Artificial intelligence (AI) is being increasingly used to design and predict antibody candidates and to identify novel correlations and patterns in biological data. This raises a fundamental patent-law question: which AI-generated or AI-discovered results constitute a sufficient technical contribution to justify patent protection? This article distinguishes two analytically distinct cases: whether functionally defined genus claims covering AI-predicted antibodies are commensurate with the technical contribution actually disclosed, and whether AI-discovered correlations or patterns qualify as inventions at all. The former concerns claim scope, and the latter subject matter. The first case is examined through the enablement (Amgen, Wands) and written description (Ariad, Baxalta) requirements of §112(a), and compared with Korean case law under Article 42(3) and 42(4) of the Korean Patent Act. Korean law keeps the two requirements analytically separate, whereas the §112(f) means-plus-function approach in the U.S. has no counterpart in Korean law. The second case is analyzed through §101 and the Alice-Mayo framework, drawing implications for Article 2(1), which defines an invention as a creation of a technical idea utilizing a law of nature. The article argues that AI changes the factual premises to which established patent doctrines apply rather than the doctrines themselves. For functional claims, what matters is the ratio of verified to prophetic examples, the representativeness of the disclosed candidates, and whether the experimentation left to a skilled artisan is confirmation or new search. For claims on correlations or patterns, whether the claim stops at understanding the relationship or applies it through concrete technical means. Both questions return to two conventional boundaries of patent law: correspondence between disclosed technical contribution and patent scope, and the distinction between discovery and invention.