoning.
Indeed, both weak and strong AI (see Jacques Baudron for distinction) are based on algorithms that allow the machine to act rationally. Whether it fails or succeeds in matching - or even exceeding - natural intelligence, AI has the human as its established benchmark.
That's what the philosopher Pascal Chabot referred to as a "mimesis operative". Chabot does not dissect the concept further, but we find in the philosopher Adriano Fabris a more detailed and accurate demonstration of imitation: "The concept of 'imitation' (mimesis) is a dynamic and relational category. It links two or more elements that have a specific relationship to each other. It is a relationship in which a particular element is recognized as having primacy, insofar as it is taken as a model [...]. In other words, the "copy" is considered to be a copy in so far as it resembles, and increasingly seeks to resemble, something or someone taken as a "model". In this, then, lies the dynamic of imitation. »
Essentially, it is the imitating element (the AI in our case) that mimics the behaviour of the imitated element (the human) according to a programmed method. However, it goes further than that.
Of the human who imitates
Humans can also be made to imitate AI. Suppose he wants to interact with a machine, he must adapt his pronunciation to be understood at first sight by his vocal assistant Siri, Alexa or Google Home. The examples are as numerous as the various interactions between the AI and the human (see on this subject the experiments carried out in cognitive psychology).
I propose the adoption of a new concept to express this reality of the modern human: the technomimicry.
By technomimicry I mean the process of modification and engineering by which humans draw inspiration from the forms, materials, properties, processes and functions of technology more generally, and from technology (and AI) more specifically. It aims at the implantation - conscious and unconscious - of the logic of machines in the sphere of activity of the natural.
A pensive robot. Who imitates who in the end?
Technomimicry breaks with biomimicry because it covets predictability while natural systems tend towards unpredictability. It is animated by the principle of being the artist of one's life rather than being subjected to it as a spectator. On this subject, the works of the science historian Bernadette Bensaude-Vincent pose a fair distinction between biomimicry and technomimicry.
The human, in this technomimetic conjecture, is moving towards the standardization and homogenization of his behavior. He will have to be more and more predictable - and more robotic - so that AI can process the information received quickly. By the way, it is relevant to reflect on the question of the predictability (or not) of a strong AI. No one, not even programmers and engineers, can analyze the reasoning behind the choice of "x" over "y" or the other way around. This is a critical issue.
The implementation of a technomimetic reading makes a "robot becoming human" appear. However, this has been little studied. The counter-intuitive nature of such an observation certainly has something to do with it. The book What Computers Still Can't Do: A Critique of Artificial Reason by Hubert L. Dreyfus provides precise and lucid arguments on this subject.
New perspectives for reflection
Looking at the relationship between humans and AI from the perspective of technomimicry contrasts with most studies on the subject.
Their starting premise: the machine reproduces what the person does. The reverse relationship is therefore unconsciously neglected. However, the philosopher Charles Taylor asked himself the question and ventured into it in his book Human Agency and Language.
It seems that a dynamic approach, i.e. one that is aware of the mutual influence exerted between human and machine, will make it possible to study the ethical question of AI development more pertinently: technological development (and therefore of AI today) must become a means - of human fulfilment - and not an end in itself. The work of Martin Heidegger,Ivan Illich more importantly Gilbert Simondon on technology as a tool for human beings - and not the other way around - are essential here.
A curious encounter
Let's go back to our initial question: who imitates whom? It seems that the imitation goes both ways, but to different degrees: AI imitates the human automatically, while the human imitates AI (almost) automatically.
Certainly, the latter can still preserve its independence from the machine, but for how long? This is a rather worrying question.
In order to be able to use artificial intelligence effectively, humans align - initially consciously and then more unconsciously - their behaviour (and also their thinking) to the more standard AI model. It is a human who robotizes himself by anticipation.
In this context, I note a relatively curious dynamic at work: while scientific research on artificial intelligence is pushing (with great success) towards a rapprochement with natural intelligence, it is the human who is getting closer (through technomimicry) by aligning himself with the principles of communication and AI practices.
The human increased (and decreased)
There would be a price to pay for all this.
Let us take the example of oblivion. It is one of those things that has always characterized us, whether we like it or not. Nevertheless, in the age of AI and digital technology, we forget even the existence of this deficiency. One could say that humans forget forget forgetfulness. Today, everything is saved in databases and memory becomes material, i.e. it becomes (almost) impossible to forget.
It is quite worrying that such changes are not being discussed more. Let us salute here the work of Doueihi who explores such subtle mutations in his work. To the best of my knowledge, the horizon of the increased human is much more controversial than that of the decreased human, and yet...
Thus, it seems to me that the human who imitates AI a little too much (and too closely) may be heading down a dead-end road, that of a humanity both increased and decreased, depending on the point of view.
And it is with this observation that we conclude our brief critical reflection on ethics, the development of AI, and the technomimicry of man.
I propose that we take more ownership of this issue because it affects us all. In fact, the development of AI is not just a matter for "experts" and investors.
We'll all be the judge of that.
Mario Ionuț MaroșanResearch and Teaching Assistant in Political Philosophy - Faculty of Arts and Science - Department of Political Science Université de Montréal