AI systems like modern language models operate as mysterious "black boxes" with billions of neural functions that researchers cannot fully understand or explain, unlike early transparent computers like Deep Blue, posing trust and safety problems.

We Don’t Really Know How A.I. Works. That’s a Problem.

You have been granted access, use your keyboard to continue reading.Credit...Illustration by Max GutherFor us to trust it on certain subjects, researchers in the growing field of interpretability might need to learn how to open the black box of its brain.Credit...Illustration by Max GutherApril 15, 2026When Deep Blue, IBM’s chess-playing supercomputer, beat Garry Kasparov in 1997, computers were still just computers. Deep Blue weighed more than a ton, had 32 central processing units and could ev...