Robustness to the vagaries of natural language
This substack looks at the capabilities which the current wave of AI actually produced, that is even if all advances in AI ended today, what things would likely never go back to being done the same way or just no longer be done at all?