Jeff Kaplan has little patience for artificial intelligence evangelism. Five years after leaving Blizzard Entertainment, the legendary game designer sat down with podcaster Lex Fridman and delivered a blunt assessment of the technology's role in gaming: it's oversold, it's unreliable, and it will never replace the human spark that makes great games work.
Speaking to Fridman in a lengthy conversation about his career and upcoming projects, Kaplan was asked about the future of AI in game development. The examples offered were typical of tech industry pitch decks: AI-generated NPC dialogue that would allow richer storytelling, procedural systems to replace routine tasks, the kind of futuristic vision Silicon Valley uses to convince investors and executives that artificial intelligence is the next big thing in interactive entertainment.
Kaplan's response was dismissive. According to reporting by Kotaku, he described current attempts at integrating AI into game development as "mostly a hot mess," noting that the technology and its proponents are "overconfident" in what they can actually deliver.
He shared a telling example from his own experience. When faced with a user interface design challenge, Kaplan decided to test ChatGPT despite it not being his area of expertise. The AI returned an answer with complete confidence. It was wrong. The experience crystallised his view: the technology mistakes certainty for accuracy.
There's a legitimate debate about where AI can be useful in creative industries, and Kaplan doesn't dismiss it entirely. He sees potential value in automating routine logistical work within game development, the kind of tedious tasks that consume developer time without adding much creative substance. But he's sceptical of fantasies about entire studios being run by algorithms to cut costs.
The ethical dimension matters equally to him. Kaplan was direct on this point: using artists' and voice actors' work to train AI models without permission is "immoral." It's a principle worth lingering on, especially given how pervasive such practices have become across the tech sector. Training data has value. The people who create that data deserve a say in how it's used.
When pressed on whether AI could eventually match human creativity, Kaplan drew a clear line. "No matter how good AI gets, it's never gonna draw a picture like Arnold Tsang, it's never gonna tell a story like Chris Metzen," he said, referring to artists and writers who defined Overwatch and other Blizzard titles. "Human spirit is irreplaceable."
The comment matters coming from Kaplan specifically. He spent 19 years at Blizzard, helping to build two of gaming's most influential franchises. World of Warcraft and Overwatch succeeded not because of superior technology or clever algorithms, but because teams of humans made creative decisions that connected with millions of players. That required taste, empathy, risk-taking, and the kinds of judgements machines cannot replicate.
Kaplan's perspective gains additional weight because he's now running his own studio, Kintsugiyama, preparing to launch The Legend of California, a multiplayer action-survival game set during the gold rush era. According to Steam, the title combines exploration and crafting with historical storytelling on what the studio describes as a mythical island of California. It's a deliberate departure from the fantasy and sci-fi work he built his reputation on.
Kaplan's decision to build a smaller studio of 34 developers and pursue an original creative direction signals something important. He's chosen autonomy and creative control over scale. That choice reflects a philosophy fundamentally at odds with the silicon Valley perspective that bigger, faster, and more automated always wins. Instead, Kaplan is betting that thoughtful design and human judgment remain the real competitive advantages in gaming.
His recent comments on AI should be read in that context. This isn't a Luddite rejecting all technological change. It's a veteran designer saying that not everything can be outsourced to machines, and that chasing efficiency above all else is precisely what drained the joy from some of his work at Blizzard. The real lesson isn't about AI specifically. It's about recognizing what artificial systems can and cannot do, and resisting the temptation to pretend they're a solution to every problem.