AI: When Machines Challenge Human Expertise

The Rise of AI as a Decision-Making Authority

The rise of artificial intelligences, such as ChatGPT and Gemini, has transformed their role from mere assistance tools to authoritative figures for many users. This evolution raises questions about reliability and accountability, as these systems are often perceived as experts, despite being simple aggregators of statistical data.

The Personification of Machines and the Authority Bias

Generative AI systems, initially designed as assistants, are now consulted for medical or legal advice. This blind trust in AI, seen as a single and reliable source, obscures the fact that it possesses neither consciousness nor accountability. The personification of machines creates an authority bias, where the user views AI as a competent interlocutor, to the detriment of human experts.

The meteoric success of conversational interfaces relies on a perception asymmetry. The modern user no longer sees a database but rather an interlocutor. This personification of the machine creates an unprecedented authority bias. Where a traditional search engine would provide a multitude of sources to filter, artificial intelligence offers a unique, structured, and assertive response. This form of response eliminates doubt and the effort of verification, instilling the idea that the machine possesses an intrinsic expertise superior to that of the craftsman, lawyer, or doctor.

This trend is particularly evident in technical sectors. An individual unsure about their electrical installation will often prefer the immediate response of an algorithm over consulting a qualified electrician. The sense of security provided by a well-formulated answer outweighs the awareness of real danger. A new phenomenon is also observed where AI is used as a mediator to validate or contest a human professional's diagnosis, thus reversing the legitimate hierarchy of knowledge acquired through field experience.

The Dangers of a Single Response

The shift from a multiple search model to a single response overly simplifies complex questions. In fields like law, where rules depend on specific contexts, this simplification can be misleading. The absence of legal accountability for AI means that users who follow its advice without verification bear the consequences of potential errors alone.

Ecological Consequences and Knowledge Depletion

The intensive use of AI for complex queries requires considerable computing power, with significant environmental impact. Moreover, by favoring AI responses over human experts, we risk depleting the quality of future data, creating a vicious cycle of knowledge degradation.

The commonplace use of artificial intelligence systems for trivial or ultra-complex queries conceals a material reality often overlooked. Each interaction requires phenomenal computing power, far beyond simple web indexing. The training and operation of these language models rely on data centers whose electricity and water consumption for cooling is colossal. Seeking virtual expertise for a subject that could be addressed through traditional research or direct human consultation contributes to massive environmental pressure, often disproportionate to the added value of the response obtained.

It is imperative to understand that artificial intelligence feeds on pre-existing data without ever creating new knowledge. It recycles, compresses, and redistributes. By considering it the ultimate answer, we risk depleting the sources of original knowledge produced by humans. If we stop consulting experts in favor of AI syntheses, the quality of future data that the machine will rely on will mechanically diminish, creating a vicious cycle of global knowledge degradation.

Redefining the Role of AI

It is crucial to redefine the uses of AI to preserve human expertise. AI should remain an assistance tool, excelling in reformulating and structuring ideas, without supplanting the analysis and intuition of professionals. Decision-making expertise must remain the domain of humans, who integrate elements that machines cannot comprehend.