Escalating Hard Labels to Experts

Why is obtaining labeled data easier for tasks that humans perform well?

For tasks that humans perform well, human labelers can provide high-accuracy labels to train ML algorithms.

Andrew Ng writes that since people recognize _____ images well, it is straightforward for human labelers to provide high-accuracy labels.

Match each concept to its correct description regarding human labeling of human-solvable tasks.

Order the steps for deciding whether human labelers are a good fit for collecting labeled data for an ML task.

In Andrew Ng's medical imaging example, what error rate can a team of doctors achieve when providing labels?

According to Machine Learning Yearning, the ease-of-labeling advantage from human labelers applies only to image recognition tasks.

In Andrew Ng's medical imaging example, a team of _____ can provide labels at a 2% error rate.

Match each labeling scenario to the property that best explains it, based on Machine Learning Yearning.

Order Andrew Ng's reasoning steps for why human-solvable tasks benefit from human labelers.