For years, volunteers spent their evenings reconstructing, line by line, the code of old video games. Slow, difficult, rewarding work. This week, their reference site closed: an AI now does in a few days what used to take them years.
Their story raises a question that concerns all of us, well beyond video games. Why carry on doing ourselves what a machine does better and faster?
The chess precedent
In 1997, the computer Deep Blue beat the world chess champion, Garry Kasparov. Many predicted the end of the game. Why play, when the machine is unbeatable?
Almost thirty years later, more people than ever are playing chess. Millions of games are played online every day. Computers did not kill them: they became training partners, teachers, analysis tools. Humans carry on playing each other, because what matters is not finding the best move. It is searching for it, against someone.
In 1923, the British mountaineer George Mallory was asked why he wanted to climb Everest. His answer became famous: because it is there. In 2005, a French test pilot, Didier Delsalle, landed a helicopter on the summit of the world's highest mountain, a first. Since then, thousands of mountaineers have nevertheless carried on climbing on foot, at the cost of immense effort. Nobody has ever climbed Everest in order to get to the top as fast as possible.
The result and the path
We need to distinguish between two kinds of activity.
Some we do for the result. We decompile a game to obtain its code, we translate a text to understand it, we calculate to know the answer. For those, if a machine does better, it is reasonable to hand it the task.
Others we do for the path. For the pleasure of solving, the pride of getting there, the company of those searching alongside us. Those, no machine can take from us. The pleasure of solving a crossword does not disappear because solutions exist at the back of the newspaper.
In the eighteenth century, the German philosopher Lessing wrote that if he were offered the choice between all truth, handed over in one go, and the mere search for truth, he would choose the search. It was not affectation: it was the idea that searching is a way of being alive.
What really broke
The case of the decompilers is instructive, though. What broke their community was not only AI. It was a counter: the percentage of progress displayed for each project. As long as it measured human effort, it motivated. When it became possible to push it up with an AI, it no longer measured anything.
The lesson is valuable. AI mainly threatens activities whose meaning we have ended up measuring by the result. If you play chess for the ranking, or if you search only for the ticked box, the machine will win. If you search for the pleasure of searching, it has no hold.
That said, we should not play down what has been lost. A community dissolving means friendships, habits, a collective pride. That grief is real. Other reasons to come together will have to be invented.
What to take away
AIs are finding more and more answers. That is good news, when you need the answer.
But they will never be able to give us the pleasure of having searched for them. That one, we will have to keep giving ourselves.