AI imperialism and the prospects of data-centred and automated natural science
Descargar
- Descargar este vídeo (143.32MB)
AI is imperialistic because it encroaches upon any human endeavour, including scientific research. AI’s epistemic imperialism is fast-paced, encompassing and ground breaking, as testified by the recent encroachment on mathematics. In the natural sciences, it is founded on an old-fashioned empiricist gamble dressed up in new clothes: using current databases to extract inferences through automated processing, dispensing of obscure causes and mechanistic insight. The irony is that the outcomes of AI-based automation are often obscure, compromising human intelligibility. A chief example of AI-based natural science is AlphaFold. It is often suggested that AlphaFold has solved the protein folding problem, one of the major theoretical problems in the life sciences. I shall first illustrate the nature of AlphaFold’s contribution to biochemistry. I shall then highlight some of its limitations and some of the problems that it engenders. On this basis, I shall turn to the more general issue concerning the prospects of data-centred automation, particularly in other life sciences. I end up by raising some general questions: is the hype concerning AI-based science justified? Is there a difference between encroaching on mathematics or on the natural sciences? What is the scope of AI-based natural science? And is the new empiricist gamble justified?
-
Davide Vecchi University of Bologna