The Great Bifurcation
When we began the Biologically Inspired Model (BIM) project, our goal was singular: to build an artificial brain that learns the way a mammalian brain does. We wanted to move away from the static, frozen oracles of the Transformer era and build a continuous-learning substrate.
As we progressed through BIM 0 to BIM 4, we succeeded beyond our initial expectations. BIM 4, our "Digital Species Prototype," possessed a working chemical Brainstem, an Actor-Critic Basal Ganglia, and Hippocampal Sleep Consolidation.
However, we realized that the architecture now possessed two very different, but equally profound capabilities:
- A highly efficient sequence learning engine. The sparse cortical representations and local predictive mechanisms could mitigate catastrophic forgetting in natural language tasks, running on standard CPU architectures with low power requirements.
- A true digital organism framework. The chemical drives and sensorimotor loops allowed for a subjective, autonomous entity that learns from its environment to survive.
Honoring Both Destinies
A commercial language model does not need to sleep, and it does not need Dopamine. Conversely, a synthetic lifeform does not exist merely to autocomplete text.
To honor both paths, we have officially initiated The Great Bifurcation.
The original "BIM" moniker now serves as the umbrella for our ancestral research. Active development has split into two distinct tracks:
- BILM (Biologically Inspired Language Model): Focused on sequence memory and applied NLP. It strips away action-selection neuromodulation to optimize for processing text and sequential data.
- BIO (Biologically Inspired Organism): Our "Long Game". The undiluted realization of the UnikAI mission to build a continuous-learning artificial lifeform.
We are excited to pursue both tracks. You can read the technical specifications for both architectures on their respective pages.