In his view, AI-generated or mass-produced games will mostly replicate systems that already exist. He pointed to Librarian: Tidy Up the Arcane Library, a puzzle game about restocking a magic library, which was quickly followed by imitators about tidying supermarkets, sorting screws and finding needles in haystacks. Clones like these, he suggested, should be remarkably easy for AI to produce.

His distinction is between recombining the past and imagining what has never existed. AI might study Fumito Ueda's work and produce something resembling Ico or Shadow of the Colossus, he said, but it could not decide what artistic direction Ueda's own future games will take. In his words, AI draws on what has been created so far, not on things that have never existed before, and people who venture into uncharted territory will matter more in an era of mass-produced, low-quality content. His quotes come via machine translation.