Not only this, but I think we miss the more important detail that AGIs like our brains are probably so generally successful because they are relatively sloppy.

Not only this, but I think we miss the more important detail that AGIs like our brains are probably so generally successful because they are relatively sloppy. So while they can reliably solve all sorts of problems for which they have little or no preparation, they are also remarkably unreliable for solving specific problems requiring specific answers. But these traits are likely two sides of the same coin. And any software solution that replicates the desirable traits is likely going to come with some of the undesirable ones as well.