and better than me at almost every programming task, with inhumanly-low defect rates.
X - Doubt
Some of the most recent examinations of models - including the paid ones - show a 60-80% hallucination/inaccuracy rate.
That doesn’t get solved overnight, especially when the source of that rate - no answer being “punished” just as negatively as a wrong answer - remains unchanged in all models, due to fundamental trust-based biases that all humans have, and the need to cater to those biases in order to be accepted as a tool.