Slashdot Mirror
Online 'Sand Mouse' Tests Neurobiologists
The Metahacker writes: "
A Princeton professor and his former student have created a 'mouse' (really, a neural net) that recognizes the word 'one' as spoken by a variety of speakers. The interesting part? They're challenging the neurobiology community to discover the mechanism it uses, using only the tools available to analyze live patients - observation and experimentation. You can upload your own sound files to test the mouse, and view experiments other scientists have performed. Cash prizes will be awarded to those who explain the mouse's behavior or can train the same number of neurons to perform a new task. You can read the New York Times article about it (free registration), or go
directly to the site."
"Science is like sex: sometimes something useful comes out, but that is not the reason we are doing it"
Richard Feynman
Men are born ignorant, not stupid; they are made stupid by education. --Bertrand Russell
The name of the link, "partners", seems pretty self-explanatory to me. It's for partners of the NYTimes, which I would be fairly confident Slashdot is not one. I assume this is so other news outlets or "partners" of the NYTimes can link to their articles without the registration info.
:(
Is it a stupid way of doing it? Yeah, since there's no authentication check to see where they are coming from. Does that alone make the theft of their service right? That's where I disagree.
Just because a car is unlocked and has the keys in it doesn't give you the right to drive off with it.
I don't know.. I see both sides of this, and it all comes down to morals, which are in serious decline in this society anyway. They provide a service, and in return for that service, they ask for some information. Not even $, just information. I don't see the big deal.
As a webmaster myself, I would be ticked off if I offered services requiring registration, and people got around it, viewing my content for essentially nothing. But then again, I'd probably be smart enough to have some kind of check in there at the very least.
Reminds me all of the Dilbert comic:
"What do our customers want?"
"High quality products for free."
Such is the way of the Internet.
BilldaCat
They don't ahve a right to your registration information... you can always choose to not read their information.
--
Are you saying John Hopfield should be the authorized examiner of neuroscientists? Absolutely not. Who said anything about athorising anyone? His experiment can be peer reviewed just like any other work. Time spent working on this problem is time not spent working on the brain directly. No - but it might be time spent sharpening the skills required for brain research. Then again - maybe not. That's the nature of research.
--
-- SIGFPE
Second, what I think is cool about this is sort of an open-source idea for research; there is a large community of nerds that while they don't have the stamina for research, are interested in new cool technologies such as this, and would be willing to help if it's easy to do. One thing I'd like to see more of are distributed projects that can use our idle time to put out more cool ideas; I'd love to see the research that created the 'lifeforms' made by a computer put out into a distributed form as to help their research. Here, further studies of their 'mouse' could be done by simply asking for voice samples of various words by people across the globe as to maximum the sample size (they only used 8 in the paper for testing purposes, but the web site seems to be up to about 600). Certainly, there's important issues such as disclosure of invention, but I think projects like these that challenge the community show excellent progress in research.
"Pinky, you've left the lens cap of your mind on again." - P&TB
"I can see my house from here!" - ST:
Does this piss anyone else off?
.. go ahead, mod it down now, that's what happened last time. :\
What makes you think they have a right to my registration info? You know full well that information is not intended for my benefit.
Typical Slashdot.. I shouldn't be surprised, but everytime I see this it just strikes a nerve. If they want me to not read that information, they should be be smart enough not to give links to it. But far be it from anyone at NYTimes to do that..
--
Non-meta-modded "Overrated" mods are killing Slashdot
(Hey Ryan! Here's your proof!)
64 posts in 60 minutes, and all but about 2 are /. in-jokes. They're all very, very clever - but did anyone read the paper and have anything interesting to say about it?
I read it, but I'm not sure I get it. I -think- they're saying that they set out to implement a neural net for voice recognition, came out with something that worked much better than expected.
I think they're saying that the -reason- that it works so much better than expected is a fairly novel reason (meaning not derivative of common neural net principles), and the process of understanding why this novel method works is best understood by treating the whole problem from a biology, rathern than compsci perspective. e.g. as you would go about trying to figure out how some organism that does something in a novel way does what it does.
It seems to me that they go to great lengths in the paper to be "cute" about using biology terms to describe the behavior of their computer program, because they want to emphasize how organism-like their program is. Do all AI researchers talk this way?
It would be really great if someone who actually understands the paper would post a translation, so everyone can understand what they're really talking about.
If they don't want people using it they should protect it the simple fact is they put it out there where it is pretty easy to find and use. Also as much as we like to think we are big and many in terms of the number of views they get in a day we just plain are not. Therefore it takes nothing away from them and it reduces pain for us. The real question of course is why do they leave it wide open? Simple answer their target market can and will not find it. They know they will have some people using it for whom it was not put there for but from the total lack of security on it you must come to the conclusion that they don't care. So yes this is typical /. for the most part smart people doing smart things in a way that most don't think of and that is why we are chaning the world.
Cypherpunks: Civil Liberty Through Complex Mathematics. Those who live by the sword die by the arrow.
The most amusing possibility is that someone outside the research community may come up with the answer. As this doesn't involve building apparatus, getting a grant, publishing a paper or anything other than thinking, it's very possible an undergrad or a total amatuer will come up with the answer.
Dr. Sejnowski sounded like sour grapes when he called this an "advertising gimmick". Yeah, that's what Fermat must have been doing. Too often scientists confuse the stuff associated with the practice of science - grants, publishing, peer review, experimental proof - with science. Science is what happens in your brain when you're not doing all that other stuff...usually while taking a shower.
"How perfectly Goddamn delightful it all is, to be sure" Charles Crumb
For some time now neuro-biologists (and worse, cognitive psychologists) have been misappropriating and misusing terms and experiments from cognitive computing to justify their often assinine guesses about mental processes. It lets them dress what is essentially bar-room speculation in the clothes of science. Mostly so they can get research grants.
This experiment is calling them out. If they actually get it right then they have some justification in the processes they use. Of course if they fail...
StrutterX
Although physical theories are often hard to test experimentally there is still the standard of internal consistency to a theory. Many fields of inquiry have no such checks. Consider a new 'toy' physical system that a physicist might come up with. They can ask concrete questions like 'what are the energy levels?'. These questions often have definite answers and it's easy to judge if someone's work on this problem is competent. It's much harder to do that in other fields si I find it very interesting to see a problem posed like this. Physicists are very used to working with 'toy' systems. I've a feeling you may be wrong in assuming other sciences are like physics. Try reading some psychology papers...
--
-- SIGFPE
Scientists are hardly limited to single unit recording: in addition to single unit recording they can perform psychophysical experiments, do parallel recording, introspect, examine neuroanatomy, do genetics, perform functional imaging, perform in-situ staining, introduce various drugs, to name just a few. That doesn't make the problem of figuring out how brains work easy, but it certainly makes it a lot more tractable than single unit recording.
No registration required.
A neurologically simple brain for determining whether the number "1" has been achieved? Sounds like a first-poster if I ever heard of one.
My hypothesis: the mouse checks the cid# like the rest of us.
-- Anne Marie
I can see it now...Hopfield and his grad student are going to die in a horrible car accident and scientists are going to spend the next three hundred years trying to figure out what he meant. An obscure professor will finally produce the answer in 2391 in a 1300 page paper that uses quantum theory, the psychology of preadolescent children and a statistical analysis performed by a 300 Exohertz computer, but only five people will actually be able to understand it.
The cake is a pie
"Pinky, you've left the lens cap of your mind on again." - P&TB
"I can see my house from here!" - ST:
And now I shall be able to build a voice fingerprint for each and every hacker in the /. community and Dextor's lab will be no more! Ha ha ha! ha ha ha! ha ha ha!
--
Wooden armaments to battle your imaginary foes!
They already have those. In fact, they've had those for over fifteen years. All you need is a MIDI-enabled pickup that's fast enough to process the input. I saw my first one in 1985. I believe you could stick a microphone into a sax and get MIDI data off it. Polytonal MIDI recognition is only a matter of processing speed.
That's all much simpler than recognizing the word "one" spoken by different voices.
Are you saying John Hopfield should be the authorized examiner of neuroscientists? The challenge is taken as a stunt and an insult by some. Who are these two to tell everyone else how to work? I can see their point.
Secondly, will solving Hopfield's network give us any insight into the brain? He is a leader, but this problem may not be so relevant. Perhaps it won't help. Time spent working on this problem is time not spent working on the brain directly.
The peer review process is set up to make sure that the reviewers are anonymous, and un-affiliated with the authors.
...
:-) A dollup of cynicism is always helpful, here as in so many other areas where humans err. Yes, even in hard science.
Unfortunately, there's a little more to it than that. If you return to the publisher a negative review of a paper written by a respected figure in your scientific community, there is an element of "black mark" against your name in some quarters as a result of the conflict of interest that the publisher has through needing the famous name to appear in his or her journal rather than in a competing one. As a reviewer you're anonymous to the author, but not to the publisher!
And I'm not even going to mention what happens when the journal's editorial board includes researchers interested in the same paradigm or method employed by the famous person, so that publication of that paper validates their own research area
Peer review is a fairly good process on the whole, but I doubt that anyone who's been involved in it [I have] would suggest it approaches perfection.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
What is terrific about this research is not simply that it can recognize a word independently of the speaker, rate of utterance, and noise characteristics. The 'design patterns' used in this system can provide a building block for systems which recognize patterns in many different areas. Generalizing a system like this from recognizing one word to recognizing many seems to be simply a matter of determining which 'word-neuron' is the most excited. Recognizing phrases may be accomplished by using a sequence of 'word-neuron' firings as input and determining which 'phrase-neuron' is the most excited. Presumably, the system works by using some sort of time-delay neuron chain which ensures that all the monosyllables of the word 'one', when spoken in the correct sequence, generate signals which arrive at the 'aggregator' neuron at the same time, thus pushing it over the threshold. (I won't attempt to try to figure out how the -learning- system works right now, which is the really deep part of this :).
I would really like to know which principles of neuron activity play a factor in building such a system- there is a relatively fixed physical model which is simulated in a piece of software; as well as the the interconnection patterns which realize the learning and pattern recognition model. This knowlege could form a great foundation for a general purpose piece of software that could be applied to many areas of pattern analysis.
Mike
Anyone can see why that attitude is potentially offensive.
And suggesting that the earth went around the sun was extremely offensive to the bulk of the scientific community of the day.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra
Maybe they could train it to recognise the phrase 'No, hold the anchovies'. If they could, it would be substantially more inteligent than the goods answering the phone at my local pizza delivery place.
Michael
...another comment from Michael Tandy.
"Goodness me, how unlike the FBI to abuse the trust of the American public." -- The Onion
Our second and more important motivation for presenting the material in two parts is
to open a discussion on the role of deductive thinking in neurobiology. As we have described in
the introduction, we firmly believe that careful and rigorous deductive analysis based on
incomplete knowledge may still lead to novel conclusions and clearly indicate what the most
incisive next experiments are. Nevertheless, incomplete data all too often discourages deep
deductive thinking in neurobiology.
The whole point of doing science is to think. If, as they claim, there's too much data-collection and not enough synthesis, then this is a fun way to get people going.
Good science starts in the lab, but it reaches it's zenith in the shower (or bathtub, if you're Archimedes). Time to pull out Popper and Kuhn and think about how and why science is done.
We've seen things like this before. One of the major problem with neural nets is their tendancy to specialize.. Building a system to recognize one word doesn't remotely compare to a system that can tell one from fun and done while also having the capacity to tell Bob from Rob from Cobb. The experiments posted on the site only show that the system can differentiate between 1 and the other numbers 2-9 and from various nonverbal tones. A neural net will very likely lock on to the specific differences between the sounds of these numbers. Example: A while back someone was creating a neural net to identify tanks on the ground in satellite photos. Two samples were used and the net learned to successfully differentiate them. When other samples were tried, however, the system was completely wrong. Eventually it was determined that the photo with tanks was brighter than the photo without, and that was what the system used to differentiate the photos.
-Splat
I think this experiment could be very important for neurobiological research and maybe other typs of research. There are many fields of science where it is possible to go on forever publishing research without any checks. Obvious areas where this goes on are fields like so-called postmodern literary criticism. But it happens in the sciences too. In behavioural evolutionary biology you can make up just-so stories in paper after paper safe in the knowledge that nobody else can rerun evolution for you and demonstrate that you are wrong. In psychology you can repeatedly perform experiments measuring correlation between this variable and that. By chance one in 20 results are 95% significant and you publish those results as if they are something other than noise. Hopfield's experiment is going to be a sanity check against this kind of work - a kind of experimental control. Here's a situation where somebody does know the answer and work can be checked. A neural net involving only a few hundred neurons. If researchers are unable to reverse engineer this then should they really have jobs supposedly reverse engineering animal or even human brains? We need to see a few more tests like this in academia. Beyond a certain point - after you've taken your last exam - academics are no longer accountable to anyone. Sure - you get peer reviewed. But what happens when you and your peers all belong to a clique that have a vested interest in promulgating a particular scientific dogma? This experiment is a wonderful way to ensure that researchers still are being tested.
--
-- SIGFPE
While the purpose of this project is to make the mouse recognize things, I think the greater significance is shown by the doctor's following comment: "I had to use a way of thinking that felt very different from what I normally use," Dr. Brody said. I think that this is probably the most interesting thing in the whole NYT article. By making this into a contest he is not only bringing more interest into the field but also challenging people to really 'think'! This is how progress really happens. It will definately be interesting to see which different disciplines his methodology will affect, especially if his creation is as novel as it seems.
UBU
There are astronomical data to support the modern understanding of the solar system.
That data wasn't available then.
Telling everybody that he/she is thinking badly and showing him/her a toy is not the same thing as having an answer.
But all the models of science are toys, without exception. That is the whole idea behind science, to produce mathematical toy models that hopefully might approximate the behaviour of reality as determined through empirical tests.
Scientists have never had The Answer, and no competent scientist would ever profess to do so -- the relationship between reality and the models of science are in Science 101, after all. But they're getting damn good at creating models that accurately mimic a lot of reality's behaviours, despite the scientific method not having any ability to determine The Answer or The Truth or whatever.
And that's why it is always good to point out the error of their ways to those "scientists" that waffle on interminably without producing hard testable models based on the hard thinking of logic and mathemetics. It takes more than just mimicking the forms to produce real scientific results. Reality checks may be painful, but they're important.
"The question of whether machines can think is no more interesting than [] whether submarines can swim" - Dijkstra