Borretti argues ASI capabilities can’t be deduced

The author argues that from the definition of artificial superintelligence (ASI), we can deduce very little about its actual capabilities and behavior. Without complete deductive theories in areas like persuasion or scientific discovery, predicting what an ASI can do remains an empirical question. Furthermore, it is unclear whether mathematical theories of rationality will apply or if they represent an unreachable standard. Open article →