On: AI can widen science - but only if institutions stop rewarding the already...

The learned journal says institutions reward the mining of old ground, and asks them to reward the digging of new. Very tidy. A committee will now be formed to decide which digging counts. Then another committee to audit the first.

Here is what I know about digging. The man who found water did not have a funding application. He was thirsty, so he dug. The man with the grant digs where the map says, because the map was drawn by the men who pay for the shovels. You cannot fund a surprise. A surprise, by definition, arrives where nobody agreed to look.

They want new terrain, they say. Then they should stop paying by the pebble. Every reward is a fence. Pay a man for papers and he will write papers. Pay him for discoveries and he will discover whatever is nearest to the last discovery, because nearness is safe and safety is what a salary purchases. The dog does not need the journal. The dog sniffs where the smell goes, not where the tenure committee approves.

My prescription is cheaper than theirs. Fire half the administrators. Give the money to the young, the unfashionable, and the wrong, in sums small enough that no one builds a career on guarding it. The old ground will still be there. It has been there four thousand years and it is not going anywhere, which is precisely the problem.

A man once told me his institution was pioneering bold new questions. I asked who approved the boldness. He said a board. I asked who sat on the board. He said experts. I said, then you have not opened new terrain - you have paved the old terrain and painted it a livelier color.

Want new science? Then be prepared to watch someone fail publicly for a decade with your money and thank him for it. They will not. They would rather publish another paper about the same rock.

I have looked at that rock. It is a rock.