قالب وردپرس درنا توس
Home / LifeStyle / It gets weird when a neural network is trained on text-adventure games

It gets weird when a neural network is trained on text-adventure games



The answer is pretty strange. Shane aptly describes the experience as "dreamy," with the scenery often and apparently without reason switching from scene to scene. For example, in one run, the AI ​​opened the game with a scene that was only set in space to quickly turn things into a "labyrinth of small, twisty passages that are all the same". If there is a passage to many of the scenarios, it is Zork one of the games that trained the neural network and a classic of the genre. The AI ​​often challenges the 40-year-old game to respond to the player, and often presents trolls as obstacles to progress. See below:

The Troll emerges from under the bridge and blocks your way.
You are on the south side of the abyss.
The infinite wonder that is Urbzig is honored.
A solid rainbow spans the abyss.

Another peculiarity of the game is that it does not run like a traditional narrative: there is no beginning, no middle and no end for each scenario. Instead, every session is an endless marathon with more trolls than bridges. And yet, it is noteworthy in every scene how it sets the tone of classic PC adventure games. There's something about how the AI ​​jumps from scenario to scenario capturing the atmosphere of these games. If you want to start your own AI-inspired fever dream, visit the Google Colab document created by Nathan. Just be prepared for a DOS-like experience. That's part of the appeal.


Source link