A heated debate on X (formerly Twitter) between tech giants Elon Musk and Yann LeCun has reignited questions about what constitutes science. The clash began on May 27 when Musk invited followers to join his company xAI, emphasizing a "maximally rigorous pursuit of the truth, without regard to popularity or political correctness." xAI aims to build advanced AI capable of reasoning, with its first product named Grok.
Yann LeCun, Meta's chief scientist and a pioneer in deep learning, criticized Musk, pointing out the contradiction between Musk's claim of rigor and his promotion of conspiracy theories on the platform. The argument quickly escalated, with Musk questioning LeCun's recent scientific contributions. LeCun responded by citing over 80 technical papers published since January 2022 and argued that unpublished research is not science.
The Core of the Debate
LeCun's assertion that research must be published to be considered science sparked a widespread backlash. Critics argued that science is fundamentally a method rather than merely a collection of published papers. Palmer Luckey, the founder of Oculus, condemned the notion that unpublished research would render scientists forgotten and bitter. Many noted that significant scientific work in the private sector often remains unpublished, and even in academia, a large portion of data goes unpublished.
Peter Coveney, a computer scientist at University College London, emphasized that LeCun's view gatekeeps the broader understanding of science as a method applicable in daily life. He and other critics argue that feedback and openness are crucial components of scientific progress, not just publication.
Clarifying Science
LeCun later refined his stance, highlighting that science advances through idea exchange, verification, and reproduction. He stressed that without sharing research in some form, its impact would be minimal. He also advocated for greater transparency in AI research, especially regarding the publication of source code.
Janet Stemwedel, a philosopher of science, and Coveney agree that openness is vital, particularly for AI algorithms like ChatGPT, Sora, and AlphaFold3, which have been criticized for lacking transparency. Stemwedel pointed out that responsiveness to feedback is a cornerstone of modern science, along with its predictive and explanatory capabilities.
Coveney highlighted the development of generalist AI tools, which interpret data and produce advanced reasoning. He questioned the scientific rigor of these methods, even when scrutinized by scientists. He noted that Musk's vision of using explainable AI to replace conventional scientific methods is flawed, as AI's statistical inferences do not equate to true understanding.
The Ongoing Debate
Stemwedel emphasized that defining science will always be contentious. She noted that pre-Musk Twitter fostered productive scientific discussions that showcased science's responsiveness to feedback. However, she expressed concern that X under Musk has become less conducive to reasoned debate.
Coveney underscored the importance of maintaining the fundamental principles of science, originating from the Enlightenment, which prioritize objective discussion over mere opinion. He noted the irony that the debate on X itself exemplified the very issue of conflating opinion with science.
As the conversation continues, it underscores the need for clarity and openness in defining and practicing science, especially in an era where AI plays an increasingly significant role.
