Computing Beyond Turing - Jeff Hawkins
Coaxing computers to perform basic acts of perception and robotics, let alone high-level thought, has been difficult. No existing computer can recognize pictures, understand language, or navigate through a cluttered room with anywhere near the facility of a child. Hawkins and his colleagues have developed a model of how the neocortex performs these and other tasks. The theory, call Hierarchical Temporal Memory, explains how the hierarchical structure of the neocortex builds a model of its world and uses this model for inference and prediction. To turn this theory into a useful technology, Hawkins has created a company called Numenta. In this talk, Hawkins will describe the theory, its biological basis, and a software platform created by Numenta that allows anyone to apply this theory to a variety of problems. Part of this theory was described in Hawkins' 2004 book, "On Intelligence".
This talk is by the Chairman of the Redwood Neuroscience Institute and co-founder of Palm Computing and Handspring, and is co-sponsored by Calit2 at UCSD, the Jacobs School's Computer Science and Engineering (CSE)department, and the Institute for Neural Computation (INC).
Пікірлер: 26
So great to watch Jeff's old presentation. 12 years passed. I will watch every Jeff's video available online. Thanks for The Thousand Brain Theory, Jeff. You are my hero.
I'm glad i'm still 20, and the world will drastically change in my lifetime. Jeff is one of those historical figures, it seems
Hawkins is heroic. What a wonderful human being! I wish him the best of success, but he does't need my wishes. People like Hawkins and Monica Anderson should be the heroes of pop culture.
we will be able to built very sophisticated general AI system using HTM, heuristic based learning algorithm soon.. some people have already done it..but there is still lot to do.. regarding machine consciousness..I guess we need a good theoretical framework first.. Cristoph Koch, Giulio Tononi and many other scientists are already working on it
0:19.00 . That is true but baby's move while their body is in development in the womb. So there have to be some deep neural activity that will always continuously stimulate and gets stimulated in response. When your born you don't know anything. Thus why do we recognize structure, what is it that drives a young developing body to go beyond auto-stimulus into sensing the world seeing , feeling, hearing. A baby lie down , A baby roles ,A baby crawls ,A baby grabs and baby explores an object.
this is really good! many people are still getting confused with artificial intelligence and machine consciousness..
It contradict with Turings' theory about AI. But not with Turing's theory about computing (Turing machines).
Wow. He crammed a 3 hour lecture into one.
But I don't understand how can this theory contradict with Turing's theory if they use a Turing machine to implement the model?
This is interesting. Doing beyond algorithmical approach in problem-solving. I had these ideas myself, though not as clear as all that. I wonder how many people could watch and understand this video, and whether we will need a new terminology for computer science soon, because the current language does more to obfuscate the general picture then it does to explain the details.
Regarding object recognition at around 14.00: computers can tell dogs from cats now flawlessly, as a consequence of repeated exposure to images. His question of object imagination was solved simultaneously I think - the moment a computer has a knowledge base, it can 'dream' up objects from this knowledge base.
interesting
He explains in the last few minutes of the video. Turing machines are universal, so they can simulate (or emulate) any other type of machine.
@PornoSatan I guess we'll just have to have people compose a sonnet instead.
GPT-3 is a pre-trained linguistic prediction model that is accomplishing a lot of things speculated by Jeff in this talk. We’ve come a long way and surprisingly a lot of predictions are coming true.
@TheBilly
2 жыл бұрын
GPT-3 doesn't understand a thing. It's a cute parlor trick though. When you do very big sequence modeling of text strings fun things come out but it's no more "on its way" to understanding than a pogo stick jumping higher and higher is on its way to the moon
This theory contradicts Turing's theory about Artificial Intelligence (namely that behavior characterizes artificial intelligence) but not his theory that Turing machines can compute any computable problem. Hawkins believes prediction demonstrates intelligence not behavior. Of course the algorithm Hawkins touts could be simulated in a turing machine. After all, the brain is a finite state machine. Turing formulated more than one theory in his life.
@Sciencegek while there is truth in the idea that science is modern religion, it should be acknowledged that certain beliefs are not congruent with a reality that involves man made machines capable of any feat of mental awareness that man has achieved. we are lucky that people are willing to suspend the construction of their belief system until more reliable evidence can be provided by science.
@321lawc This has nothing to do with atheïsm. This is just scientific evolution : ) Don't bring in the old intelligent design (creator) vs "evolutionary" debate. The scientific community (in particular cognitive science) is making nice progress, but we still understand shit :) (and no i'm totally not religious). btw Atheism or science are also just matrices for filtering information and structure our environment for better "understanding". They are religions in essence to.
Very impressive but not "Beyond Turing". All his ideas are computable in the Turing sense and his programs run on regular computers, i.e. Universal Turing Machines. In fairness Hawkins is careful to clarify that he has potentially advanced the intelligence of machines, but has not created machine awareness.
@IceyGrooves ...and invents prosperity linear to peaceful time.
Turning was a weird
mommy toomee bowload