Our Aims
We are a group of neuroscientists, engineers, and AI researchers working to reconceptualize neuroscience and to safeguard human agency in an increasingly AI-shaped world. We believe that:
- Basic science and preclinical research should move faster: Automation and open-source tools already available today can accelerate breakthroughs blocked by manual workflows and siloed data in traditional academic and industry approaches.
- Creating foundation neural models of biological organisms will make for better and less harmful science. Better sensors and smarter analysis allow us to learn more from fewer animals in naturalized habitats. We will generate peta-byte and exa-byte neural-behaviour datasets and build models that capture dynamics sufficiently to do science in the cloud.
- Frontier science can be more open: We're creating the
Neurotech Academy and a
Whole-brain-emulation course of study to onboard a community of like-minded scientists.
- Human agency is precious. We're concerned about AI tools and systems replacing human control over our future, and have written extensively on this:
functional WBE,
human agency loss,
non-reversal agency loss to AIs,
and ran
AI safety hackathons on agency loss,
Transparent, biologically inspired models can help us study goal formation and alignment in AI systems. Whole-brain emulation (WBE) is an ideal tool for this work.
Our present work has two intertwined and ambitiuos aims:
- Create advanced neuroscience research platforms focusing on naturalistic behaviours.
We want to help solve the
translation and
replication crises in biology. We believe that overcoming these requires
studying animals in increasingly natural and confound-free paradigms. So we build naturalized behaviour rodent habitats that generate petabyte-scale multimodal datasets.
At the core of this approach is the replacement of single-task, low-bandwidth behaviour assays with high-resolution, long-term behavioural recording.
- Create foundation models of the brain and whole-brain emulation technologies while evaluating them as a principled platform for AI safety.
We are working on building the first foundation models of the C57BL/6 mouse strain. Our approach relies on scalable hardware and data generation approaches
and AI acceleration. We seek to analyze and decompose models using neuroscience and AI interpretability tools –
towards a biologically grounded baseline for evaluating alignment proposals. Our work aligns with the NeuroAI-for-Safety
roadmap and current foundation and whole-brain-emulation model advocacy efforts.
Looking Ahead
We seek to start a revolution in neuroscience: building a functioning mouse FM capable of novel task learning while emulating mammalian brain dynamics that can be causally perturbed. This first model can lay a new foundation for basic and preclinical research vastly reducing the need for biological organisms in basic and preclinical research. We view achieving human FMs and WBEs as the next goal - and a critical component of AI alignment and human fluorishing strategies.
Join Us
- If your work is in behavioural or computational neuroscience, systems engineering, ASIC design or ethics — we want to hear from you.
- We welcome collaborators committed to rigour, open science, and long-term thinking.
- We're actively seeking contributors to our Ethics Board: If you've spoken publicly about AI safety,
machine consciousness or human agency — please reach out.
- Send us your introduction email along with your CV to: info@netholabs.com