TFL Vsebine / Zbornik znanstvenih razprav

Po kratki uvodni predstavitvi orodij umetne inteligence avtor s tremi pomembnimi raziskovalnimi dosežki, ki jih teoretiki v celoti pripisujejo umetni inteligenci, preuči pravilnost takega poročanja v luči izvirnih raziskovalnih poročil. Nato predstavi prak- so Evropskega patentnega urada, Japonskega patentnega urada, Državnega urada LR Kitajske za intelektualno lastnino ter Urada ZDA za patente in blagovne znamke, ki načeloma podeljujejo patente za izume, povezane z umetno inteligenco. Sledi kritičen premislek o akademski razpravi o zmožnosti uveljavljenih paradigem patentnega prava, da obvladajo izzive tehnologije umetne inteligence. Dejstva široko objavljenih patentnih prijav »DABUS« kažejo na slabosti trditve, da je umetna inteligenca avtonomno iznašla prijavljene izume, ter opozarjajo na pomanjkljivosti sistema in kažejo na ukrepe za us- trezno določitev na primer stanja tehnike, in oceno zahtevane stopnje inventivnosti za izume, povezane z umetno inteligenco. V skladu s prakso vodilnih svetovnih patentnih uradov, razumevanjem in presojo te prakse s strani njenih glavnih uporabnikov avtor meni, da bodo pod vplivom tehnologije umetne inteligence ustaljene patentnopravne paradigme nekoliko evolucijsko prilagojene, vendar ne bo prišlo do bistvenih sprememb. Avtor prispevek sklene s kritičnim komentarjem načrta EU za sprejem uredbe, ki bi posebej obravnavala vse kompleksne vidike patentiranja izumov, povezanih z umetno inteligenco.

After a brief introduction into the tools of artificial intelligence (AI), the author uses three important research achievements attributed by science writers entirely to the AI to examine the correctness of such reporting in the light of the original research reports. The author presents the practice of the EPO, JPO, SIPO and USPTO, which all in principle grant patents for AI-related inventions. A critical reflection on the academic debate ques- tioning the ability of the established patent law paradigms to master the challenges of the AI technology follows. The facts of the broadly publicised “DABUS” patent applications serve to demonstrate the weakness of the assertion that AI autonomously invented the claimed inventions and to point to the deficiencies of the system and suggest measures to adequately determine, e.g. the prior art, and assess the inventive step requirement for AI- related inventions. In line with the practice of the leading patent offices of the world, the understanding and assessment of that practice by its main users, the author believes that under the impact of the AI technology, the settled patent law paradigms will undergo certain evolutionary adaptations, but will not change as to their substance. Finally, the article critically comments on the EU plans to adopt a regulation, which would specifi- cally deal with all complex aspects of patenting AI related inventions.