In narrow places this is already measured. In others it is nowhere near true. The sentence above is a forecast, and here is what it rests on.
Where machines already beat our best. GPQA Diamond is a set of graduate-level science questions. People who hold or are studying for a PhD in the relevant field score 65 percent on it.1 The leading systems now score around 94 percent.2 In July 2025 a Google DeepMind system reached a gold-medal score at the International Mathematical Olympiad.3
Where they are still behind a child. On ARC-AGI-3, humans solve 100 percent of the environments. As of March 2026, frontier systems scored below 1 percent.4
So the sentence is about a day that has not arrived. In the largest survey of its kind, 2,778 researchers who publish in top AI venues put the chance of machines outperforming humans at every possible task at 10 percent by 2027, and 50 percent by 2047.5
Nobody can hand you the date, and I am not going to pretend otherwise. The question is worth sitting with before that day arrives rather than after it.
Current as of 8 August 2026
What should we dig into?
Tap any that matter to you. Press and hold one to crown it your number 1.
Will you be there?
Sign in with Google so we know who you are and you can change your answer later. Nothing is posted anywhere.