How does a neural network-driven ghost work?
We've been receiving messages from players who are intrigued that "Bureau of Contacts" will feature the possibility of being killed by a neural network for the first time, but also questioning - how is this possible?
Indeed, this is the first time that generative technology has been used in a game in such a way as to control behavior based on players' words and actions.
Think back to what good old traditional AI was like - a large set of scripts and a bit of randomization magic yielded a well-designed set of conditions and possible outputs.
That's how it used to be, already 15 years ago...
In our case, we use not that classical, but a generative model that can realize not only content, but also meaning. In other words, the ghosts in the game:
- analyze context;
- simulate moods and attitudes towards what they hear/occur;
- make decisions on their own will and reasoning, parodying human behavior and "cause and effect relationships".
All of this is realized to the extent of how well the model is trained and how powerful the neural network is.
In other words, in "Bureau of Contacts" you will have to watch your words, because you will not only get a meaningful response for these words, but may also be subjected to appropriate "actions" if IT think you deserve it.
Let's look at one possible reaction to the provocation:
Bureau of Contacts
"Bureau of Contacts" - first horror game in which neural network is trained to kill
Status | In development |
Author | MiroWin |
Genre | Action |
Tags | Co-op, Exploration, Ghosts, Horror, investigation, Multiplayer, Singleplayer, supernatural, Survival Horror |
Languages | English |
Leave a comment
Log in with itch.io to leave a comment.