It’s really weird that this article walks past the 2 likely explanations and then declares them both insufficient in favor of an assertion that maybe it’s not any fun.

Speculative generation is expensive and time consuming and in most cases will just be the game company writing a check to a provider every time the player does something. It’s very difficult to imagine a revenue model which allows that to make sense. Even if you did get it to make sense, you then have to worry about their being a market risk to people associating your game with AI. I don’t think you need more than those two explanations to understand why in two or three years we have not seen games that the author describes come out.

Is it a possibility that talking to a chat bot is not any fun from a games per perspective? Yes it might even be very likely. But we’re not gonna have a real answer to that question tested out by real game developers until it becomes pragmatically possible to ship a game using these tools.

The cost of generation will come down and people will find clever uses for it and one way or another opinions about AI will change. Then maybe we’ll see whether or not these are any fun.