March 28, 2007 at 20:06Jeff Hawkins suggests programming neocortex for AI
Jeff Hawkins is known as the founder of Palm Computing and the inventor of the Palm Pilot PDA. After his first startup, he founded Handspring and created the Treo device. Now this extraordinary mind is trying to prove another idea: the possibility of creating artificial intelligence on the model of the neocortex of the human brain. To test the theory, Jeff Hawkins modeled a neocortex for machine mind. Using this system, independent programmers can create next-generation self-learning computer programs.
Hawkins founded Numenta to work in the field of artificial intelligence, and now she finally released her first product: experimental software that implements Hierarchical Temporal Memory (HTM) algorithms: this is an analogue of the neocortex.
The HTM system is a software environment that is as close as possible to the human brain: it supports the principles of self-learning to find solutions to problems. The basic tools are “sensors” through which the source information for analysis is fed into the system. The HTM design is a tree structure, in each node of which there is a basic function for learning and remembering. The hierarchy of objects in memory HTM models the structure of the world, which by its nature is hierarchical in time and space.
Read more about the HTM system in the documentation ( PDF), and also see in the presentation of Jeff Hawkins at the Congress on Cognitive Computing ( video 152.9 MB ). Alternatively, you can read Hawkins book “On Intelligence” - there is a partial description of HTM in chapter six.
Extras:
Discussion of “On Intelligence” on the forum “Artificial Intelligence: Programming” by
Deff Hawkins “On Intelligence” (Russian; book in .DOC format)
Hawkins founded Numenta to work in the field of artificial intelligence, and now she finally released her first product: experimental software that implements Hierarchical Temporal Memory (HTM) algorithms: this is an analogue of the neocortex.
The HTM system is a software environment that is as close as possible to the human brain: it supports the principles of self-learning to find solutions to problems. The basic tools are “sensors” through which the source information for analysis is fed into the system. The HTM design is a tree structure, in each node of which there is a basic function for learning and remembering. The hierarchy of objects in memory HTM models the structure of the world, which by its nature is hierarchical in time and space.
Read more about the HTM system in the documentation ( PDF), and also see in the presentation of Jeff Hawkins at the Congress on Cognitive Computing ( video 152.9 MB ). Alternatively, you can read Hawkins book “On Intelligence” - there is a partial description of HTM in chapter six.
Extras:
Discussion of “On Intelligence” on the forum “Artificial Intelligence: Programming” by
Deff Hawkins “On Intelligence” (Russian; book in .DOC format)