A similar conundrum took place in the world of AI research in the early 90s'.
Most AI researchers up to this point insisted on storing a "map" of reality in the AI's memory as a way of "remembering" where in the world the robot is.
If robot walked 1 meter forward - robot's internal map was updated to reflect its new position on the map.
Or if the robot was to perform a multi-step movement it had to "plan" 5 or 6 steps ahead while taking into consideration all the things that COULD go wrong. Multi-step planning was like playing chess.
Except when it wasn't. Because even though it may have intended to take 1 meter step, it only took a 98cm step (because the floor is slippery), or because you lifted it off the ground. Eventually the robot's internal state of "where it is in reality" became misaligned with where it actually was and even though it knew how to handle SOME contingencies, it had no clue how to deal with unexpected scenarios (like when its internal state no longer represents reality). Robots couldn't "think on their feet" as we say and they couldn't deal with unexpected external events. Chess has rules (limits) to what is a "valid move" - reality doesn't.
So now the focus is on real-time information-processing and orientation. Once you have a map of the "world" in memory - you orient/calibrate yourself using landmark recognition. If you think you are lost - take a reading. Calibrate yourself.
Once you have a goal set, algorithms can help you work towards that goal even if you encounter unexpected circumstances. It just creates a sub-goal: get back on track!
You can read more about it here:
https://en.wikipedia.org/wiki/Nouvelle_AI
The original paper was called "Elephants don't play chess" - which is to signify that real-time decision-making is cheaper/easier than long-term planning given all the variables in a complex world.
In practice it's a space-time trade-off:
https://en.wikipedia.org/wiki/Space–time_tradeoff
You need less space (memory) to represent reality - but you need more time for calibration and effective decision-making.