Presented by Christof Koch, Ph.D, Chief Scientist, MindScope Program, Allen Institute, Seattle, Washington US.
The best understood cortex is primary visual cortex of the mouse. This is witnessed by state-of-the-art simulation of ~1 mm3 of V1 from the adult mouse, using two different levels of granularity (point neurons, spatially extended HH models). These models, based on the massive Allen Institute databanks functional connectivity and in vivo recordings, replicate in vivo spiking data the model was not trained on in a quantitative manner (Billeh et al., Neuron 2020). It is being extended to include detailed connectivity of electron-microscopically reconstructed data from mouse V1. It is likely that within a few decades such models could be extended to faithfully simulate the brain and the behavior of mice, predicting genuine new phenomena and system-level properties.
The human brain is three orders of magnitude bigger than the mouse brain. There is currently little evidence to suggest that it is, per unit volume, significantly more complex than the mouse brain. The field is now in the first stages of assembling a dataset of individual human pyramidal neurons and interneurons, based on in vitro data from neurosurgical samples. This provides a first, but limited view, onto the human brain circuits at the cellular level. For the foreseeable future, we will not have access to in vivo cellular data nor synaptic learning rules. This will impose unique limits onto our ability to faithfully simulate the human brain at the micro-functional level over the thirty or more years.