Conférence de Matthew Battles (Harvard University)

Value of the Information & Strength of the Argument

The lecture offers a valuable and original perspective by bridging the fields of natural history, cultural history, and computational technology. Battles effectively argues that the practices of taxonomy are deeply embedded in human culture and craft, using concrete examples like Nabokov’s scientific work and the demonstration of a neural network to illuminate how meaning is constructed through naming. His argumentation is solid and nuanced, particularly in his refusal to fully endorse the ideology of machine learning as a natural phenomenon, instead viewing it as a cultural artifact. The use of literary and historical references, from Jules Verne to W.S. Merwin, enriches the discussion and provides a poetic layer that reinforces the central themes. However, the lecture is more reflective than rigorously analytical, with some points relying on poetic associations rather than empirical data, which slightly weakens the overall argumentative force.

Scientific Rigor, Source Quality, Title Accuracy

Battles demonstrates strong intellectual rigor by referencing a range of sources, including William James, David Weinberger, and André Karpathy, and he explicitly notes that he will share a list of resources after the talk. The quality of sources is high, drawing from established academic and public intellectual work. There is no evident fabrication or misrepresentation of facts; the historical and scientific claims are generally accurate. The title ‘Earth Measurer’ directly ties to the theme of taxonomy and to the winter moth’s family name Geometridae, meaning ’earth measurer,’ and the subtitle accurately previews the discussion of biodiversity, affect (emotional resonances), and technology amid mass extinction. The title is well-adapted to the content, and the presentation successfully delivers on its promise. The talk is not a formal scientific study but rather an expert opinion, which is clearly communicated.

The title 'Earth Measurer: Biodiversity, Affect, and Technology in the Midst of Mass Extinction' accurately reflects the content, as the talk explores the intersection of natural history collections, machine learning, and the cultural dimensions of biodiversity loss.

8/10

The speaker is a recognized scholar with a strong publication record in the cultural history of libraries and writing. The lecture is well-argued, referencing a range of historical and contemporary works, and demonstrates a thoughtful integration of theory and practice. However, it is an opinion-based talk rather than a peer-reviewed study, and some interpretations are subjective, which slightly reduces the reliability score.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 02:00Introduction of the talk topic and the William James quotation about nature as excess.
• 10:00Discussion of the Catalogue of Life and the taxonomic framework for biodiversity, including the dark abundance of undiscovered species.
• 18:00Introduction of Vladimir Nabokov as an entomologist and his discovery of the Karner blue butterfly, illustrating craft practices of naming.
• 42:00Transition to machine learning, referencing David Weinberger and the concept of post-scarcity computing.
• 50:00Demonstration of a recurrent neural network trained on moth genus names, generating plausible but invented names.
• 1:00:00Conclusion with W.S. Merwin's poem 'Learning a Dead Language' and reflections on the fate of names and species.

• Catalogue of Life — Mentioned as a meta-database of biological taxonomy, used to explore the diversity of insect orders.
• André Karpathy's char-rnn GitHub repository — Source of the Python script used in the demonstration of a recurrent neural network generating moth genus names.