To find √2 without a root key, iterate: take a guess x, average it with 2/x. Starting from 1.5 the first step gives 1.4166666666667 (second link), and squaring that result gives 2.0069444444445 (third) — the amount by which it overshoots two is the amount by which the guess is wrong, and it is already down in the third decimal. The true root, on the first link, is 1.4142135623731, and the last link squares it and prints exactly 2.
The doubling is not luck. Write the error as x = √2(1 + ε); then 2/x = √2(1 − ε + ε² − ⋯) and the average of the two is √2(1 + ε²/2 + ⋯), so the error is replaced by its square. Two correct decimals become four, four become eight, eight become sixteen — and since the screen carries fourteen digits, four steps from 1.5 is the whole of the work: the first lands on 1.4166666666667, the third is right in its first dozen places, the fourth cannot be told from the root. The fourth link shows the limit case: feed the true root back into the map and the same fourteen digits come out again, because √2 is a fixed point of the averaging, exactly what the algebra says it must be. This is the oldest numerical algorithm still in daily use, and every square-root instruction in a processor today is either this iteration or a variant of it started from a lookup table.
Answer
1.4142135623731
How to read it
The root itself, and what the map does to it. The second link is one Newton step from 1.5, the third squares that step, the fourth iterates on the root, the fifth squares the root.
Source
Newton's method of 1671 as published in Raphson's 1690 account, applied to x² − 2; the iteration x → (x + a/x)/2 is the Babylonian square-root algorithm, on the tablet YBC 7289 around 1800 BC
Work it in the calculator
These links open the right calculator with this problem's numbers already typed in. The line above is what the screen should show — check it yourself.