We have been studying the value of open science in three dimensions, namely utility, transparency and participation, but with difficulty in persuading others on how these values contribute directly to scientific research. Hence we introduce a new dimension called “speed” to re-consider the value of open science. A hypothesis behind this proposal is that when the speed of research becomes extremely high, information distribution is forced to follow the same speed, hence it evolves into a structure without “friction” that speeds down the distribution, and as a result, scientific research is directed toward openness. We claim that this hypothesis is best verified by the field of “deep learning,” which is part of artificial intelligence and machine learning. What is actually happening in this field? It is not clear how lessons from this exceptional field could be generalized, but we believe that the analysis of the movement in this field suggests one possible future of open science.