The real problem is --- progress toward this "breakthrough" is not happening. Instead, the evidence shows that most of the easily available data has already been consumed and capabilities are reaching a plateau.

The probabilistic design is revealing itself as a fundamental constraint. Any significant breakthrough will likely require a completely different architecture.

It's unlikely that real intelligence is a probabilistic function connected to a huge database.