Reading some of the case studies cited in the survey, it's a pity all those insights can't somehow feed back into more recurring, systematic 'real-world evals' of the systems' usefulness + risks. So much nuances and little benefits/risks that probably get overlooked in evals that are 1-2 steps removed from real-world use.
Reading some of the case studies cited in the survey, it's a pity all those insights can't somehow feed back into more recurring, systematic 'real-world evals' of the systems' usefulness + risks. So much nuances and little benefits/risks that probably get overlooked in evals that are 1-2 steps removed from real-world use.