Square Enix recently developed an experimental remake of the 1983 Enix mystery-adventure game, The Portopia Serial Murder Case, reimagined as a technical showcase for AI technology. The original game was released during a boom period for classic text-based adventure games, and allowed players to enter specific word prompts to find clues. The 2023 remake aimed to expand this edict by allowing AI to interpret player commands in creative ways – but, as it turns out, the AI just doesn’t work very well at all.
Players jumping into the game have reportedly discovered the AI is extremely limited, to the point where it can barely interpret even basic commands. While it uses a hearty Natural Language Processing (NLP) system to attempt to translate player inputs, users report that it can’t even tell the difference between phrases like ‘go to study’ and ‘go to the study’.
One player simply asked a character, ‘Who are you?’ for basic context, and received the generic pushback answer of ‘Maybe we should focus on the task at hand.’ Interrogating the same character about a murder victim’s enemies and friends achieved similar results, and repetition of unhelpful dialogue.
Read: Ubisoft faces major backlash after announcing AI writing tool
‘This is basically a text parser that looks for certain keywords such as names, or specific words such as “Investigate” or “Go”. Any variant that even strays remotely from the known list of keywords will return a nonsensical answer,’ the user reported via Steam.
Elsewhere in the game’s reviews, a similar sentiment is shared. Players claim they can’t get anything tangible out of the game’s AI, and many agree that its dialogue system is far more limited than even classic text adventures, which were usually hand-written and programmed by meticulous human beings.
Yet despite this negativity, it appears Square Enix is confident in this game’s release, and how AI might aid games in future.
‘At the time of the game’s original release, most adventure games were played using a “command input” system, where the player was asked to type in text to decide the actions of their character,’ Square Enix said in the game’s description. ‘Free text input systems like these allowed players to feel a great deal of freedom. However, they did come with one common source of frustration: players knowing what action they wanted to perform but being unable to do so because they could not find the right wording.’
’40 years have passed since then, and PC performance has drastically improved, as have the capabilities of NLP technology.’
According to the game’s ‘Very Negative’ Steam rating, and a cavalcade of reviews flagging the game as barely operable, that ‘drastic improvement’ has not quite been realised in this experiment. The Portopia Serial Murder Case was an early success for Enix, and did much for the adventure game genre – but this remake appears to be a poor reflection of what made the game great.
As much as the current conversation around AI focusses on its future potential, this is one case that highlights just how far it must evolve before it becomes viable as a tool for robust storytelling.