Square Enix has launched a reboot of arguably the most influential AI game yet, but it’s an incredible failure of both a classic revival and an AI experiment.
Last week, the publisher announced the forthcoming release of Square Enix AI Tech Preview: The Portopia Serial Murder Case. The original Portopia, a murder-solving text adventure game, is one of the most influential titles in the Japanese gaming canon, but despite this, it has never been officially released in English. The idea of getting an official translation along with the innovative use of an AI language model to replace the old text parser seemed potentially exciting.
That excitement evaporated almost as soon as I started typing in the new Portopia’s text field. As in the original game, you don’t control your character directly, instead you give orders to your assistant, who takes them on your behalf. So you can say “question” to have him ask questions to local passers-by, or say “tell me about the victim” to get some details about the character.
The problem is that AI is hardly capable of answering all but the most narrow questions and commands. If you’ve ever played an old school text adventure and had to struggle to figure out what vocabulary the computer actually understands, you’re starting to understand the frustration here. In the new Portopia, things are about a thousand times worse, because instead of learning a few specific commands that the game recognizes, you have to work around the whims of the algorithm, trying to understand your words.
If you type something that the AI doesn’t recognize, which is about 95% of what you type, you’ll get one of several out-of-the-box responses from the game, like “I’m not sure what to say about this”, “hmm” or “maybe we should focus on the current task.” Not exactly the AI-driven future we’ve been told.
Portopia currently has just under 300 user reviews on Steam, of which less than 20% are positive. This is the worst percentage. any game that Square Enix has ever launched on Steam (will open in a new tab), below even such infamous disasters as Balan’s World of Wonders, The Fall of Babylon, The Survivors, and The Quiet Man. You can read these reviews yourself on the game’s website. Steam page (will open in a new tab) if you want more examples of how this technology went wrong.
The strangest thing about this new version of Portopia is that there is no AI-generated dialogue in response to your input. Writing AI is controversial – and for good reason – but it seems like if you’re going to release an “AI Technology Preview” you’ll want to go through with it. Having an AI companion that will give you hints when you stumble over the next story trigger might even be the detail that makes the whole game fall into place.
“This tech preview originally included a feature based on natural language generation technology, where the system would generate natural responses to questions that didn’t have a pre-written answer,” Square Enix said in a post on the game’s Steam page. “However, the NLG feature is omitted in this release because the risk remains that the AI will generate unethical responses. We thank you for your understanding. We will consider re-introducing this feature as soon as our research allows us to create an environment where players can enjoy the game with peace of mind.”
Yes, the fear of “unethical answers” is what keeps it all from making sense. Most AI chat tools use some sort of filter to keep responses from getting too horny or offensive – ChatGPT was very insistent on protecting the honor of the Pokémon icon, Ash’s mom – but I keep wondering why this game even exists if it will exist in such a compromised state. This is a free preview meant to showcase the future of AI technology in games, but if that’s what we can expect, then you can count me out.
Hideo Kojima loved Portopia so much that part of the game’s code is hidden in MGS 5.
