Slashdot Mirror
Jeff Hawkins' Cortex Sim Platform Available
UnreasonableMan writes "Jeff Hawkins is best known for founding Palm Computing and Handspring, but for the last eighteen months he's been working on his third company, Numenta. In his 2005 book, On Intelligence, Hawkins laid out a theoretical framework describing how the neocortex processes sensory inputs and provides outputs back to the body. Numenta's goal is to build a software model of the human brain capable of face recognition, object identification, driving, and other tasks currently best undertaken by humans. For an overview see Hawkins' 2005 presentation at UC Berkeley. It includes a demonstration of an early version of the software that can recognize handwritten letters and distinguish between stick figure dogs and cats. White papers are available at Numenta's website. Numenta wisely decided to build a community of developers rather than trying to make everything proprietary. Yesterday they released the first version of their free development platform and the source code for their algorithms to anyone who wants to download it."
Someone needs to put this Cortex Simulator in an 8-legged, hydraulic-actuated, 10 ton spider-machine. If you think that's a crazy idea, you suck.
I'm still a bit confused as to how he is so confident that this is how the neocortex works given that this is still one of the 23 unsolved problems in system neuroscience. But hey, he made a lot of money off Palm, that gives him way more street cred than people who have been working on this problem for their whole lives.
Do not try to read the dupe, thats impossible. Instead, only try to realize the truth
What truth?
There is no dupe
I read that and thought "a new, more advanced algorithm for breaking CAPTCHAs"
[Fuck Beta]
o0t!
I for one welcome our open source neocortical robot overlords...
High quality versions of Jeff Hawkin's talk at UC Berkeley are available here.
NEOcortex - Begin the Matrix jokes/analogies now...
Libertarian Leaning Political Discussion Forum.
"Yesterday they released the first version of their free development platform and the source code for their algorithms to anyone who wants to download it."
And can understand it.
Yeah, but can it distinguish the invention of PalmOS Graffiti from the invention of PARC Unistroke? That would have been handy...
[
This sounds REALLY cool. Even if all it amounts to is a set of computationally-expensive toys, it's still the basis for being able to boil down the essentials and costs of self-learning systems. That, and perhaps the stepping stone to being able to have the hitchhiker-like "real people personalities".
Then, of course, there's always the dream of eventually being able to really 'get into the code' and debug it from the inside, leading to the soviet joke where "the code debugs you."
Ryan Fenton
Buy Steampunk Clothing Online!
How unusual to see software that will run on OS X or Linux, but there is no Windows version. Shape of things to come I hope.
You need to check out http://www.fakenamegenerator.com/.
Here is an example (all information is fictional):
Emily F. Garza
1587 McDowell Street
Nashville, TN 37238
Email Address: Emily.F.Garza@spambob.com
Phone: 931-299-3591
Mother's maiden name: Mennier
Birthday: August 12, 1950
of course, you might want to use a disposable e-mail like pookmail.com so you can retrieve any e-mail they might send.
Having read the Hierarchical Temporal Memory (HTM) white papers, and knowing something of the area prior to that, it looks like Jeff Hawkin's and his company have take a lot of ideas and algorithms that exist, and hacked them together to implement his neocortex ideas.. there's a bits and pieces of graphical models, time recurrent neural nets, Boltzmann machines, etc.. It does some cool stuff but nothing that AI and machine learning people haven't been doing for years. The difference is that Jeff has taken the entrepreneurial approach to AI. Instead of publishing and allowing the academic community (the original open source movement!) to peer review and contribute, he's formed a company to announce his ideas to the world -- ready or not. This isn't necessarily bad, but the proof of his ideas will be scaling them up to start solving some useful problems. Bring on the face recognition that isn't fooled by dark sunglasses and a false mustache!
Someone needs to immediately train this to catch /. dupes and/or run Linux.
As a current student I neuroscience I would love to see this happen however there are a few major problems.
1) All the research into cortical circuitry is done in non-humans. There are definite similarities between our cortex and that of a rat, but there are also drastic differences, if there weren't then rats would be able to talk, think, and reason like we do. (Yes lots of research is being done in non-human primates, but this work is EXTREMELY expensive and even non-human primates have different cortical circuitry then we do)
(Not only are the cortices of different species drastically different, scientists often chose regions of cortex that have no correlation in humans. Many neuroscientists are studying the Barrel Cortex. It is a region of cortex that is specifically designed to integrate the signals from the whiskers of a Rodent. Humans don't have whiskers and we also don't have Barrel Cortex. Anything learned about the circuitry of the Barrel Cortex will not necessarily correlate to human cortex.
2) Intra-population Circuitry research examines very small subsets of neurons that make up a bigger populations. When studying neurons in the visual cortex for example the best anyone can do is look at the firing of about 150 neurons. When you consider that there are over 10,000,000,000 (BILLION) neurons that make up the human brain a small set of 150 neurons is almost nothing. We don't have sufficient technology to examine what each neuron in a specific population is doing.
3) Inter-population circuitry research only looks at what populations are connected to each other. Yes we know what type of neurons project from one area of the brain to the next, however, this only gives a very rough schematic of the circuitry. The circuitry of both the cerebellum and the hippocampus have been described beautifully (they have both been known for well over 50 years). However once we no this circuitry it yields no light on how the circuitry actually accomplishes its task.
4) Failure to integrate both intra and inter population circuitry. I have yet to read a paper that does a good job of integrating these two studies. Most neuroscientist pick one emphasis and stick with it. In order to understand exactly what the cortex is doing you must integrate all levels of research into your studies.
5) Study of the cortex is insufficient. The cortex projects to many regions of the brain whose functions are still unknown. These connections to these brain regions might not appear necessary but if they really weren't necessary why are they there? Back in the day people who had really bad seizures would have what is called a "Corpus Callosomy" This is the cutting of the fibers that connect the two hemispheres of the brain. At first the procedure was called a success. However, after further investigation it turned out that the people on whom this operation was performed had drastic problems. (Example, if a person was holding an object in their left hand (the sensory fibers project from the left hand to the right side of the brain) and if they weren't allowed to see the object, upon request of the examiner of what the person was holding they would respond there is nothing in their hand. ) This example is only to illustrate that upon initial examination many regions of the brain appear to have no function as lesioning these structures has no aversive effects, this is what many people thought about the corpus colosum, however upon further examination this proved untrue. Before we can understand how the cortex fully functions we must understand how the entire brain works with it.
Sorry to be a nay sayer but I have serious doubts whenever someone claims to have figured out how the cortex works.
That book was published over a year ago, lots can and has changed in that time.
Actually, its content was produced seven or eight years ago.
Its publishing date was "December 2005". But publishers will lie about the publication date of a book if it allows them to sell more books. And in this case, I wouldn't be surprised if the book came out hot off the presses in December 2004 with a postdate of "December 2005"
Furthermore, this book was based on the scientific proceedings of a conference which occurred six years before the book was finally edited (or finally published). I'm actually not sure of the year of the scientific conference itself, because the information supplied to sell the book doesn't give the actual year.
http://en.wikipedia.org/wiki/Baum-Welch_algorithm http://en.wikipedia.org/wiki/Viterbi_algorithm
The first is an alogorithm which utilizes forward and back-tracking "to find the unknown parameters of a hidden Markov model." The second is a similar algorithm used for learning 'known' causes (for reference).
I work in computational linguistics and the time an algorithm takes to run and the amount of memory it requires are serious limitations. That's why ad-hoc systems are so common.
Hawkins' solutions likely overlap with Dr. Stephen Thaler's patents for neural networks(NN). In particular, Thaler's algorithms inject noise into a proprietary NN system (actually 2 or more NNs conjoined) to generate novel patterns (that is, to _discover_ new patterns). For example, Thaler trained and used such a NN system to generate thousands of possible musical riffs which he has now copyrighted. Thaler is in business and making money.
And yet again, we see the potential of the patent system to retard progress instead of stimulate it; to favor cashing in over invention; to stifle, crush and force back progress, however isolated from the original inventor such progress may have originated. The PTO is a hive of scum and villainy.
Abolish it. It is out of hand.
I've fallen off your lawn, and I can't get up.
A truly sentient software program is mere child's play compared to the awesome potential of this guy. As the hybrid clone of Stephen Hawking and Richard Dawkins, Jeff Hawkins is destined to become one of the leading minds of the 21st century.
That's not a "debunking", that's a closed-minded opinion-fest. Reminds me of Papert's and Minsky's huge rants on how neural nets couldn't do this and that, exemplified by the (incorrect) claim they couldn't even be made to do an XOR. They published, just ran off at the mouth like college kids with their first exposure to ideas orthogonal to their thinking, then were proved soundly wrong by the facts.
Some advice for the closed minded: Judge this fellows work by his actual results; not what other people think his results may turn out to be. He's published the code, and those of us who are working in this area are very interested. That still doesn't mean we'll use his work the way he will, or that we'll get the same results. Just be a little patient and just a little less judgmental. Or not; after all, even Minsky and Papert couldn't change the facts. They turned out to be well educated, highly opinionated, deeply respected fuckups. You want to join them? Jump to conclusions. Nature's got a place for you, too. :-)
I've fallen off your lawn, and I can't get up.
I did not rtfa, who has time for that anymore ;) But regardless, in the Slashdot fashion, here is my opinion:
I played around with some of his publicly available code a few months ago. It was pretty impressive on a toy problem (recognizing a small set of characters) but was very, very slow at training (on the order of hours or days to learn the simple problem).
But on the other hand, I can't think of any sort of technology that could do better than it (I am into machine learning and AI.) Also, it is not a big deal if it trains slowly if it can compute fast - the human brain took millions of years to evolve.
From what I could tell, his technology is (was?) a glorified bayes net with time forced into the model. To train the net to recognize images he just moved an image around a bunch and had the net brute-force learn all the possible patterns. Training this could get tedious, to say the least. In theory it sounds like he's on to something, but it came off as a pretty simple modification to an old algorithm. I found his book entertaining, and insightful at times, but not revolutionary. He reasons out some really cool ideas, but it comes off more as philosophy more than science sometimes.
Hawkins is a rich guy, and no-one feels like telling him that his stuff is crap. He had a few smart people working for him at some point, but when they told him his ideas were half baked and not new, he just fired their asses.
Here is what many people in machine learning and computer vision think about Hawkins stuff:
- it's way, way behind what other people in vision and machine learning are doing. Several teams have biologically-inspired vision systems that can ACTUALLY LEARN TO RECOGNIZE 3D OBJECTS. Hawkins merely has a small hack that can recognize stick figures on 8x8 pixel binary images. Neural net people were doing much more impressive stuff 15 years ago.
- Hawkins's ideas on how the brain learns are not new at all. Many scientists in machine learning, computer vision, and computational neuroscience have had general ideas similar to the ones described in Hawkins's book for a very long time. But scientists never talk about philosophical ideas without actual scientific evidence to support them. So instead of writing popular book with half-baked conceptual ideas, they actually build theories and algorithms, they build models, and they apply them to real data to see how they work. Then they write a scientific paper about the results, but they rarely talk about the philosophy behind the results.
It's not unusual for someone to come up with an idea they think is brand new and will revolutionize the world. Then they try to turn those conceptual ideas into real science and practical technologies, and quickly realize that it's very hard (the things they thought of as mere details often turn out to be huge conceptual obstacles). Then, they realize that many people had the same ideas before, but encountered the same problems when trying to reduce them to practice (which is why you didn't hear about their/your ideas before). These people eventually scaled back their ambitions and started working on ideas that were considerably less revolutionary, but considerably more likely to result in research grants, scientific publications, VC funding, or revenues.
Most people go through that "naive" phase (thinking they will revolutionize science) while they are grad students. A few of them become successful scientists. A tiny number of them actually manage to revolutionize science or create new trends. Hawkins quit grad school and never had a chance to go through that phase. Now that he is rich and famous, the only way he will understand the limits of his idea is by wasting lots of money (since he obviously doesn't care about such things as "peer review"). In fact, many reputable AI scientists have made wild claims about the future success of their latest new idea (Newell/Simon with the "general theorem prover", Rosenblatt with the "Perceptron", Papert who thought in the 50's that vision would be solved over the summer, Minsky with is "Society of Minds", etc......).
No scientist will tell Hawkins all this, because it would serve no purpose (other than pissing him off). And there is a tiny (but non-zero) probability that his stuff will actually advance the field.
- Anonymous Scientist
After participating in the neural network hype in the 1980s (I spent 1 year on a DARPA committee for NN tools, and was the original author of the SAIC ANsim NN software product) I found Hawkin's book to be light technically, but I really enjoyed reading it.
m /
His work might have been inspired by Kohonen's classic Springer-Verlag book "Self-Organization and Associative Memory".
I downloaded their software last night but have had little time doing anything but building and running two examples. When I get 20 hours to really kick the tires, I will blog about it on my AI blog http://artificial-intelligence-theory.blogspot.co
I am hoping that NTA will really simulate temporal memory and spacial invariance that the neocortex apparently has.
A little off topic, but I love the way they package the NTA software: most of the low level code is C++, that builds into sharable libraries loaded and used in a Python wrapper. Neat stuff. The free license is only for non-commercial use, BTW.
Some people spend their entire adult lives trying to overcome alcohol addiction, or trying not to beat their spouse. To others, it comes naturally.
He may have founded Palm - and I do encourage him to push forward with Numenta - but I'm still trying to get my Palm to sync on 2 different computers ..............
Its not the years, its the mileage
I'm no expert on this topic, but this doesn't sound very new or revolutionary to me. It looks very familiar with known model theories of neuronal networks. Aren't concepts of backpropagation or pattern recognition known for ages?
http://thedialogs.org/2007/03/02/jeff-hawkinss-adv isor/
If he said that, he's very wrong. The cortex consists of dozens areas with different cyto-architectonic (that means cellular structural) properties, see http://spot.colorado.edu/~dubin/talks/brodmann/bro dmann.html for a nice map. Brodmann counted 46 of them and modern views distinguish sub-areas in most of them. E.g., BA44 (Brodmann's Area 44) is considered to be involved in language processing (amongst other things), but is usually divided into 44a, b and c (there are different ways of naming these, too; e.g. the pars opercularis or Broca's Area for BA44).
So, the cortex contains many different areas with different physical properties and these are commonly tied to specific functions: e.g. language processing involves a few areas, and motoric processing involves a few different areas, and these never overlap. Consequently, any model that wants to approach the cortex at neuronal level should account for this.
And the cortex of a whale is much larger than that of humans...
>Numenta's goal is to build a software model of the human brain capable of face recognition, object identification, driving
Great, they're going to invent the Johnny Cab
Here's a Q&A with Jeff Hawkins from May 2006 that asks him to better explain how his theory works and how it can be applied.
"Perceptrons" by Minsky and Papert was correct regarding perceptrons' limited computational expressiveness. Following that they incorrectly conjectured that their negative result would hold for 3 or more perceptron layers, which later turned out not to be true. So cut them some slack, they gave the research community significant useful results, and probably acted with best judgment.
I flipped through that (several times!) and found it interesting, but then I ended up buying Douglas Hofstadter's Fluid Concepts and Creative Analogies instead. It's a follow-up to the philosophical Godel, Escher, Bach, in which his research group tries to model creativity using computers. His general technique is different from neural network modeling. Stephen Pinker's How the Mind Works is also interesting, and well-written.
I hang out, via e-mail, with people involved in the Loebner Prize Contest, so I have kind of a skewed view of AI. The people there focus on the "chatterbot" approach, descendants of ELIZA, and some of them actually think that's a good model of intelligence. (They're wrong.) I'd like to see some kind of open-source AI project, but what I know of the existing ones is that each backer has their own fixed theory of the mind, and the confusion of programming languages and other details make it hard to coalesce around any one idea.
Revive the Constitution.
"Fiduciary" on whose behalf? Please, look the word up. It's commonly misused, so don't feel bad.
I think you meant "fiscal" or simply "financial".
If you can't take the time to look it up, remember this: "fiscal" has has to do with money, while "fiduciary" has to do with acting on someone's behalf or in their trust.
The reason someone has a "fiduciary" responsibility to shareholders is that the company officers work in the interests of the shareholders. At least they are supposed to -- that's the point of fiduciary responsibility -- it's so important to actually be responsible to people that there should be trustworthy people involved.
Other examples of fiduciary arrangements include trust funds managers, real estate brokers, attorneys at law, executors of estates, auto mechanics, computer security analysts, and police. These people work in the trust of their clients (or the public) for the supposed benefit of their clients (or the public).
Not too long ago two of IT's top original thinkers and innovators, Jeff Hawkins and Ray Kurzweil, appeared at an MIT emerging tech conference to discuss artificial intelligence. Both see computing mirroring the functions of the human brain. But they disagree on how fast scientists and engineers will develop technologies that exhibit the most complex cerebral traits of humans: self-awareness, emotion, and even a sense of one's own mortality.
Because of technology's exponential growth, Kurzweil sees emotion-laden, self-aware machines being developed by mid-century.
Hawkins' view on technology patterned after the human brain is more limited than Kurzweil's prognostications, saying such artificial beings will take centuries, not decades, to create. The brain is just too complex to replicate that quickly. In this video, Hawkins says robots that run amok, will remain science fiction for a very long time. In a recent magazine interview, Hawkins discusses his theories on building an intelligent machine.
Here's a series of podcasts of Kurzweil's vision of a computer that reasons and shows emotion.
High grade scientists use equation to express ideas, weak minds use ton of words.
If you need help with multivariate (non-)linear algebra, I can strongly recommend trying out books from chemistry and psychology (or more specifically, chemometrics and psychometrics) because these are basically somewhat "dumbed-down" descriptions of the most common algorithms to study large datasets.
I say "dumbed-down" in the nicest possible sense, in that they focus on solving practical problems in industry and laboratory, as opposed to rigorous statistical proofs as to why these algorithms work.
Well, it helped me, at least (many years ago). YMMV. Frits.
To be, or not to be: isn't that quite logical, Slashdot Beta?
I really don't think they deserve any slack, particularly Minsky. There were huge personal issues, they published an outright attack based on Minsky's emotional bias (it certainly wasn't based on the facts, because they weren't in possession of them), they didn't do any testing of the claims they made, and they were wrong. Their results affected research for years in the wrong direction. It was a bone-headed stunt with nothing about it that redeems it. I watched it happen; it seemed vindictive and puerile then, and it doesn't look any better today.
I've fallen off your lawn, and I can't get up.
I haven't had a chance to finish watching the video yet, I'm only a half hour in, but even at that stage it seems to make predictions that match my personal experience. He talks about a model where belief models get fed up the hierarchy, and shows an example of a square constructed out of short dashes, when you look close enough. One level of the hierarchy says those little dashs might be horizontal, or diagonal, or vertical, and lets the next level resolve the issue. (picture an ascii square, with lines formed from \ characters). To me this suggests that congition would work even when low levels details are fuzzed out. Some recent, separate, posts I've read suggests 10x10 (or even 6x6) pixel faces are recognisable. Also, I find some images are more recognisable when I unfocus my eyes a bit. I don't know if he is completely right, but these are 2 examples of a prediction being made and verified.
I have been reading non-disclosure agreements for 40 years. I have been learning to understand not only the meaning of the words but is what is behind it. In my spare time I read presidential proclamations of GWB. The non-disclosure said a lot of the usual stuff about what is their's is their's. I expected that, then it said you can only use the materials for approved uses as stated previously. Then I read that the agreement could be terminated by them on 60 days notice at their will, without breaching any conditions. It stinks. What it is says to me IMHO IANABCL, is that they arent sure what they have, and if you use their stuff and start to see more into its uses than they do, they can stop you in your tracks right there with no recourse. I realize that at that point, you could renegotiate with them for a new contract. I stopped right there. I didn't need to read further. If they are so paranoid about this technology, I fail to see why they are announcing it. Maybe they should be discussing it in peer review journals of the appropriate science. I think they are planting seeds for a future patent trolling activity, should someone else develop something along these lines in the future. Right now its a waste of time.