lArtificial intelligence (AI) awakens a sense of wonder comparable to that of the industrial revolution of the 19th century.my century. But calls for caution are increasing, as evidenced by bills aimed at regulating the development and use of AI in California or the French government’s recent initiative to measure its environmental impact. These concerns illustrate fundamental questions about our relationship with technology, our pride, and our desire to believe in unlimited progress.
The ability of humans to transform their environment through technology and language (through the stories that structure our view of the world) defines their condition. This freedom is manifested in the constant desire to push the limits of what is possible, as illustrated by research in artificial intelligence and the emergence of applications such as ChatGPT (and its recent incremental improvement o1), Gemini or Llama.
These events, as charming as they are disturbing, are sometimes more fantasized than questioned. Because if the basic techniques of deep learning – a branch of AI in which machines learn from data – have been known since the 1950s, the recent appearance of specific applications such as large language models (LLM) in English) large language models), formed in massive amounts of texts since 2018, surprised the general public. But not AI researchers, who already knew its potential.
Public enthusiasm
The surprise comes from the public’s enthusiasm for these systems. Following the failure of Microsoft’s Tay chatbot a few years ago and criticism of Galactica, Meta’s scientific writing tool, ChatGPT surprised by experiencing dazzling success since its launch. This enthusiasm has no doubt been fueled by its ease of use, its ability to provide consistent answers, and effective marketing focused on potential for use.
For Yann Le Cun, head of artificial intelligence at Meta and father of deep learning [l’apprentissage profond] – a subfield of deep learning that uses artificial neural networks to analyze and interpret complex data – interviewed on the “Do It Yourself” podcast, this mastery of language by LLMs may seem impressive, but it is actually based on simple statistical models and does not reflect true intelligence.
You have 55.12% of this article left to read. The rest is reserved for subscribers.
