The third age of computing - The Computing Universe: A Journey through a Revolution - page 330

ball in only about a hundred steps. The brain clearly solves the problem a dif-
ferent way than relying on conventional computation. According to Hawkins,
it uses memory:

How do you catch the ball using memory? Your brain has a stored memory
of the muscle commands required to catch a ball (along with many other
learned behaviors). When a ball is thrown, three things happen. First, the
appropriate memory is automatically recalled by the sight of the ball. Second,
the memory actually recalls a temporal sequence of muscle commands. And
third, the retrieved memory is adjusted as it is recalled to accommodate the
particulars of the moment, such as the ball's actual path and the position of
your body. The memory of how to catch a ball was not programmed into your
brain; it was learned over years of repetitive practice, and it is stored, not
calculated in your neurons. 17

To account for the fact that the position of the ball needs to be constantly
adjusted as the ball comes toward us, Hawkins uses the idea that the mem-
ories stored in the cortex are actually invariant representations . Artificial auto-
associative memories can recall complete patterns when given only a partial
image as input. But ANNs have a hard time recognizing a pattern if the pat-
tern has been rescaled, rotated, or viewed from a different angle - a task our
brains can handle with ease. If you are reading a topic, you can change your
position, rotate the topic, or adjust the lighting, so that the visual input of
the topic to your brain can be constantly changing. Yet your brain knows
that the topic is the same, and its internal representation of “this topic” does
not change. The brain's internal representation is therefore called an invari-
ant representation . The brain combines such invariant representations with
changing data to make predictions of how to perform tasks, such as catching
a ball.
Our understanding of the world is tied to our ability to make such predic-
tions. Our brain receives a constant stream of patterns from the outside world,
stores them as memories, and makes predictions by combining what it has
seen before with the incoming stream of information. Hawkins says:

Thus intelligence and understanding started as a memory system that fed
predictions into the sensory stream. These predictions are the essence of
understanding. To know something means that you can make predictions
about it. 18

This idea is the basis of Hawkins's memory-prediction framework of intelligence:
“Prediction not behavior is proof of intelligence.” 19 According to this view of
intelligence, intelligent machines could be built that have just the equivalent
of a cortex and a set of input sensors. There is no need to connect to the emo-
tional systems of the other, older regions of the brain. Such intelligent sys-
tems will not resemble the humanoid robots of science fiction but would be
able to develop an understanding of their world and make intelligent predic-
tions. However, the technical challenges of building such systems in silicon still
remain formidable, both in terms of the number of neurons required and their
vast connectivity requirements.

B.16.6. Daniel Dennett is a philoso-
pher and cognitive scientist who has
written popular topics on evolu-
tion and consciousness - Darwin's
Dangerous Idea and Consciousness
Explained . He is codirector of the
Center for Cognitive Studies at Tufts
University in Massachusetts.