Which option best describes a common drawback of machine learning approaches?

• A. They are less predictable, require substantial data and computational resources
• B. They are less predictable but require minimal data
• C. They are completely deterministic
• D. They never require data

Answer: (A)

A common drawback of machine learning approaches is that they can be less predictable and require substantial data and computational resources. In practice, models often behave differently when exposed to new or shifted data, because performance depends on how well the training data represent real-world situations. This unpredictability is tied to the need for large, varied datasets to learn robust patterns and to the inherent randomness in training processes (like initialization and stochastic optimization). Additionally, many powerful models demand significant computing power and storage, both for training and sometimes for deployment, which can be costly and time-consuming.

Other statements don’t fit because machine learning is not completely deterministic—randomness in training can lead to different outcomes between runs. It also isn’t data-free; learning requires data to identify patterns. And training effectively typically requires more than a minimal amount of data to generalize well.