Machine Learning in Science

A machine, receiving at distant times and from many hands new combinations and improvements, and becoming at last of signal benefit to mankind, may be compared to a rivulet, swelled in its course by tributary streams until it rolls along a majestic river, enriching in its progress provinces and kingdoms.
In retracing the current, too, from where it mingles with the ocean, the pretensions of even ample subsidiary streams are merged in our admiration of the master-flood. But, as we continue to ascend, those waters which, nearer the sea, would have been disregarded as unimportant, begin to rival in magnitude, and divide our attention with, the parent stream; until, at length, on our approaching the fountains of the river, it appears trickling from the rock, or oozing from among the flowers of the valley. So, also, in developing the rise of a machine, a coarse instrument or a toy may be recognized as the germ of that production of mechanical genius whose power and usefulness have stimulated our curiosity to mark its changes and to trace its origin. The same feelings of reverential gratitude which attached holiness to the spots whence mighty rivers sprung, also clothed with divinity, and raised altars in honor of the saw, the plough, the potter's wheel, and the loom.

Take a look at George Gamow, who is now recognized as one of the great cosmologists of the last hundred years. I speculate that he probably didn't win the Nobel Prize because people could not take him seriously. He wrote children's books. His colleagues have publicly stated his writing children's books on science had an adverse effect on his scientific reputation, and people could not take him seriously when he and his colleagues proposed that there should be a cosmic background radiation, which we now know to be one of the greatest discoveries of 20th-century physics.

We are becoming the servants in thought, as in action, of the machine we have created to serve us.