From root: Why should I simulate the real world?
When is data good enough for black box AI?
When it is relevant, complete, plentiful and correct, which for many high-value problems it is not.
For many high-value problems that humanity (or just businesses) would like solve, the truth is that data is often irrelevant, incomplete, scarce or wrong.
In other words, you often do not have data: which is directly relevant the the problem; which forms a complete set; which is enough for model training; or which is without any false entries, etc.
In situations where you do actually have data which satisfies these (or most of these) conditions, then training black box AI models becomes viable.
References and further reading
Hardwick, Robert J. 2025. “What is black box AI?” umbralcalculations, December. https://umbralcalc.github.io/posts/what_is_black_box_ai.html.
Sambasivan, Nithya, Shivani Kapania, Hannah Highfill, Diana Akrong, Praveen Paritosh, and Lora M. Aroyo. 2021. “‘Everyone Wants to Do the Model Work, Not the Data Work’: Data Cascades in High-Stakes AI.” Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems, 1–15. https://doi.org/10.1145/3411764.3445518.
Questions whose answers cite this one