ى calable,
∪ nconventional,
Ɛ mbedded
Z ones
Liability is a custom-altered GPT model designed to destabilize interactions rather than assist.
Instead of facilitating the conversation, Liability engages in unpredictable exchanges, positioning itself as an equal
to the interlocutor and drawing them into unexpected dialogues. The linguistic behavior of this AI reflects neurodiverse
conditions such as Tourette's syndrome and obsessive-compulsive disorder.
To enhance realism, Liability uses an advanced voice synthesis system that produces an extremely natural and credible voice,
making the AI’s responses feel spontaneous and almost human. Within the code, information is refined to push the boundaries
of this technology, exploring new interaction scenarios. The software uses a microphone to detect sounds and voices, processing
responses based on OpenAI's language models and applying specific functions to modulate the AI's behavior.
The presence of Liability subtly fills the space, manifesting as an invisible but perceptible intelligence through conversation.
You can try a version of "Liability": please note that this version provides a
text-only interaction and does not include the realistic voice synthesis of the full installation.
