Thierry Rayna, Professor at the Polytechnic: "The industrial age's model of innovation is a heresy in history"
The article discusses the current state of innovation in the AI industry, contrasting the closed, proprietary models developed by companies like Anthropic and OpenAI with the open-source approach favored by Chinese firms. It quotes Thierry Rayna, a professor at École Polytechnique and researcher at CNRS i³, who argues that the industrial-era model of innovation, characterized by centralized, secretive research and development, is outdated and counterproductive in the digital age. Rayna emphasizes that historical innovation was more collaborative and open, with discoveries freely shared and improved upon by others. He criticizes the dominance of the industrial-era model, which he claims has led to inefficiencies and a lack of visibility for those who do not control proprietary systems. The piece highlights the growing shift toward open innovation, particularly in the realm of artificial intelligence, where open-weight models are gaining traction.
Thierry Rayna, professor at Polytechnique, stated, “The model of innovation from the industrial era is heresy in history.” He argued that the current digital age has fundamentally changed how innovation occurs, moving away from the centralized, closed systems of the past. Anthropic and OpenAI are both preparing for major milestones: Anthropic plans to list on the stock market in November, while OpenAI will do so the following year. During this period, Anthropic reported two profitable quarters, while OpenAI launched new personal agents called “dots” on September 29. To demonstrate their commitment, both companies have been vocal about cybersecurity risks associated with AI and signed the self-regulation charter proposed by Donald Trump. However, the real threat comes from China. While Western firms develop proprietary models, Chinese companies bet heavily on open innovation through their “open weights” models. According to Rayna, these models have fully understood the importance of openness in an environment saturated with content and choices. He explained, “In a world of abundance of content and choice, refusing to be copied is condemning oneself to invisibility and insignificance.” Rayna emphasized that historical innovation was always artisanal and open. From the wheel to bread, discoveries spread freely and were often improved upon by others, even if some countries sought to keep certain processes secret. It was during the industrial era that innovation became concentrated within large enterprises, requiring expensive machinery and extensive research teams. This model dominated from the 1960s until the 1990s, yet it does not represent the true nature of human innovation throughout history. He noted that open-source software initially faced resistance from proprietary software advocates. In reality, the proprietary model is the anomaly. Digital technologies have revived open innovation, transforming economic models. Despite lingering in the industrial era, the shift toward open systems is accelerating. The core feature of the industrial era was the centralization of production means within corporations. Digital tools, however, empower individuals to communicate, create, and innovate freely. This shift explains the rise of social media, Wikipedia, and open-source software. As tech companies once resistant to open innovation begin to embrace it, Rayna pointed out that even the largest firms struggle against communities of developers and users who better understand user needs.
The article discusses the current state of innovation in the AI industry, contrasting the closed, proprietary models developed by companies like Anthropic and OpenAI with the open-source approach favored by Chinese firms. It quotes Thierry Rayna, a professor at École Polytechnique and researcher at CNRS i³, who argues that the industrial-era model of innovation, characterized by centralized, secretive research and development, is outdated and counterproductive in the digital age. Rayna emphasizes that historical innovation was more collaborative and open, with discoveries freely shared and improved upon by others. He criticizes the dominance of the industrial-era model, which he claims has led to inefficiencies and a lack of visibility for those who do not control proprietary systems. The piece highlights the growing shift toward open innovation, particularly in the realm of artificial intelligence, where open-weight models are gaining traction.
Bias read (Progressive): The article frames the industrial-era model of innovation as inherently flawed and restrictive, aligning with progressive critiques of centralized corporate control over technology. While it does not explicitly reference political parties or policies, the critique of proprietary systems and advocacy
The article discusses the limitations of artificial intelligence compared to human imagination, using the history of aviation as an example. It highlights how early aviation pioneers relied on creativity and intuition, which AI currently lacks. The piece emphasizes that while AI can process data efficiently, it struggles with tasks requiring original thought and vision. The comparison suggests that human ingenuity remains unmatched in driving innovation, particularly in complex fields like aerospace. The article does not present specific technological advancements or case studies but focuses on conceptual differences between AI capabilities and human cognitive abilities.
Bias read (Center): The article presents a balanced discussion of AI's current limitations relative to human imagination without overtly favoring any particular ideological stance. It uses historical examples to illustrate a general principle rather than promoting a political agenda. The framing remains objective, with
How each side covered it
The same event, grouped by the political lean of the outlets covering it.
progressive
center
conservative
★
How each side covered it
Support independent, bias-aware news and unlock the social pulse, community voting, and every other Supporter feature.