The Brain vs. AI: Chronicle of an Organ Overwhelmed by the Data Deluge

The brain, a predictive machine hungry for data

An organ selected to reduce uncertainty

Contemporary neuroscience describes the brain as a predictive machine. According to the free energy principle formalized by Karl Friston (1959–), the brain constantly seeks to minimize surprise, that is, the gap between its internal predictions and the sensory signals it receives. This theory, called the FEP, unifies perception, action, and learning under a single mathematical framework.

In this model, the hunger for data is not a defect: it is an evolutionary strategy. Each new piece of information makes it possible to refine internal models, improve predictions, and thus better survive. The brain is therefore wired to actively seek information, especially that which reduces uncertainty or signals a reward.

The work of Donald Hebb (1904–1985) on synaptic plasticity showed that this hunger leaves a physical trace: neural connections strengthen or weaken according to their use. The brain is therefore not a passive receptacle, but an organ that constantly reconfigures itself, based on the data it processes. It is this plasticity that makes adaptation possible, but also vulnerability.

The reward circuit and dopamine

Seeking information activates the brain's reward circuit. A small deep area, the ventral tegmental area, sends dopamine to another area, the nucleus accumbens. Result: the brain wants to keep searching. Dopamine does not encode pleasure itself, but the reward prediction error: it signals that something unexpected and positive has just occurred. However, modern digital platforms exploit precisely this mechanism by delivering new, unpredictable, and personalized content.

The result is a feedback loop: each notification, each scroll, each AI-generated response produces a small positive prediction error, which reinforces the search behavior. The brain, designed to economize on information, finds itself exposed to a nearly unlimited flow of stimuli calibrated for its own reward system.

The three cases of reward prediction error

The information deluge: a rupture of scale

Unprecedented volumes

To measure the magnitude of the phenomenon, a few orders of magnitude suffice. In 2024, humanity produced more than 400 exabytes of data, or about 400 billion billion bytes. This volume roughly doubles every two to three years. A growing share of this data is generated by AI, whether text, images, videos, or code.

The human brain, however, has not evolved at this speed. Its conscious processing capacity is estimated at about 50 bits per second, while its sensory systems capture several million bits per second. Most of this flow is filtered, but the filtering itself consumes attentional resources. The gap between the information supply and the cognitive absorption capacity is at the heart of the problem.

The notion of information overload

The concept is not new. As early as 1970, Alvin Toffler (1928–2016) described in Future Shock the psychological shock linked to the acceleration of change. In 1971, Herbert Simon (1916–2001), Nobel laureate in economics, formulated the fundamental idea: in a world rich in information, the scarce resource becomes attention. Information overload therefore does not come from a lack of data, but from an excess of data relative to available attention capacity.

With generative AI, this overload changes in nature. It is no longer only a quantitative excess, but a qualitative excess: content produced on demand, personalized, plausible, and impossible to distinguish from the real without effort. The brain must not only sort, but also understand and verify, which mobilizes costly prefrontal resources.

Documented cognitive impacts

Fragmentation of attention

Numerous studies in cognitive psychology have measured the effects of digital multitasking on attention. The results converge: repeated solicitation by notifications and continuous streams reduces the duration of sustained attention, increases the number of errors, and lengthens the time to resume after an interruption. A phenomenon known as the switch cost.

fMRI shows that digital multitasking excessively solicits the dorsolateral prefrontal cortex, a key region of executive control. In the long term, this overuse could promote decision fatigue and lower tolerance for boredom, a state nevertheless necessary for memory consolidation and creativity.

Saturation of working memory

Working memory, a capacity limited to about four to seven simultaneous elements, is the bottleneck of human cognition. Continuous information flows saturate it permanently, preventing deep encoding into long-term memory. The result is a more superficial memory, more dependent on external supports, and more vulnerable to distractions.

Modification of the reward circuit

The work of Daniel Kahneman (1934–2024) showed the coexistence of two systems of thought: a fast, intuitive system and a slow, analytical one. The information deluge favors the former at the expense of the latter, because it rewards immediate reaction rather than reflection. This shift has consequences for the quality of reasoning, resistance to cognitive biases, and the ability to sustain long intellectual effort.

Effects on sleep and emotional regulation

Prolonged exposure to screens and information flows, especially in the evening, delays falling asleep and fragments sleep. However, sleep is essential for consolidating learning and eliminating cerebral metabolic waste. Chronic sleep deprivation in turn aggravates irritability, anxiety, and difficulty regulating emotions, creating a vicious circle.

Table of documented impacts

The table below summarizes the main observed effects of the information deluge on cognitive functions, as well as the neural mechanisms probably involved.

Neuroplasticity: adaptation or damage?

A double-edged plasticity

The brain is plastic: it reorganizes itself according to use. This property, essential to learning, also means that prolonged exposure to a saturated information environment leaves lasting traces. Imaging studies show modifications of functional connectivity in intensive users of digital platforms, particularly in the circuits of attention and cognitive control.

Should this be seen as damage or adaptation? The answer is nuanced. The brain adapts to its environment: if this environment is saturated and fragmented, it develops fast, superficial processing strategies, but efficient for navigating this flow. The cost appears when these strategies become dominant, at the expense of deep reflection and sustained attention.

The role of cognitive rest

The brain does not only need stimulation: it needs cognitive rest. Moments of mind-wandering, breaks without solicitation, walks without screens activate the default mode network (DMN), involved in memory consolidation, projection into the future, and the construction of the self. The systematic suppression of these moments by a continuous information flow could be one of the deepest effects of the current deluge.

AI and brain: symbiosis, dependence, or submersion?

Cognitive prostheses

AI can also act as cognitive prostheses: they externalize memory, assist reasoning, filter information, translate, summarize. Used as tools, they free attentional resources for higher-value tasks. This is the scenario of augmented symbiosis, in which human and machine complement each other.

But this symbiosis has a condition: the user must retain control of the process. If AI decides what is relevant, summarizes, guides, and personalizes, the user risks gradually losing the skills of sorting, analysis, and memorization. This is the scenario of cognitive dependence, where the tool replaces the function instead of assisting it.

The risk of functional atrophy

In neuroscience, the rule is simple: a function that is little used degrades. Studies on London taxi drivers showed that the posterior hippocampus develops in those who memorize the city map, and regresses after retirement. By analogy, if AI takes charge of memory, orientation, mental calculation, or writing, the corresponding circuits could weaken from lack of training.

This phenomenon, sometimes called cognitive offloading, is not inevitable. It depends on use: an AI that explains, guides, and makes the user work strengthens the circuits; an AI that answers in their place weakens them. The distinction is crucial for education and training.

New scientific frontiers

Brain-machine interfaces and augmented cognition

Brain-machine interfaces (BMIs) are progressing rapidly. They already allow paralyzed patients to control a robotic arm or a cursor by thought. Ultimately, they could offer a direct channel between the brain and AI, bypassing sensorimotor bottlenecks. But they raise major ethical questions: who controls the flow? Who guarantees the integrity of thought?

fMRI studies on the use of generative AI

Since 2023, several teams have been using fMRI to compare brain activity during tasks performed with or without generative AI. The first results suggest reduced activation of working memory and executive control regions when AI provides the answer, and increased activation of verification and vigilance regions. These works, still preliminary, open an essential field of research.

Toward information hygiene

Faced with the deluge, a new discipline is emerging: information hygiene. It consists of actively managing one's exposure to flows, preserving periods of cognitive rest, prioritizing depth over quantity, and regularly training the threatened functions (attention, memory, reasoning). Neuroscience can guide these practices, as it already guides sleep hygiene or physical activity.

Key takeaways

The brain is not a passive machine. Its plasticity allows it to adapt to very different environments, including informationally saturated ones. The question is therefore not whether it will survive the deluge of artificial intelligences, but in what ways and at what functional cost. The challenge is as much scientific as it is cultural and educational.

Natural selection, meanwhile, acts on very long timescales, on the order of several tens of generations. It therefore cannot respond in a few decades to such a rapid technological change. If there is selective pressure, it will probably bear on traits favoring resistance to information overload: better attention regulation, less dependence on immediate dopaminergic rewards, greater tolerance for boredom and slowness. But these traits are polygenic, difficult to measure, and subject to many cultural influences. Biological evolution will therefore be neither the only nor the main mechanism of adaptation.

What evolves fastest is culture: usage norms, digital education, information hygiene, collective regulations. Natural selection could eventually favor certain cognitive profiles, but it will not erase the need for immediate cultural and educational adaptation. The human brain will not be replaced: it will be reconfigured, partly by plasticity, partly by the collective choices we will make.