It's quite possible that the future of artificial intelligence is here. And if it is, as some in the know are coming to believe, then it is called HTM, or Hierarchical Temporal Memory.
The Future of Artificial Intelligence
HTM is a machine learning model developed by Jeff Hawkins and Dileep George of Numenta, Inc. Jeff Hawkins is an engineer, entrepreneur, and author of the 2004 published book about A.I. titled On Intelligence 
.
Dileep George is a former electrical engineering graduate student who was working part-time at the Redwood Neuro-Science Institute, a small think tank founded by Hawkins in 2002 (it's now part of UC Berkeley). At the same time he was working at the Institute he was looking for a PhD topic about the brain and decided he wanted to team up with Hawkins.
The possible future of artificial intelligence is based on a key insight that Hawkins had in the classic "Eureka" moment back in his UC Berkeley PhD study days. As he pondered the question of how he would react if a blue coffee cup were to quite suddenly materialize on his desk it came to him.
Hawkins' insight may hold the key to the future of artificial intelligence development: that all former A.I. models and theories eventually crumbled or came up against impassable walls because they did not grasp the core fact that the human brain's capacity is centrally based on predicting the future; both the immediate future and the far-flung future.
Hawkins is completely fascinated by the concept that "when you are born, you know nothing," and he and George have begun creating and expanding a software platform that transfers this principle to artificial intelligence design: when it's "born", it knows nothing. But, like human brains, this A.I. will have "instincts"--the human-programmed protocols for learning.
HTM is the likely future of artificial intelligence because with this platform, an A.I. must create memories from sensory input, learn from experience, and then use those remembered and processed experiences to make predictions about the future.
This goes right down to the basics of recognizing what an object is from only obscured, partial, or distorted sensory data--something that human beings take doing for granted--but which the world's chess champion computer program would hopelessly fail at.
Just as a little tiny infant begins doing instinctively from the second it emerges from the womb, such an A.I. grows itself--it starts from knowing nothing and learns from having and then assimilating experiences.
Future of Artificial Intelligence
HTM programming rests on the core concept of hierachical knowing. This practical insight comes from the observation that the human brain's neocortex--the top layer of our three layers of gray matter--makes up at least 60% of the human brain by itself.
Almost all higher level insights and calculations, such as Hawkins' own revelation about the basic structure of human intelligence, are made within the neocortex. Fun Fact: Humans is the largest neocortex, and takes up the highest percentage, of that of any mammal on earth... by far.
The future of artificial intelligence will be built upon something out of the ancient past: the pyramid. This means that A.I. "neurons" are arranged hierarchically, which Hawkins, backed up by a great deal of neurological research, presumes to be the situation with human brains. Lower-level neurons are like those which receive the basic sensory input from, say, a musical instrument.
These tones are first picked up by the ear, then interpreted by those lower-level neurons, which process information about them and then kick them up to mid-level neurons. The mid-level neurons then follow suit but with higher processing, and then kick them up to the top.
Then the top makes the ultimate processing and then translates signals about those stimulations back down to the low-level neurons, "teaching" them like a "more experienced" guide or mentor. Just as there are fewer geniuses than average-intelligence human beings, so there are fewer higher-level neurons than mid-level or lower-level neurons. The future of artificial intelligence is hierachical in structure, just like human brains, just like human society.
It was George, over the course of several weekends, who constructed the original basic, elemental A.I. model of the process used in the human visual cortex. The usual modeling of a program is linear; it processes data and makes calculations in just one direction.
George then created multiple, parallel layers of nodes, with each node embodying thousands of neurons in cortical columns and each in and of itself a tiny computer program with a unique ability for processing information, remembering patterns, and making predictions.
Hawkins' company Numenta has been at work with Edsa Micro. Adib Nasle, Edsa's president, says of HTM-based A.I., "We've seen some incredible speed improvements. Some approaches, you give too many examples and they get dumber. HTM seems not to suffer from that. It's pretty impressive."
The most likely future of artificial intelligence
