Koch and Buice make the generous point that all this might not matter if one could devise the biological equivalent of Alan Turing's Imitation game, seeing if an observer could tell whether output they observe for a given input is being generated by the simulation or by electrical recording from living tissue. Here are some interesting clips from their article in Cell.
...the current BBP model stops with the continuous and deterministic Hodgkin-Huxley currents...And therein lies an important lesson. If the real and the synthetic can’t be distinguished at the level of firing rate activity (even though it is uncontroversial that spiking is caused by the concerted action of tens of thousands of ionic channel proteins), the molecular level of granularity would appear to be irrelevant to explain electrical activity. Teasing out which mechanisms contribute to any specific phenomena is essential to what is meant by understanding.
Markram et al. claim that their results point to the minimal datasets required to model cortex. However, we are not aware of any rigorous argument in the present triptych of manuscripts, specifying the relevant level of granularity. For instance, are active dendrites, such as those of the tall, layer 5 pyramidal cells, essential? Could they be removed without any noticeable effect? Why not replace the continuous, macroscopic, and deterministic HH equations with stochastic Markov models of thousands of tiny channel conductances? Indeed, why not consider quantum mechanical levels of descriptions? Presumably, the latter two avenues have not been chosen because of their computational burden and the intuition that they are unlikely to be relevant. The Imitation Game offers a principled way of addressing these important questions: only add a mechanism if its impact on a specific set of measurables can be assessed by a trained observer.
Consider the problem of numerical weather prediction and climate modeling, tasks whose physico-chemical and computational complexity is comparable to whole-brain modeling. Planet-wide simulations that cover timescales from hours to decades require a deep understanding of how physical systems interact across multiple scales and careful choices about the scale at which different phenomena are modeled. This has led to an impressive increase in predictive power since 1950, when the first such computer calculations were performed. Of course, a key difference between weather prediction and whole-brain simulation is that the former has a very specific and quantifiable scientific question (to wit: “is it going to rain tomorrow?”). The BBP has created an impressive initial scaffold that will facilitate asking these kinds of questions for brains.