Outlook
Future NeuroAI research
Where NeuroAI is likely to advance through 2030 — and where realism demands patience.
Key takeaways
- Expect rapid progress on efficiency, slower progress on continual learning.
- Neural foundation models — large models pretrained on neural recordings — are an emerging category.
- Closer alignment between systems neuroscience and AI evaluation is the field's main institutional need.
Three directions worth watching
Neural foundation models that learn from large-scale neural recordings; efficient training algorithms inspired by local synaptic rules; tightly co-designed neuromorphic hardware/software stacks.
Where to be patient
True biological-level continual learning, energy parity with the brain, and rigorous mechanistic interpretability of large models are all multi-decade challenges, not multi-year ones.
Sponsor this work
BRAINMATTER is independent and reader-supported. Everything here stays free, ad-light, and open to everyone. A one-time sponsorship covers research time, editorial review, and hosting so we can keep publishing without paywalls.