Learning in the age of artificial intelligence

"The sadness of artificial intelligence is that it is without artifice, and therefore without intelligence."

The challenges

Artificial intelligence poses 5 major challenges for training. The first concerns the way we look at it. Either artificial intelligence is a prosthesis that fills a gap, in which case we still consider that learning consists in being supplemented as if we were knowledge-disabled. Or we see artificial intelligence as an orthosis, a kind of exoskeleton that enhances our ability to think, in which case we see learning as a continuous increase in our capacities. We probably need to revisit the myths that guide our representations, and perhaps move away from the Terminator and transhumanism of Hollywood films to imagine Calinator, an artificial intelligence that stimulates our desires and increases the release of dopamine in our brains through the continuous experience of mini digital and creative feats.

The second issue is ethical. It concerns the way in which AIs are designed and progressively encroach on our decision-making capacities, based on parameters we know nothing about. First, they finish our sentences when we type text, then place us under their influence in a "nudge" capitalism that builds paths and glides from an argument to a site, from a proposition to a choice, and gradually guides us from interface to interface in our ways of relating to the world. Ergonomists refer to the "affordance" of an environment that induces our behavior. One of the consequences of this is the risk of weakening our conscious choices, which in terms of learning leads to a loss of self-direction of our learning - essential, according to psychologist Albert Bandura, to the creation of intrinsic motivation. Conversely, the more I feel in control of my learning choices, the greater my persistence in the act of learning.

The third issue is economic. When ChatGPT burst onto the market, it won over 3 million users in 1 week, and passed the 100 million mark in 1 month. Never before seen. It's an upheaval for the major digital operators that we thought were well established, for example, for search engines. It's a twofold move into a market for intellectual techniques and a way of accelerating the construction of reasoning, as well as into the creation and fluidification of content. However, Noam Chomsky, the linguist and father of generative grammar, reminds us that "language is an inner human power that allows us to generate and understand, thanks to a finite number of rules, an infinite number of propositions that express thought". This is very different from the production of "probable character strings" in generative AI. In other words, there is no such thing as artificial intelligence, but rather an artifice of intelligence.

The fourth issue is pedagogy. It's as if a "fossil pedagogy" made up of sediments accumulated by the river of knowledge, which year after year deposits a layer of knowledge from which everyone draws to live their world, is now being challenged. Generative AI simply draws on the past and rearranges data to produce probable information. AI has the good fortune to remind us that learning is more than remembering and articulating the facts of the past. It's also, and above all, about drawing on one's singular experience and projecting one's desires into the future to build the world to come. AI speaks to us of two root emotions. Fear leads educational models to reproduce the past. As a result, the first reflex has been to be wary of AIs and possible cheating on exams, where copy-and-paste would conceal a lack of individual reasoning. Here, learning is about dominating nature, combining information and controlling the means by which knowledge is constructed. The second root emotion is the joy and movement of exploring the world. Probably, by pushing us to relieve ourselves of the task of remembering, AI invites us to move towards our human singularity and our power of exploration.

The fifth challenge is technological, and concerns the inclusion of AI in social and economic models. In this respect, it's worth remembering that AI consumes more energy than the brain to process calculations, a little like a leaf and the principle of photosynthesis, because living organisms, unlike machines, operate thriftily and continuously recycle energy. Synaptic circuits consume less energy than power plants. So there is an electrical cost to be paid for producing calculations and data whose use is hardly frugal. Another aspect of this issue is the water consumption used to cool data centers. Finally, by being everywhere, AI has become much more than a tool, it's a context. AI is invasive; it is like water, slipping in everywhere and playing on the porosity between systems. A techno-conference is taking hold without our knowledge, as screens increasingly form a screen between us and the world.

The opportunities

Spectacular job losses are predicted, with the spectre of machines replacing humans as looms once replaced weavers. Unless new professions emerge, such as "prompt designer", and others take on roles yet to be invented, or only repetitive tasks disappear, while creative tasks develop for more creative professional profiles. Let's recall the story of vellum, monks and the invention of printing. Vellum was the calfskin on which monks wrote sacred texts. They were so expensive to produce that no gaps were left between the words. This popularized the reading of texts during meals in the refectory. One of the monks would take on the mental burden of reading the text aloud for everyone.

With the invention of paper and mobile printing presses, texts became more airy, margins appeared at the edges of pages, individual reading became possible, notes could be taken and the conditions were ripe for distance and critical thinking. What if AI freed up brain time for something else? There are two possibilities. Either an increase in screen time, with already more than 5 hours a day, which risks disconnecting us from the concrete; or creative, meditative time, an opportunity to rethink our lives and the world. In training, it's possible to push towards the creative option. This involves further exploration of the sensitive world of dreams, imagination, emotions and so on.

Incidentally, we are now sliding towards a phenomenological approach to training, in which singular experience takes on its rightful place. Normative frames of reference and the neuro-cognitivist paradigm are fading into the background, to be replaced by a welcoming of human singularity and all approaches that evoke experience. After the triptych of objective/content/pedagogical method inherited from the industrial era that marks our engineering practices, other approaches are beginning to emerge in which the trainer's mediation changes register. Whereas they used to be content experts, they now take a more frequent interest in learning processes. He becomes a facilitator, helping groups of learners to construct more meaning in a world fraught with a variety of expected transitions.

In this way, vocational training continues to evolve towards learning. This apprenticeship, nourished by the desire to learn, leads to approaches in which movement and the development of critical thinking take on greater importance. Training is less a setting in which the learner is placed, and more an environment that he or she is encouraged to co-construct. This leads to open pedagogies in which AI can play a role through its power to aid creativity and reasoning. Alongside cold, impersonal knowledge, warm, contextualized knowledge takes its rightful place. The fragility of human knowledge has its place, because "it's the flaw that lets the light through".

The questions we are asked

We have Paleolithic emotions, medieval institutions and divine technology; this is the dilemma we face. At the same time as AI is humanizing itself with a humanized appearance, avatars, voices, intonations and expressions, we are mechanizing ourselves. We've become button-pushers for organic robots. Why continue this self-fulfilling prophecy and accept this reduction? Incidentally, generative AI pushes us to distinguish between knowledge as a result of learning and learning as a process. It's as if consulting the algorithmic oracle takes the place of learning, when in fact it's merely an exposure to data it doesn't even understand.

AI ignores ethics, cannot conceive of concepts and is incapable of self-grasping. MIT's moral machine helps us to pose ethical dilemmas, but for the moment, only programmers are committed to solving them according to their own codes of values. Once again, the human experience is at the heart of the matter, and training is called upon to integrate a more critical spirit, while systems gain in capacity to create fertile knowledge, i.e. knowledge capable of generating new knowledge and not just reproducing the past. Let's conclude with Paul Virilio: if the industrial era marked the warming of the planet, the AI era is characterized by the warming of minds. We need to take a cool-headed approach to this technology, which promises as much as it worries.

Sources
Albert Bandura (2003). Self-efficacy. The feeling of personal efficacy. Deboeck
https:// journals.openedition.org/osp/741
Author of Apprendre à l'ère de l'intelligence artificielle to be published in January 2024 https://www.esf-scienceshumaines.fr/accueil/445-apprendre-a-l-ere-de-l-intelligence-artificielle.html
Philomag. Chatgpt, Chomsky and the banality of evil
https:// www.philomag.com/articles/chatgpt-chomsky-et-la-banalite-du-mal