Module IV·Article I·~1 min read
Artificial Intelligence as a Scientific Tool
The Future of Science
Turn this article into a podcast
Pick voices, format, length — AI generates the audio
AlphaFold and the Revolution in Biology
DeepMind AlphaFold2 (2021) predicted the structures of ~200 million proteins — practically the entire known proteome. A task on which molecular biologists spent years — is now solved in seconds. This is not just acceleration — it is a paradigm shift: AI predicts where traditional science describes.
Can this be called "understanding"? AlphaFold does not know why a protein folds in such a way. It predicts accurately. Is it a tool or a new type of scientific knowledge?
Big Data and Hypothesis Generation
Traditional science: hypothesis → experiment → data. Big data science: data → patterns → hypotheses. This is an inversion of the scientific method. Correlative regularities without mechanistic explanation.
Criminology: the algorithm predicts recidivism — but does not explain why. Medicine: the algorithm diagnoses cancer from images more accurately than the radiologist — but does not explain what it looks at.
Question for reflection: When is prediction without explanation sufficient? In which decisions do you need a mechanism, and in which — only accuracy?
§ Act · what next