Early research indicates that BLUM can identify which parts of language models can be simplified or eliminated without sacrificing performance, promising substantial cost savings and efficiency gains.

ALLT.AI plans to form strategic partnerships with major AI companies to incorporate BLUM’s neuroscience insights into next-generation language models, with the technology set for patent release in late 2025.

ALLT.AI, a joint venture between NXTLLM and MKHSTRY Labs, has developed BLUM (Brain-LLM Unified Model), a groundbreaking technology that translates neural damage patterns into AI models to significantly reduce computational requirements.

This innovative approach not only advances AI efficiency but also enhances understanding and treatment of neurological language disorders, positioning it as a transformative breakthrough in both AI and medical research.

BLUM's foundation lies in studying how stroke-damaged brains recover language, revealing essential neural pathways and mimicking these in AI systems to improve performance and biological relevance.

Led by neurobiology expert Dr. Julius Fridriksson, ALLT.AI emphasizes that the human brain remains the ultimate language model—faster, more efficient, and more intuitive—and that understanding brain damage and recovery can inform AI development.

The technology aims to create more efficient, biologically inspired AI models that could drastically cut energy consumption compared to current large language models, which are known for their high electricity use.

BLUM employs a translation algorithm that maps impaired language outputs from 'lesioned' models to neuroimaging data of brain lesions, enabling the simulation of neurological conditions like aphasia in AI systems.

Beyond language, the technology has potential applications in modeling neurological diseases such as Alzheimer’s and dementia, offering new insights into their treatment and understanding.