TALUS™ (Trained Artificial Logic Used for Safety) is 2-Dooz’s foundational Mental Health Safety AI model. The name is inspired by the talus, which is a small bone in the ankle that is instrumental for balance and that holds the weight of an entire body. Likewise, TALUS AI is a core model that serves as the foundation for a family of multimodal AI models which target specific aspects of mental health.
The catalyst for the development of the TALUS AI foundational model is our work within the Resilience Tech-Forum on suicide prevention among military veterans. Within the RTF, datasets which are specific to suicidality are used to extend the TALUS AI model to estimate a veteran's suicide risk level.
A key differentiator for TALUS is that it is trained on high-quality, clinician-curated datasets in accordance with Clark’s Law, which postulates that for the same level of model performance, the higher the quality of the training dataset, the lower the requirement for training and inferencing capacity.
This allows for a model implementation that is highly accurate while requiring minimal processing and energy consumption for training and inferencing.