This could be combined with a model of memory management, allowing you to more clearly map who's leaky.
Build memory wrappers around the whole system so you know who last touched what memory. Lay this memory out in 3d space, colored by who last touched'em.
First, all science fields have to make efforts to connect with normal Joes. If they don't, public support falls, and funding falls. This is an excellent example of such an attempt, more power to them.
Were they hoping to gain fame? Of course they appreciate the prestige of making a public show, they're still human. Why are they less deserving of the fruits of their labor than anyone else revered for their contributions to science, technology, or the arts?
As for the "seeing like a cat" comments. Yes it's an oversimplified statement, but that doesn't mean this work is not important. It's part of an attempt to understand the changes that occur in visual information as it progresses through the brain, which is critical for understanding vision (and thereby make artificial eyes, understand consciousness, yadda yadda).
You can debate the importance of that if you want, but if you accept that that's a valid field of study, this is a necessary stepping stone IMO.
-Illserve
A bit of info about this technology
on
The Cat Cam
·
· Score: 2
First of all, about the cats. They are probably a bit less miserable than you (like to) think. Their heads are possibly restrained to impede movement artifacts on the electrode's recording. They are probably trained to stare at a fixation point for a food reward. I *doubt very much* that their eyelids are clamped open as someone here suggested. The data during eyeblinks is probably *extremely* important for their studies.
And FYI, not all scientists are inhuman monsters.:) Some of us actually try to minimize discomfort as much as the paradigm will allow.
Unfortunately, If they are using standard implant electrode technology, it's not very stable. This means that over time(days-weeks) you will lose access to those cells. Also, they had to laboriously catalog (I assume) the response of each and every cell, and then apply each cell's response as a filter on a small area of space as shown in the figure.
Any inevitable drift in the electrode would screw this up big time, requiring a complete recataloging of spike waveforms and their spatial filters.
So I wouldn't expect this to be field mountable without some major advances in neuron recording technology. The metal wire in the brain is just too invasive in the long term.
As for implanting them in cats and soldiers as mobile spies, why not just use a camera mounted to their head (with maybe a gaze tracker if you want to monitor precisely what they're looking at)? I don't understand what the benefit of this technology is, compared to a small camera. There is an immense "pain-in-the-ass" factor of using it practically.
Not to say it's not important, this is awesome news. I just don't think it has much practical application as a spy cam in the near future. Now creating artificial eyes, that's something, and this is right on target.
I'm sure similar complaints were made about Automobiles in their embryonic stages. The rate of fatal accidents was atrocious for years, until laws and safety features were hammered out. Now they play a vital role in most of our lives, and have for decades.
Perhaps the problems lies in a mismatch between aspects of computer technology. We don't yet know how to build devices that interface cleanly with human sensory and motor capabilities.
We also lack the AI to build sufficient filters for the various input channels that we bombard ourselves with. Imagine a big time CEO without a secretary, all calls can go straight to his desk. That's an apt anology to most of us these days with our cell phones and email accounts. Sure there are the basics of filters, but they have a long way to go.
Give geeks a chance. We'll make the world better, probably.
You've got some fairly strong genralization going on in this reply, much of which I doubt you can substantiate given our current body of knowledge.
Learning occurs over a long period of time, and it is entirely possible that new cells are responsible for the way in which information representations are codified over days or weeks.
To further my previous point, synapses and even spines have been found to sprout on CA pyramidal dendrites in response to learning.
In fact Kristin Harris here at BU thinks that *perhaps* a morphological change(spine growth) is one of the early links in the LTP chain, which is later subsumed by "permanent" biochemical alterations. Counter-intutive? absolutely, but it's hardly the first time the brain has caused a double-take.
In addition, dendrites might also seek (by secreting axonal growth factors) out new connections when its spontaneous firing rate is too low, thereby normalizing the cell's activity with respect to the network.
While a familiar myth in the science fiction community, don't expect sentience to start popping up anytime you throw a bunch of computing power together, regardless of the quantity.
Our own intelligence is the product of carefully deisgned feedback loops to allow us to use our resources more effectively. If you want a more accurate depiction of AI(IMNSHO), check out Greg Bear's book Slant. In it, they have to be coddled and tweaked constantly and are painstakingly constructed over years by teams of scientists.
Newe insights suggest that synapses and even cells are sometimes created in addition to biochemical learning(in rat hippocampus).
Every theory you've heard about learning in the brain is probably true somewhere in there.
Computer nerds should stay away from Neuroscience
on
Leech Neuron Computers
·
· Score: 1
I was hoping for some slightly better commentary from this lot, but was a bit disappointed by the rash of hysterics that broke out (apologies to to the many of you who did not share them).
Point 1: Nothing about Biological computational devices is inherently superior. In point of fact, neurons are atrociously slow compared to silicon. Evolution did the best with what it had, but apparently doped silicon was not a tool it had discovered when it needed to add command and control structures to its little beasties. So it built the fundamental structure of its computers with cells, and has been locked into that track ever since(evolution is always tightly constrained by its own history). We are better than computers at general intelligence because we are unfathomably complex(so far).
#2: Intelligence is no accident. Put a couple (or trillion) neurons in a dish and you do not get a spontaneous potential for a super intelligent entity. Building an intelligence is *HARD*, it does not happen by accident. It is not an emergent property of billions of neurons interacting. Billions of neurons thrown into a dish and kept alive will give you a machine capable of generating epileptic-like seizures, and commiting mass suicide from excitoxicity.
So take it easy folks. Terrorist controlled viral or nuclear weapons will kill us all long before we're capable of building super intlligent entities that will declare us obsolete ala terminator.
Slashdot Mirror
This could be combined with a model of memory management, allowing you to more clearly map who's leaky.
Build memory wrappers around the whole system so you know who last touched what memory. Lay this memory out in 3d space, colored by who last touched'em.
I'm giddy.
First, all science fields have to make efforts to connect with normal Joes. If they don't, public support falls, and funding falls. This is an excellent example of such an attempt, more power to them.
Were they hoping to gain fame? Of course they appreciate the prestige of making a public show, they're still human. Why are they less deserving of the fruits of their labor than anyone else revered for their contributions to science, technology, or the arts?
As for the "seeing like a cat" comments. Yes it's an oversimplified statement, but that doesn't mean this work is not important. It's part of an attempt to understand the changes that occur in visual information as it progresses through the brain, which is critical for understanding vision (and thereby make artificial eyes, understand consciousness, yadda yadda).
You can debate the importance of that if you want, but if you accept that that's a valid field of study, this is a necessary stepping stone IMO.
-Illserve
First of all, about the cats. They are probably a bit less miserable than you (like to) think. Their heads are possibly restrained to impede movement artifacts on the electrode's recording. They are probably trained to stare at a fixation point for a food reward. I *doubt very much* that their eyelids are clamped open as someone here suggested. The data during eyeblinks is probably *extremely* important for their studies.
:) Some of us actually try to minimize discomfort as much as the paradigm will allow.
And FYI, not all scientists are inhuman monsters.
Unfortunately, If they are using standard implant electrode technology, it's not very stable. This means that over time(days-weeks) you will lose access to those cells. Also, they had to laboriously catalog (I assume) the response of each and every cell, and then apply each cell's response as a filter on a small area of space as shown in the figure.
Any inevitable drift in the electrode would screw this up big time, requiring a complete recataloging of spike waveforms and their spatial filters.
So I wouldn't expect this to be field mountable without some major advances in neuron recording technology. The metal wire in the brain is just too invasive in the long term.
As for implanting them in cats and soldiers as mobile spies, why not just use a camera mounted to their head (with maybe a gaze tracker if you want to monitor precisely what they're looking at)? I don't understand what the benefit of this technology is, compared to a small camera. There is an immense "pain-in-the-ass" factor of using it practically.
Not to say it's not important, this is awesome news. I just don't think it has much practical application as a spy cam in the near future. Now creating artificial eyes, that's something, and this is right on target.
Go humans.
-Illserve
I'm sure similar complaints were made about Automobiles in their embryonic stages. The rate of fatal accidents was atrocious for years, until laws and safety features were hammered out. Now they play a vital role in most of our lives, and have for decades.
Perhaps the problems lies in a mismatch between aspects of computer technology. We don't yet know how to build devices that interface cleanly
with human sensory and motor capabilities.
We also lack the AI to build sufficient filters for the various input channels that we bombard ourselves with. Imagine a big time CEO without a secretary, all calls can go straight to his desk. That's an apt anology to most of us these days with our cell phones and email accounts. Sure there are the basics of filters, but they have a long way to go.
Give geeks a chance. We'll make the world better, probably.
-Illserve
You've got some fairly strong genralization going on in this reply, much of which I doubt you can substantiate given our current body of knowledge.
Learning occurs over a long period of time, and it is entirely possible that new cells are responsible for the way in which information representations are codified over days or weeks.
To further my previous point, synapses and even spines have been found to sprout on CA pyramidal dendrites in response to learning.
In fact Kristin Harris here at BU thinks that *perhaps* a morphological change(spine growth) is one of the early links in the LTP chain, which is later subsumed by "permanent" biochemical alterations. Counter-intutive? absolutely, but it's hardly the first time the brain has caused a double-take.
In addition, dendrites might also seek (by secreting axonal growth factors) out new connections when its spontaneous firing rate is too low, thereby normalizing the cell's activity with respect to the network.
-Illserve
While a familiar myth in the science fiction community, don't expect sentience to start popping up anytime you throw a bunch of computing power together, regardless of the quantity.
Our own intelligence is the product of carefully deisgned feedback loops to allow us to use our resources more effectively. If you want a more accurate depiction of AI(IMNSHO), check out Greg Bear's book Slant. In it, they have to be coddled and tweaked constantly and are painstakingly constructed over years by teams of scientists.
Newe insights suggest that synapses and even cells are sometimes created in addition to biochemical learning(in rat hippocampus).
Every theory you've heard about learning in the brain is probably true somewhere in there.
I was hoping for some slightly better commentary from this lot, but was a bit disappointed by the rash of hysterics that broke out (apologies to to the many of you who did not share them).
Point 1:
Nothing about Biological computational devices is inherently superior. In point of fact, neurons are atrociously slow compared to silicon. Evolution did the best with what it had, but apparently doped silicon was not a tool it had discovered when it needed to add command and control structures to its little beasties. So it built the fundamental structure of its computers with cells, and has been locked into that track ever since(evolution is always tightly constrained by its own history). We are better than computers at general intelligence because we are unfathomably complex(so far).
#2: Intelligence is no accident. Put a couple (or trillion) neurons in a dish and you do not get a spontaneous potential for a super intelligent entity. Building an intelligence is *HARD*, it does not happen by accident. It is not an emergent property of billions of neurons interacting. Billions of neurons thrown into a dish and kept alive will give you a machine capable of generating epileptic-like seizures, and commiting mass suicide from excitoxicity.
So take it easy folks. Terrorist controlled viral or nuclear weapons will kill us all long before we're capable of building super intlligent entities that will declare us obsolete ala terminator.
-Illserve