Deep Thinking: Where Machine Intelligence Ends and Human Creativity Begins

by Garry Kasparov

Later, programming techniques were developed that allowed programs to “fantasize” a little by looking at hypothetical positions away from the search tree, but this came at the cost of slowing the main search. Much more success was had with ways to make the search smarter and deeper with techniques like “quiescence search” and “singular extensions” that tell the algorithm to deeply examine variations that meet special conditions, such as piece captures or the king being in check. It’s a slight wave toward the old Type B programs and the dream of playing chess like a human and prioritizing certain moves early on, but it’s still search, not knowledge. These clever techniques went a long way toward eliminating the horizon effect in practical play, as did ever-faster chips.