To somewhat state the obvious - the problem isn’t the amount of data, it’s the algorithms.

We need to discover the set of learning algorithms nature has, and determine whether they’re implementable in silicon