Commentaries and excerpts about "On Intelligence" by Jeff Hawkins

An excellent book for those that are in search for a real world (a.k.a.materialist) explanation for the conscience and reasons for us being able in the near future of building intelligent and / or conscient machines

• Intelligence must not be defined in terms of behavior but in terms of prediction capability. Behavior is something that comes from prediction capability and not the other way around.

• Our brain is a correlation machine. It matches inputs from the senses with series of inputs stored in the synapses and forecasts what will happen next. If in the short term, forecast matches the new input sequences, an "OK" feedback is passed by higher regions of the neocortex to its lower regions. If not, adjustments need to be done. This is called "attention to something".

• The stored sequences are called "Invariant Representations". This idea comes from Plato, although the forms are neither perfect nor stored in heaven but stored in the brain itself. They are invariant in the sense that a long series of inputs matching new patterns (and consequently making void the stored forms) must be presented to the brain for the "Invariant Representation" to be changed.

• What comes from the definition explained above is that our brain is composed of similar parts grouped together in an hierarchical system. The correlation and prediction / correction algorithms are the basic part of our neocortex (a kind of machine language programming).

• Also, on a mathematical point of view, our brain works continuously making Fourier and Principal Components Analysis.

I have posted several quotations from the book itself below, and leave them for your comments:

• "Intelligent machines will arise from a new set of principles about the nature of intelligence."

• "They [engineers] will continue to fail as long as they keep ignoring the differences between  computers and brains"

• "Redwood Neuroscience Institute dedicated to brain theory and  finding a theoretical understanding of the neocortex"

• "The biggest mistake is the belief that intelligence is defined by intelligent behavior"

• "The brain uses vast amounts of memory to create a model of the world. Everything you know and have learned is stored in this model. The brain uses this memory-based model to make continuous predictions of future events. It is the ability to make predictions about the future that is the crux of intelligence. I will describe the brain’s predictive ability in depth; it is the core idea in the book."

• Scientific American - September, 1979 - Special issue on brain

• John Searle - Chinese Room Experiment

• Three key factors: a)  the inclusion of time in brain function, b) the importance of feedback, c)  any theory should account for the physical architecture of the brain

• "Backpropagation doesn't qualify as feedback because it only works in the learning phase"

• "Almost none of them [neural networks, circa 2003] attempt to capture the overall function or architecture of the neocortex."

• "Both [neural networks and AI programs] are fatally burdened by their focus on behavior. Whether they are calling these behaviors “answers,” “patterns,” or “outputs,” both AI and neural networks assume intelligence lies in the behavior that a program or a neural network produces after processing a given input. The most important attribute of a computer program or a neural network is whether it gives the correct or desired output. As inspired by Alan Turing, intelligence equals behavior."

• "[There were] networks that didn’t focus on behavior. Called auto-associative memories, they were also built out of simple “neurons”."

• "they were interconnected differently, using lots of feedback. Instead of only passing information forward, as in a back propagation network, auto-associative memories fed the output of each neuron back into the input"

• "[To retrieve a memory from it] you don’t have to have the entire pattern you want to retrieve in order to retrieve it."

• "unlike most other neural networks, an autoassociative memory can be designed to store sequences of patterns, or temporal patterns. This feature is accomplished by adding a time delay to the feedback. With this delay, you can present an auto-associative memory with a sequence of patterns, similar to a melody, and it can remember the sequence."

• the brain uses circuits similar to an auto-associative memory to do so

• "Auto-associative memories hinted at the potential importance of feedback and time-changing inputs. But the vast majority of AI, neural network, and cognitive scientists ignored time and feedback."

• "Neuroscientists ... know about feedback ... but ... no theory ... beyond ... talk of "phases" and "modulation". ... time has ... no central role in most of their ideas on overall brain function. They tend to chart the brain in terms of where things happen, not when or how neural firing patterns interact over time."

• "[functional imaging tells you WHERE things are happening but not WHEN and HOW they are happening. Telling somebody to repeatedly perform a task is not good enough of an experiment exactly because the second time you perform a task your brain is not in the same mental state as before."

• "We don’t have machines that think, we have machines that do. Even as we observe our fellow humans, we focus on their behavior and not on their hidden thoughts."

• "However ... our intuition is often the biggest obstacle to discovering the truth. Scientific frameworks are often difficult to discover, not because they are complex, but because intuitive but incorrect assumptions keep us from seeing the correct answer."

• "Intelligence is something that is happening in your head. Behavior is an optional ingredient. [Since] behavioral equivalence is not enough... intelligence is an internal property of a brain, [therefore] we have to look inside the brain to understand what intelligence is."

• "there is an underlying elegance of great power, one that surpasses our best computers, waiting to be extracted from these neural circuits"

• "Connectionists intuitively felt the brain wasn’t a computer and that its secrets lie in how neurons behave when connected together."

• [EU: interessante perceber que toda a teoria de redes neurais se deu a partir de redes com três camadas. Eu sempre pensei que o ideal seria aumentar a quantidade de níveis para justamente aumentar a capacidade de decisão da rede. No entanto nunca imaginei algo parecido com as redes auto-associativas. Feedback para mim era sinônimo de "back-propagation", no entanto o que o autor propõe é implementar o feedback como parte da arquitetura operacional da máquina em modo permanente!]

• FUNNY: "Your neocortex is reading this book"

• "I am not interested in building humans. I want to understand  intelligence and build intelligent machines. Being human and being intelligent are separate matters."

• "Every moment in your waking life, each region of your neocortex is comparing a set of expected columns driven from above with the set of observed columns driven from below. Where the two sets intersect is what we perceive"

• "behavior and perception are almost one and the same"

• "motor behavior must also be represented in a hierarchy of invariant representations"

• "Conscience is what we feel when we have a neocortex!"