It’s strikes me that AGI is no longer a useful metric. We’ve seemingly achieved Artificial General Knowledge, but not Artificial General Capability. We should start specifying which of these a model has achieved, since without the second I’m wagering these models have limited macro impact.