Slashdot Mirror
Building Brainlike Computers
newtronic clues us to an article in IEEE Spectrum by Jeff Hawkins (founder of Palm Computing), titled Why can't a computer be more like a brain? Hawkins brings us up to date with his latest endeavor, Numenta. He covers progress since his book On Intelligence and gives details on Hierarchical Temporal Memory (HTM), which is a platform for simulating neocortical activity. Programming HTMs is different — you essentially feed them sensory data. Numenta has created a framework and tools, free in a "research release," that allow anyone to build and program HTMs.
Because it would signal the end of civilization...if computers can look like women (porn), feel like women (Realdolls), and think like women (have a brain, at least in some cases), then all procreation would cease and humans would suffer the same fate as the dinosaurs.
Next you'll say that we're incapable of growing ears on rats right?
Under the influence of Post-Cyberpunk Gonzo Journalism
...comps get lazy and start reading /. instead of working?
This quote from the article is telling: HTM is not a model of a full brain or even the entire neo-cortex. Our system doesn't have desires, motives, or intentions of any kind. Indeed, we do not even want to make machines that are humanlike. Rather, we want to exploit a mechanism that we believe to underlie much of human thought and perception. This operating principle can be applied to many problems of pattern recognition, pattern discovery, prediction and, ultimately, robotics. But striving to build machines that pass the Turing Test is not our mission. Well, my goal is to build machines that pass the Turing Test, so I have to think about more than cortex. But more generally, one might wonder how much of intelligence it is possible to capture with a system that "doesn't have desires, motives, or intentions of any kind".
you must be lost. this is a science website.
First, AI ignored the brain. Then, Neural Networks took off in the 80's, and during the 90's were also the 'hot thing' in AI and machine learning. Basically, by using some 'brain-like' considerations, flexible learning systems could be built. These include perceptrons, etc. However, since then, neural networks have basically been made obsolete. Both from a theoretical and a practical standpoint, methods like support vector machines and boosting are far better than neural networks; these are the current state of the art. And they return us to the 'old AI' approach of ignoring the brain, in that they are NOT 'brain-like' in any significant way. Rather, they are natural algorithms that arise once you have a mature theory of machine learning (which, one might argue, science now has, with VC theory and later developments).
I tried to read the Numenta stuff, but really I fail to see the 'point' in it. Basically all I want is to see that their methods outperform support vector machines - show me that, and I will be an instant convert. Until then, I remain skeptical.
Since they (scientists) don't really have a full understanding about how the brain works then it seems to me that building a computer to work like one is a litle far fetched.
When Fascism comes to America, it will call itself Anti-Fascism, and tell you to give up your guns.
Ofcourse we don't grow ears on rats. We grow them on mice!
http://news.bbc.co.uk/2/hi/health/1949073.stm
Because of the neocortex's uniform structure, neuro-scientists have long suspected that all its parts work on a common algorithm-that is, that the brain hears, sees, understands language, and even plays chess with a single, flexible tool. Much experimental evidence supports the idea that the neocortex is such a general-purpose learning machine. What it learns and what it can do are determined by the size of the neocortical sheet, what senses the sheet is connected to, and what experiences it is trained on. HTM is a theory of the neocortical algorithm.
While I believe that the HTM is indeed a giant leap in AI (although I disagree with Numenta's Bayesian approach), I cannot help thinking that Hawkins is only addressing a small subset of intelligence. The neocortex is essentially a recognition machine but there is a lot more to brain and behavior than recognition. What is Hawkins' take on things like behavior selection, short and long-term memory, motor sequencing, motor coordination, attention, motivation, etc...?
How does that follow? Granting, for the sake of discussion, that everything in the natural universe, including brains, was created by God, that hardly implies that we can't copy brains. We can reproduce many naturally occurring things, after all, through understanding their structure and composition.
Diamonds are things made by God, and we can copy them.
The question isn't "will we?", the question in reality be: "should we?" Do we have the right to dissect the creations of god and dupllicate them? Sure, I see no reason not to. There are certainly hazards (as most of famous sci-fi movies absolutely love to point out) but there are hazards to driving in the morning. Sure, one day we may be responsible for annihilation of all man-kind but hey, we had a good run ;)
But I think there are some good aspects to trying to replicate the brain. The best reason of all is for understanding of how we work. To duplicate something, you need to know how it works first (or at least know how in general). If we understand the brain, that could help us
Oh and one last thing. Have you ever programmed an email program? They made be fun to design, if you're a hard-core coder but they're not easy.
Please take your professional/scientific reviews to real scientific journals. Only bitter/ignorant jokes are acceptable on /.
Still based on birds though.
early jumpbo jets used the landings of pigeons as a basis for example - those techniques are still used
$_="Slashdotter";$syn="OTT";s;..;;;sub _{print shift||$_};s!ash!Perl !;s=$syn=ack=i;tr+LLEd+BLAH+;_"Just Another ";_
Some people are absolutely terrified by the fact that they are not special at all in the grand scheme of things.
You did a very good job discrediting yourself with that last paragraph.
Gamingmuseum.com: Give your 3D accelerator a rest.
Medievals didn't understand the atom or crystalline structures, but they still made carbonized steel for armour. They had the wrong ideas about exactly how metal became properly carbonized and tempered, but they still came up with correctly tempered spring-like steels (IIRC similar to tempered 1050) without getting any of the "why" of it right.
I think someday we will be viewed as the medievals of AI. We occasionally make progress even though we really don't know what we're doing. Yet.
Weaselmancer
rediculous.
Hawkins and the people he's working with have come up with an approach that lets people explore possible uses of allowing a machine to learn in a way that's inspired by a process that may be part of how humans learn. They don't need a "full understanding" of how the human brain works to do that.
Ok, according to moore's law we will get there, with a transistor based computer. I believe the idea is to create the hardware equivelant of a neuron. Something like Asimov's positronic brain. Currently the modern computer is little more than a highly programmable calculator. The idea in this case is to create a computer that can learn or repurpose it's transistors/neurons.
My colleagues and I have been pursuing that approach for several years. We've focused on the brain's neocortex, and we have made significant progress in understanding how it works. We call our theory, for reasons that I will explain shortly, Hierarchical Temporal Memory, or HTM. We have created a software platform that allows anyone to build HTMs for experimentation and deployment. You don't program an HTM as you would a computer; rather you configure it with software tools, then train it by exposing it to sensory data.The end goal is to create more advanced computers or software. You'd do better venting your religious frustrations against scientists in the genetics industry where the end goal is more advance people or thoughts.
Under the influence of Post-Cyberpunk Gonzo Journalism
I know that you're merely trolling and don't actually believe what you say. Nevertheless ...
:-)
:-)
It's worth stating that unless you believe that the human brain contains magic (which 99% of your religious bretheren don't), then it is no more than a very complex arrangement of perfectly ordinary physical components, namely atoms and molecules. And if you don't think that we will in due course be able to arrange atoms and molecules as we wish, then you're very blinkered to the direction in which science and engineering are heading.
That said, the recreation of human brains is merely an interesting challange as far as practical engineers are concerned, and not a practical approach. The vast majority of us have no intention of actually taking that route because protein is such an inferior building material. Your alleged god (aka. blind evolution) only "chose" it because carbon is so damn versatile in conjunction with O and N and H, so a million different reactions occurred in the mess of the primordial soup. And one of them happened to work.
Well we don't rely on blind chance, but coerce the reactions in the direction we want, which gives us the chance to choose our materials more strategically. And we will.
There's not a chance in hell (trying to use your frame of reference here) of us producing "brains" that are *MERELY* as good as nature created in humans, because the equations that underpin ordinary physics and chemistry (and therefore molecular nanotechnology) say otherwise. Instead, you can expect "brains" a billion times our mental capacity and a trillion times our mental speed in due course. We know that it's possible (from theory, and by observing protein nanomachines doing it very poorly), but we lack the infrastructure to do it ourselves at present. It's many decades away, but hey, we're working on it.
You'd have to contradict the maths and physics of materials and biotech that says that MNT is possible before you can validly say that it's not. And with the intellectual depth of your contribution above, my guess is that you won't.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
Diamonds are things made by God, and we can copy them.
Regardless of there being a God, brains, humans, birds, or diamonds, to be honest we don't want to create a brainlike computer.
Human brains can do amazing things, but one thing we like about computers over human brains is that human brains, even the best ones, are simply wrong from time to time, and our goal with "brainlike computers" is not to recreate these mistakes, but rather to overcome them.
With respect to our senses, again, they are amazing, but then again they are fooled much of the time. There are perceptual errors, optical illusions, selective memories (ask 10 eye witnesses and get 10 different accounts), and all of that.
Today, computers are great at being calculators, and for storing and retrieving digital data. They suck at making "decisions". Even seemingly trivial ones like telling the difference between an apple and an orange is difficult for a computer today.
Take a look at much more mature technologies, like flying. For ages, humans tried to make flying machines like birds, and now we have a handful of flying technologies that can fly faster than the speed of sound and can go beyond the earth's atmosphere. But we still can't fly like a bird with flapping wings, and I don't remember a time in my life where I saw a headline saying "Building Birdlike Planes".
It's true I tell you, feller at work's next door neighbour read it in the paper.
I take drugs for bipolar tendency and have had 5 nervous breakdowns, so I have some ideas about how the brain goes wrong, I am afraid that the search for a perfect machine learning device may be a side track compared to explaining the mistakes the brain makes.
I have an engineering degree and a masters specialising in machine learning - but that was 13 years ago, I would be delighted in more pointers of the state of the art
http://www.cnbc.cmu.edu/Resources/disordermodels/ , on bipolar and neural networks, seemed promising at one stage but I had not the time, energy or rights to read the latest papers. [The web page is dated 1996]
Be Free: Free Software Tuition
The thing is, computers can already do lots of things that brains are bad at. Making brainlike software that allows computers to do things brains are good at is something we want to do, because lots of times we'd like our computers to do tasks that involve repetitively doing things brains are bad at mixed with things that brains are good at, while our actual brains are off doing completely unrelated things rather than be interrupted everytime the computer needs someone to do the part brains are good at.
Obviously, it would be good ultimately to make computers that do things that brains are good at even better than brains do them, but since we're far from as good as brains in our computers in many areas, we've got even more distance to cover till we get to better than brains. In the short-term, we're aiming more for "close enough to brains" so that for tasks which are hard for computers but trivial for brains, we can reduce the amount of human involvement needed to get the task done.
That's not entirely true.
Actually just as much evidence contradicts that hypothesis. We have very specific brain areas for generating and processing verbal data (Broca and Wernicke's areas), and a very specific brain area for recognizing faces.
In defence of Hawkins, note that he does not disagree (RTA) that there are specialized regions in the brain. However, this does not imply that the brain uses a different neural mechanism for different regions. It only means that a region that receives audio input will specialize in processing sounds. It all has to do with how the input and the output fibers are connected. The cortex will rewire itself to accomodate any sensory modality. IMO, Hawkins is right in this regard. Even specialized areas of the visual cortex that show a gradation of recognition capabilities can be explained using a hierarchical system heavily dependent on feedback.
Well, since neural networks perform state-of-the-art results on numerous problems (see the works of Yann LeCun, Geoff Hinton or Ronan Collobert for instance), I wouldn't call them obsolete. They're also second in the Netflix prize contest.
People don't use neural networks because they not as easy to train as SVM (given that you're given libSVM or equivalent). However, SVM are basically template matchers, which are good for problems where the number of samples is big compared to the dimensionnality of the problem (which is NOT the case for real world problems), but that's it.
But using SVM just because the optimization is convex, no matter what the quality of the final solution is, just blows my mind. Besides, since we now know how to optimize deep networks (thanks to Toronto's lab and their Deep Belief Networks), I think neural nets will soon gather some interest again.
My 2 neurons.
Browse the ToCs of some recent journals and conference proceedings on ML, RL, EC, NN.
Sheesh, evil *and* a jerk. -- Jade
Hmm, I see that I might have been easily misunderstood. I meant to say that SVMs dominated the field of classification. Obviously ML journals are full of other topics (unsupervised learning, etc.). But the great majority of publications in classification are about SVMs and related tools (boosting, etc.). At least in the journals I read (JML, JMLR, for example).