But deep learning “is effective” only for a constrained list of difficulties. Altman is aware of this: humanity discovered an algorithm that might genuinely, truly study any distribution of data (or really the underlying “guidelines” that produce any distribution of data)