[agi] Lamarck Lives!(?)
Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Note also, http://sciencelinks.jp/j-east/article/200308/20030803A0129895.php Jean Baptiste de Lamarck (1744-1829) maintained that characteristics that were acquired during an organism's lifetime are passed on to its offspring. This theory, known as Lamarckian inheritance, was later completely discredited. However, recent progress in epigenetics research suggests it needs to be reexamined in consideration of DNA methylation. In this article, I summarize our observations, which support Lamarckian inheritance. Initial experiments indicate that (1) artificially induced demethylation of rice genomic DNA results in heritable dwarfism, and (2) cold stress induces extensive demethylation in somatic cells of the maize root. Based on these results, I propose the hypothesis that traits that are acquired during plant growth are sometimes inherited by their progeny through persistent alteration of the DNA methylation status. (author abst.) I wonder how this relates to adaptive mutagenesis http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1206667 which has been rather controversial http://www.genetics.org/cgi/content/full/165/4/2319 ben On Wed, Dec 3, 2008 at 11:11 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] I intend to live forever, or die trying. -- Groucho Marx --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
On 12/3/2008 8:11 AM, Richard Loosemore wrote: Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Yes, but it obviously needs a lot more confirmation first. :-) Would it mean that memories (including cultural adaptations) could be passed from mother to child? No. As far as I understand it, they are proposing changes to the DNA in the neural cells only, so it wouldn't be passed on. And I would expect that the changes are specific to the neural structure of the subject, so even if you moved the changes to DNA in another subject, it wouldn't work. Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? No bets here. But they are proposing that elements are added onto the DNA, not that changes are made in arbitrary locations within the DNA, so it's not /quite/ as bad as you suggest --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
Richard, The role played by the epigenome in genetics actually does have a slightly Lamarckian tinge. Nova had a show saying that when identical twins are born their epigenomes are very similar, but that as they age their epigenomes start to differ more an more, and that certain behaviors like drinking or smoking can increase the rate at which such changes take place. What I didn't understand about the article you linked to is that it appears they are changing the epigenome to change the expression of DNA, but as far as I know DNA only appears in the nucleus (with the exception of mitochondirial DNA), and thus would appear to affect the cell as a whole, and thus not be good at differentially affecting the strengths of different synapses --- as would presumably be required for most neuronal memory --- unless the nuclear DNA had some sort of mapping to individual synapses, or unless local changes to mitochondrial DNA, near a synapse are involved. The article does not appear to shed in any light on this issue of how changes in the expression of DNA would affect learning at the synapse level, where most people think it occurs. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 11:12 AM To: agi@v2.listbox.com Subject: [agi] Lamarck Lives!(?) Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on -your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Interesting. Note, however, that it is conceivable that those other examples of plant and bacterial adaptation could be explained as situation-specific - in the sense that the particular cause of the adaptation could have worked in ways that were not generalizable to other, similar factors. So, some very specific factors could be inherited while others could never have an effect because they just don't happen to affect methylation. But if the neural results hold up, this would be a whole new ball game: a completely general mechanism for storing memories in an inheritable form. Not just [memory-for-your-first-kiss] affecting the DNA, but the whole shebang. If it turns out that this is the correct interpretation, then this is one hell of a historic moment. I must say, I am still a little skeptical, but we'll see how it plays out. Richard Loosemore Ben Goertzel wrote: Note also, http://sciencelinks.jp/j-east/article/200308/20030803A0129895.php Jean Baptiste de Lamarck (1744-1829) maintained that characteristics that were acquired during an organism's lifetime are passed on to its offspring. This theory, known as Lamarckian inheritance, was later completely discredited. However, recent progress in epigenetics research suggests it needs to be reexamined in consideration of DNA methylation. In this article, I summarize our observations, which support Lamarckian inheritance. Initial experiments indicate that (1) artificially induced demethylation of rice genomic DNA results in heritable dwarfism, and (2) cold stress induces extensive demethylation in somatic cells of the maize root. Based on these results, I propose the hypothesis that traits that are acquired during plant growth are sometimes inherited by their progeny through persistent alteration of the DNA methylation status. (author abst.) I wonder how this relates to adaptive mutagenesis http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=1206667 which has been rather controversial http://www.genetics.org/cgi/content/full/165/4/2319 ben On Wed, Dec 3, 2008 at 11:11 AM, Richard Loosemore [EMAIL PROTECTED] wrote: Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Harry Chesley wrote: On 12/3/2008 8:11 AM, Richard Loosemore wrote: Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Yes, but it obviously needs a lot more confirmation first. :-) Would it mean that memories (including cultural adaptations) could be passed from mother to child? No. As far as I understand it, they are proposing changes to the DNA in the neural cells only, so it wouldn't be passed on. And I would expect that the changes are specific to the neural structure of the subject, so even if you moved the changes to DNA in another subject, it wouldn't work. You're right, of course. But if this holds up, it would not be quite so crazy to imagine a mechanism that uses junk DNA signalling to get the end caps of the genital DNA to reflect the changes. I admit, though, this is stretching it a bit ;-). As for the changes not working in another subject: yes, it would probably be the case that specific memories are encoded in an individual-specific way. But what about more general factors? What if there were some primitive types of musical understanding, say, that were common across individuals, for example? Like, a set of very primitive concepts having to do with links between sounds and finger movements? If such general factors could be passed across, a person could inherit above average musical ability because their parents had been active musicians all their lives. All this is fun to think about, but I confess I am mostly playing devil's advocate here. Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? No bets here. But they are proposing that elements are added onto the DNA, not that changes are made in arbitrary locations within the DNA, so it's not /quite/ as bad as you suggest It would be pretty embarrassing for people gearing up for scans with a limiting resolution at about the size of one neuron, though. IIRC that was the rough order of magnitude assumed in the proposal I reviewed here recently. Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? No bets here. But they are proposing that elements are added onto the DNA, not that changes are made in arbitrary locations within the DNA, so it's not /quite/ as bad as you suggest It would be pretty embarrassing for people gearing up for scans with a limiting resolution at about the size of one neuron, though. IIRC that was the rough order of magnitude assumed in the proposal I reviewed here recently. When I saw Todd Huffman give a presentation on brain imaging aimed toward WBE last year, he was showing images revealing individual proteins expressed around in the brain ... and the challenge was to infer higher-level stuff like synaptic potentiation from this lower-level protein-expression imaging data My recollection of the details is fuzzy, but anyway I'm clear that he and others in that field are working on lower-level imaging as well as neuron-level... ben g --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
2008/12/3 Richard Loosemore [EMAIL PROTECTED]: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? No. They are saying memories might be stored as changes *on* the DNA. Imagine a big long DNA molecule. It has little molecules attached to bits of it, which regulate which genes are and aren't expressed. That's how a cell knows it's a skin cell, or an eye cell or a liver cell. Apparently the same mechanism is used in neurons are part of the mechanism for laying down new memories. Would it mean that memories (including cultural adaptations) could be passed from mother to child? No, for two reasons: (1) the DNA isn't being changed. (2) even if the DNA was being changed, it isn't in the germ-line. (Incidently, my understanding is[*] that DNA in various cells in the mammalian immune system does change as the immune system evolves to cope with infectious agents; but these changes aren't passed along to the next generation.) * if there are any molecular biologists reading, feel free to correct me. -- Philip Hunt, [EMAIL PROTECTED] Please avoid sending me Word or PowerPoint attachments. See http://www.gnu.org/philosophy/no-word-attachments.html --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Hi Richard, Thanks for the link, pretty intriguing. It's important to note that the mechanism proposed is just a switch that turns specific genes off... so properly understood, it's likely that the resolution required to model this mechanism would not necessarily require modeling the entire DNA strand. It seems more likely that these methylation caps are being applied to very specific genes that produce proteins heavily implicated in the dynamics of synapse creation/destruction (or some other process related to memory). So modeling the phenomenon could very possibly be done functionally. Memories could only be passed to the child if 1) those DNA changes were also made in the germ cells (i.e. egg/sperm) and 2) the DNA changes involved resulted in a brain organization in the child that mimicked the parent's brain. (1) is very unlikely but theoretically possible; (2) is impossible for two reasons. One is, the methylation patterns proposed involve a large number of neurons, converging on a pattern of methylation; in contrast, a germ cell would only capture the methylation of a single cell (which would then be cloned in the developing fetus). Second, the hypothesized methylation patterns represent a different medium of information storage in the mature brain than what is normally considered to be the role of DNA in the developing brain. It would truly be a huge leap to suggest that the information stored via this alteration of DNA would result in that information being preserved somehow in a developing brain. There are plenty of other epigenetic phenomena to get Lamarck fans excited, but this isn't one of them. Terren --- On Wed, 12/3/08, Richard Loosemore [EMAIL PROTECTED] wrote: From: Richard Loosemore [EMAIL PROTECTED] Subject: [agi] Lamarck Lives!(?) To: agi@v2.listbox.com Date: Wednesday, December 3, 2008, 11:11 AM Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
2008/12/3 Richard Loosemore [EMAIL PROTECTED]: Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? No bets here. But they are proposing that elements are added onto the DNA, not that changes are made in arbitrary locations within the DNA, so it's not /quite/ as bad as you suggest It would be pretty embarrassing for people gearing up for scans with a limiting resolution at about the size of one neuron, though. IIRC that was the rough order of magnitude assumed in the proposal I reviewed here recently. It might well be. It is anyway apparent that there are different mechanisms in the brain for laying down long-term memories and for short-term thinking over the order of a few seconds. -- Philip Hunt, [EMAIL PROTECTED] Please avoid sending me Word or PowerPoint attachments. See http://www.gnu.org/philosophy/no-word-attachments.html --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
Ed, That's a good point about synapses, but perhaps the methylation just affects the neuron's output, e.g., the targeted genes express proteins that only find a functional role in the axon. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Richard, The role played by the epigenome in genetics actually does have a slightly Lamarckian tinge. Nova had a show saying that when identical twins are born their epigenomes are very similar, but that as they age their epigenomes start to differ more an more, and that certain behaviors like drinking or smoking can increase the rate at which such changes take place. What I didn't understand about the article you linked to is that it appears they are changing the epigenome to change the expression of DNA, but as far as I know DNA only appears in the nucleus (with the exception of mitochondirial DNA), and thus would appear to affect the cell as a whole, and thus not be good at differentially affecting the strengths of different synapses --- as would presumably be required for most neuronal memory --- unless the nuclear DNA had some sort of mapping to individual synapses, or unless local changes to mitochondrial DNA, near a synapse are involved. The article does not appear to shed in any light on this issue of how changes in the expression of DNA would affect learning at the synapse level, where most people think it occurs. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 11:12 AM To: agi@v2.listbox.com Subject: [agi] Lamarck Lives!(?) Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on -your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Philip Hunt wrote: 2008/12/3 Richard Loosemore [EMAIL PROTECTED]: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? No. They are saying memories might be stored as changes *on* the DNA. Imagine a big long DNA molecule. It has little molecules attached to bits of it, which regulate which genes are and aren't expressed. That's how a cell knows it's a skin cell, or an eye cell or a liver cell. Apparently the same mechanism is used in neurons are part of the mechanism for laying down new memories. Yes, I know this: I appreciate the difference between tampering with the gene regulation apparatus and affecting the codons themselves, but for my money, *any* mechanism that collects synaptic signals (to speak very broadly) and then walks over to some DNA and does anything systematic to the DNA, to record the results of those signals, is storing something on the DNA. There could have been no way to get from one to the other, but now it appears that there is. Would it mean that memories (including cultural adaptations) could be passed from mother to child? No, for two reasons: (1) the DNA isn't being changed. (2) even if the DNA was being changed, it isn't in the germ-line. This is a crucial point: has anyone definitely ruled out the possibility that state of the gene regulation apparatus could somehow affect the germ line? This I am not clear about. When the Mom and Pop DNA really start to get down and boogie together, do they throw away the scratchpad that contains all the extra information about the state of the junk DNA, the methylation endcaps, etc? Or is it still an open question whether some of that can carry over? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Terren Suydam wrote: Hi Richard, Thanks for the link, pretty intriguing. It's important to note that the mechanism proposed is just a switch that turns specific genes off... so properly understood, it's likely that the resolution required to model this mechanism would not necessarily require modeling the entire DNA strand. It seems more likely that these methylation caps are being applied to very specific genes that produce proteins heavily implicated in the dynamics of synapse creation/destruction (or some other process related to memory). So modeling the phenomenon could very possibly be done functionally. Memories could only be passed to the child if 1) those DNA changes were also made in the germ cells (i.e. egg/sperm) and 2) the DNA changes involved resulted in a brain organization in the child that mimicked the parent's brain. (1) is very unlikely but theoretically possible; (2) is impossible for two reasons. One is, the methylation patterns proposed involve a large number of neurons, converging on a pattern of methylation; in contrast, a germ cell would only capture the methylation of a single cell (which would then be cloned in the developing fetus). Second, the hypothesized methylation patterns represent a different medium of information storage in the mature brain than what is normally considered to be the role of DNA in the developing brain. It would truly be a huge leap to suggest that the information stored via this alteration of DNA would result in that information being preserved somehow in a developing brain. There are plenty of other epigenetic phenomena to get Lamarck fans excited, but this isn't one of them. I see what you are saying. I really want to distance myself from this a little bit (don't want to seem like I am really holding the banner for Lamarck's crowd), but I think the main conclusion that we can draw from this piece of research is, as I said a moment ago, that we now have reason to believe that there is *some* mechanism that connects memories to DNA modifications, whereas if anyone had suggested such a link a few years ago they would have been speculating on thin ice. I definitely agree that getting from there to a situation in which packages of information are being inserted into germ cell DNA is a long road, but this one new piece of research has - surprisingly - just cut the length of that road in half. All fun and interesting, but now back to the real AGI Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
I don' really see how a change in gene expression in the nucleus of a neuron caused by methylation could store long term memories, since most neural network models store all most all their information in the location and differentiation of they synapses. How is information in a neural net stored by making what would appear to be only neuron-wide behaviors? Such a global change might be valuable for signally that a record of recent events in the neuron at a give brief period of time, should be stored, but it would not appear to actually keep them stored over a long period of time. I think the article failed to mention an important part of the theory of what is going on. Ed Porter -Original Message- From: Terren Suydam [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 12:16 PM To: agi@v2.listbox.com Subject: RE: [agi] Lamarck Lives!(?) Ed, That's a good point about synapses, but perhaps the methylation just affects the neuron's output, e.g., the targeted genes express proteins that only find a functional role in the axon. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Richard, The role played by the epigenome in genetics actually does have a slightly Lamarckian tinge. Nova had a show saying that when identical twins are born their epigenomes are very similar, but that as they age their epigenomes start to differ more an more, and that certain behaviors like drinking or smoking can increase the rate at which such changes take place. What I didn't understand about the article you linked to is that it appears they are changing the epigenome to change the expression of DNA, but as far as I know DNA only appears in the nucleus (with the exception of mitochondirial DNA), and thus would appear to affect the cell as a whole, and thus not be good at differentially affecting the strengths of different synapses --- as would presumably be required for most neuronal memory --- unless the nuclear DNA had some sort of mapping to individual synapses, or unless local changes to mitochondrial DNA, near a synapse are involved. The article does not appear to shed in any light on this issue of how changes in the expression of DNA would affect learning at the synapse level, where most people think it occurs. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 11:12 AM To: agi@v2.listbox.com Subject: [agi] Lamarck Lives!(?) Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on -your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
I definitely agree that getting from there to a situation in which packages of information are being inserted into germ cell DNA is a long road, but this one new piece of research has - surprisingly - just cut the length of that road in half. Half of infinity is still infinity ;-] It's just not a possibility, which should be obvious if you look at the quantity of information involved. Let M be a measure of the information stored via distributed methylation patterns across some number of neurons N. The amount of information stored by a single neuron's methylated DNA is going to be much smaller than M (roughly M/N). A single germ cell which might conceivably inherit the methylation pattern from some single neuron would not be able to convey any more than a [1/N] piece of the total information that makes up M. The real significance of this research has nothing to do with Lamarckian inheritance. It has to do with the proposed medium of memory, as a network of switched genes in neurons and perhaps other cells. It's a novel idea that is generative of a whole range of new hypotheses and applications (e.g. in the pharmaceutical space). Terren --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Terren Suydam wrote: I definitely agree that getting from there to a situation in which packages of information are being inserted into germ cell DNA is a long road, but this one new piece of research has - surprisingly - just cut the length of that road in half. Half of infinity is still infinity ;-] It's just not a possibility, which should be obvious if you look at the quantity of information involved. Let M be a measure of the information stored via distributed methylation patterns across some number of neurons N. The amount of information stored by a single neuron's methylated DNA is going to be much smaller than M (roughly M/N). A single germ cell which might conceivably inherit the methylation pattern from some single neuron would not be able to convey any more than a [1/N] piece of the total information that makes up M. Now you're just trying to make me think ;-). Okay, try this. [heck, you don't have to: I am just playing with ideas here...] The methylation pattern has not necessarily been shown to *only* store information in a distributed pattern of activation - the jury's out on that one (correct me if I'm wrong). Suppose that the methylation end caps are just being used as a way station for some mechanism whose *real* goal is to make modifications to some patterns in the junk DNA. So, here I am suggesting that the junk DNA of any particular neuron is being used to code for large numbers of episodic memories (one memory per DNA strand, say), with each neuron being used as a redundant store of many episodes. The same episode is stored in multiple neurons, but each copy is complete. When we observe changes in the methylation patterns, perhaps these are just part of the transit mechanism, not the final destination for the pattern. To put it in the language that Greg Bear would use, the endcaps were just part of the radio system. (http://www.gregbear.com/books/darwinsradio.cfm) Now suppose that part of the junk sequences that code for these memories are actually using a distributed coding scheme *within* the strand (in the manner of a good old fashioned backprop neural net, shall we say). That would mean that, contrary to what I said in the above paragraph, the individual strands were coding a bunch of different episodic memory traces, not just one. (It is even possible that the old idea of flashbulb memories may survive the critiques that have been launched against it ... and in that case, it could be that what we are talking about here is the mechanism for storing that particular set of memories. And in that case, perhaps the system expects so few of them, that all DNA strands everywhere in the system are dedicated to storing just the individual's store of flashbulb memories). Now, finally, suppose that there is some mechanism for radioing these memories to distribute them around the system ... and that the radio network extends as far as the germ DNA. Now, the offspring could get the mixed flashbulb memories of its parents, in perhaps very dilute or noisy form. This assumes that whatever coding scheme is used to store the information can somehow transcend the coding schemes used by different individuals. Since we do not yet know how much common ground there is between the knowledge storage used by individuals yet, this is still possible. There: I invented a possible mechanism. Does it work? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
Ed, Though it seems obvious that synapses are *involved* with memory storage, it's not proven that synapses individually *store* memories. Clearly memory is distributed, as evidenced by brain injury studies (a situation that led Karl Pribram/David Bohm to propose a holographic storage metaphor). In other words, memories might be stored as patterns of synaptic/neural dynamics, in which the relevant scope is well higher than at the level of the individual synapse. Given that memory storage is not so simple as to depend crucially on individual synapses, I see no serious problems with a neuron-wide mechanism of memory storage. Also, think of Hebbian learning, in which synaptic strength is reinforced based on a neuron-wide signal. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: From: Ed Porter [EMAIL PROTECTED] Subject: RE: [agi] Lamarck Lives!(?) To: agi@v2.listbox.com Date: Wednesday, December 3, 2008, 1:33 PM I don' really see how a change in gene expression in the nucleus of a neuron caused by methylation could store long term memories, since most neural network models store all most all their information in the location and differentiation of they synapses. How is information in a neural net stored by making what would appear to be only neuron-wide behaviors? Such a global change might be valuable for signally that a record of recent events in the neuron at a give brief period of time, should be stored, but it would not appear to actually keep them stored over a long period of time. I think the article failed to mention an important part of the theory of what is going on. Ed Porter -Original Message- From: Terren Suydam [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 12:16 PM To: agi@v2.listbox.com Subject: RE: [agi] Lamarck Lives!(?) Ed, That's a good point about synapses, but perhaps the methylation just affects the neuron's output, e.g., the targeted genes express proteins that only find a functional role in the axon. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Richard, The role played by the epigenome in genetics actually does have a slightly Lamarckian tinge. Nova had a show saying that when identical twins are born their epigenomes are very similar, but that as they age their epigenomes start to differ more an more, and that certain behaviors like drinking or smoking can increase the rate at which such changes take place. What I didn't understand about the article you linked to is that it appears they are changing the epigenome to change the expression of DNA, but as far as I know DNA only appears in the nucleus (with the exception of mitochondirial DNA), and thus would appear to affect the cell as a whole, and thus not be good at differentially affecting the strengths of different synapses --- as would presumably be required for most neuronal memory --- unless the nuclear DNA had some sort of mapping to individual synapses, or unless local changes to mitochondrial DNA, near a synapse are involved. The article does not appear to shed in any light on this issue of how changes in the expression of DNA would affect learning at the synapse level, where most people think it occurs. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 11:12 AM To: agi@v2.listbox.com Subject: [agi] Lamarck Lives!(?) Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on -your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed:
Re: [agi] Lamarck Lives!(?)
Ed, you seem to be taking the memory as synaptic weight modification model a bit too seriously ... it's really just a simplified formal model that captures a certain percentage of what goes on in the brain (and no one knows how much) This is why I shy away from brain-modeling approaches to AGI ... we just don't know how the brain works yet... ben g On Wed, Dec 3, 2008 at 1:33 PM, Ed Porter [EMAIL PROTECTED] wrote: I don' really see how a change in gene expression in the nucleus of a neuron caused by methylation could store long term memories, since most neural network models store all most all their information in the location and differentiation of they synapses. How is information in a neural net stored by making what would appear to be only neuron-wide behaviors? Such a global change might be valuable for signally that a record of recent events in the neuron at a give brief period of time, should be stored, but it would not appear to actually keep them stored over a long period of time. I think the article failed to mention an important part of the theory of what is going on. Ed Porter -Original Message- From: Terren Suydam [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 12:16 PM To: agi@v2.listbox.com Subject: RE: [agi] Lamarck Lives!(?) Ed, That's a good point about synapses, but perhaps the methylation just affects the neuron's output, e.g., the targeted genes express proteins that only find a functional role in the axon. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Richard, The role played by the epigenome in genetics actually does have a slightly Lamarckian tinge. Nova had a show saying that when identical twins are born their epigenomes are very similar, but that as they age their epigenomes start to differ more an more, and that certain behaviors like drinking or smoking can increase the rate at which such changes take place. What I didn't understand about the article you linked to is that it appears they are changing the epigenome to change the expression of DNA, but as far as I know DNA only appears in the nucleus (with the exception of mitochondirial DNA), and thus would appear to affect the cell as a whole, and thus not be good at differentially affecting the strengths of different synapses --- as would presumably be required for most neuronal memory --- unless the nuclear DNA had some sort of mapping to individual synapses, or unless local changes to mitochondrial DNA, near a synapse are involved. The article does not appear to shed in any light on this issue of how changes in the expression of DNA would affect learning at the synapse level, where most people think it occurs. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 11:12 AM To: agi@v2.listbox.com Subject: [agi] Lamarck Lives!(?) Am I right in thinking that what these people: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on -your-dna.html are saying is that memories can be stored as changes in the DNA inside neurons? If so, that would upset a few apple carts. Would it mean that memories (including cultural adaptations) could be passed from mother to child? Implication for neuroscientists proposing to build a WBE (whole brain emulation): the resolution you need may now have to include all the DNA in every neuron. Any bets on when they will have the resolution to do that? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] I intend to live forever, or die trying. -- Groucho Marx --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed:
Re: [agi] Lamarck Lives!(?)
Does it work? Assuming that the encodings between parent and child are compatible, it could work. But you'd still be limited to the total amount of information storage allowable in the junk DNA (which would necessarily be a miniscule fraction of the total information stored in the brain as memory). And you'd still need to identify the mechanism that writes to the junk DNA, which would involve some hefty molecular machinery (snipping DNA, synthesizing the new stuff, rejoining it, all while doing error correction and turning off the error correction involved with normal DNA synthesis/repair). Finally, the idea of junk DNA is getting smaller and smaller as we identify gene targets that are not necessarily proteins, but various RNA products; or sections of DNA that are simply there to anchor other sections, or to enable other methods of gene switching. I know you're just playing here but it would be easy to empirically test this. Does junk DNA change between birth and death? Something tells me we would have discovered something that significant a long time ago. Terren --- On Wed, 12/3/08, Richard Loosemore [EMAIL PROTECTED] wrote: Okay, try this. [heck, you don't have to: I am just playing with ideas here...] The methylation pattern has not necessarily been shown to *only* store information in a distributed pattern of activation - the jury's out on that one (correct me if I'm wrong). Suppose that the methylation end caps are just being used as a way station for some mechanism whose *real* goal is to make modifications to some patterns in the junk DNA. So, here I am suggesting that the junk DNA of any particular neuron is being used to code for large numbers of episodic memories (one memory per DNA strand, say), with each neuron being used as a redundant store of many episodes. The same episode is stored in multiple neurons, but each copy is complete. When we observe changes in the methylation patterns, perhaps these are just part of the transit mechanism, not the final destination for the pattern. To put it in the language that Greg Bear would use, the endcaps were just part of the radio system. (http://www.gregbear.com/books/darwinsradio.cfm) Now suppose that part of the junk sequences that code for these memories are actually using a distributed coding scheme *within* the strand (in the manner of a good old fashioned backprop neural net, shall we say). That would mean that, contrary to what I said in the above paragraph, the individual strands were coding a bunch of different episodic memory traces, not just one. (It is even possible that the old idea of flashbulb memories may survive the critiques that have been launched against it ... and in that case, it could be that what we are talking about here is the mechanism for storing that particular set of memories. And in that case, perhaps the system expects so few of them, that all DNA strands everywhere in the system are dedicated to storing just the individual's store of flashbulb memories). Now, finally, suppose that there is some mechanism for radioing these memories to distribute them around the system ... and that the radio network extends as far as the germ DNA. Now, the offspring could get the mixed flashbulb memories of its parents, in perhaps very dilute or noisy form. This assumes that whatever coding scheme is used to store the information can somehow transcend the coding schemes used by different individuals. Since we do not yet know how much common ground there is between the knowledge storage used by individuals yet, this is still possible. There: I invented a possible mechanism. Does it work? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
I know you're just playing here but it would be easy to empirically test this. Does junk DNA change between birth and death? Something tells me we would have discovered something that significant a long time ago. Terren well, loads of mutations occur in nuclear DNA between birth and death; this is part of how aging occurs. There are specific DNA repair mechanisms that fix mutation errors that occur during the cell's lifetime It seems quite plausible that these repair mechanisms might work differently on coding and noncoding regions of the DNA ben g p.s. hmm.. relatedly, there is debatable evidence that in some cases there can be acquired mutations http://home.planet.nl/~gkorthof/kortho39.htm related to immune function, and some of these may be in genes and some not... --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Ben Goertzel wrote: I know you're just playing here but it would be easy to empirically test this. Does junk DNA change between birth and death? Something tells me we would have discovered something that significant a long time ago. Terren well, loads of mutations occur in nuclear DNA between birth and death; this is part of how aging occurs. There are specific DNA repair mechanisms that fix mutation errors that occur during the cell's lifetime It seems quite plausible that these repair mechanisms might work differently on coding and noncoding regions of the DNA Ah, hang on folks: what I was meaning was that the *state* of the junk DNA was being used, not the code. I am referring to the stuff that is dynamically interacting, as a result of which genes are switched on and off all over the place so this is a gigantic network of switches. I wouldn't suggest that something is snipping and recombining the actual code of the junk DNA, only that the state of the switches is being used to code for something. Question is: can the state of the switches be preserved during reproduction? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
On Wed, Dec 3, 2008 at 3:19 PM, Ed Porter [EMAIL PROTECTED] wrote: Terry and Ben, I never implied anything that could be considered a memory at a conscious level is stored at just one synapse, but all the discussions I have heard of learning in various brain science books and lectures imply synaptic weights are the main place of our memories are stored. Nevertheless, although it's an oft-repeated and well-spread meme, the available biological evidence shows only that **this is one aspect of the biological basis of memory in organisms with complex brains** There certainly is data about long-term potentiation and its relationship to memory ... but the available data comes nowhere near to justifying the sorts of assumptions made in setting up formal neural net models, in which synaptic modification is assumed as the sole basis of learning/memory... ben g --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
Terry and Ben, I never implied anything that could be considered a memory at a conscious level is stored at just one synapse, but all the discussions I have heard of learning in various brain science books and lectures imply synaptic weights are the main place of our memories are stored. Yes, Hebbian learning would appear to use a neuron wide signal that a neuron has fired, but the actual Hebbian learning is only believed to take place at individual synapses as a function of the relationship of the timing between the synapse's up- and downstream neurons. So all the Hebbian and Hebbian-like learning I have ever heard described distinguishes between which of a neuron's synapses are to have their weights changed by how much and/or in what direction. Now there may well be other mechanisms which would allow long term memory to be stored at a neuron-wide level, but I can't at the moment remember reading or hearing of any. That is not proof they don't exist, but it, at least it suggests that, so far, the evidence for such mechanism that has been learned is underwhelming. On the other hand, I have read or heard probably at least a thousand times about the brain storing information in synapses. Ed Porter -Original Message- From: Terren Suydam [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 2:00 PM To: agi@v2.listbox.com Subject: RE: [agi] Lamarck Lives!(?) Ed, Though it seems obvious that synapses are *involved* with memory storage, it's not proven that synapses individually *store* memories. Clearly memory is distributed, as evidenced by brain injury studies (a situation that led Karl Pribram/David Bohm to propose a holographic storage metaphor). In other words, memories might be stored as patterns of synaptic/neural dynamics, in which the relevant scope is well higher than at the level of the individual synapse. Given that memory storage is not so simple as to depend crucially on individual synapses, I see no serious problems with a neuron-wide mechanism of memory storage. Also, think of Hebbian learning, in which synaptic strength is reinforced based on a neuron-wide signal. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: From: Ed Porter [EMAIL PROTECTED] Subject: RE: [agi] Lamarck Lives!(?) To: agi@v2.listbox.com Date: Wednesday, December 3, 2008, 1:33 PM I don' really see how a change in gene expression in the nucleus of a neuron caused by methylation could store long term memories, since most neural network models store all most all their information in the location and differentiation of they synapses. How is information in a neural net stored by making what would appear to be only neuron-wide behaviors? Such a global change might be valuable for signally that a record of recent events in the neuron at a give brief period of time, should be stored, but it would not appear to actually keep them stored over a long period of time. I think the article failed to mention an important part of the theory of what is going on. Ed Porter -Original Message- From: Terren Suydam [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 12:16 PM To: agi@v2.listbox.com Subject: RE: [agi] Lamarck Lives!(?) Ed, That's a good point about synapses, but perhaps the methylation just affects the neuron's output, e.g., the targeted genes express proteins that only find a functional role in the axon. Terren --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Richard, The role played by the epigenome in genetics actually does have a slightly Lamarckian tinge. Nova had a show saying that when identical twins are born their epigenomes are very similar, but that as they age their epigenomes start to differ more an more, and that certain behaviors like drinking or smoking can increase the rate at which such changes take place. What I didn't understand about the article you linked to is that it appears they are changing the epigenome to change the expression of DNA, but as far as I know DNA only appears in the nucleus (with the exception of mitochondirial DNA), and thus would appear to affect the cell as a whole, and thus not be good at differentially affecting the strengths of different synapses --- as would presumably be required for most neuronal memory --- unless the nuclear DNA had some sort of mapping to individual synapses, or unless local changes to mitochondrial DNA, near a synapse are involved. The article does not appear to shed in any light on this issue of how changes in the expression of DNA would affect learning at the synapse level, where most people think it occurs. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent:
RE: [agi] Lamarck Lives!(?)
Richard, You asked can the state of the switches be preserved during reproduction? According to the Nova show I saw about epigenome, they were able to induce a change in a mouse's epigenome that changed its appearance, then its children would be more likely to inherit the same changed appearance. They could also unchanged that particular epigenomic trait back to what it had been in a parent or grandparent. So they were able to change and unchanged traits that were inheritable. So the answer is yes. Ed Porter -Original Message- From: Richard Loosemore [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 2:44 PM To: agi@v2.listbox.com Subject: Re: [agi] Lamarck Lives!(?) Ben Goertzel wrote: I know you're just playing here but it would be easy to empirically test this. Does junk DNA change between birth and death? Something tells me we would have discovered something that significant a long time ago. Terren well, loads of mutations occur in nuclear DNA between birth and death; this is part of how aging occurs. There are specific DNA repair mechanisms that fix mutation errors that occur during the cell's lifetime It seems quite plausible that these repair mechanisms might work differently on coding and noncoding regions of the DNA Ah, hang on folks: what I was meaning was that the *state* of the junk DNA was being used, not the code. I am referring to the stuff that is dynamically interacting, as a result of which genes are switched on and off all over the place so this is a gigantic network of switches. I wouldn't suggest that something is snipping and recombining the actual code of the junk DNA, only that the state of the switches is being used to code for something. Question is: can the state of the switches be preserved during reproduction? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Possibly... it has been shown with methylation. But I think the mechanism you're proposing could not involve methylation because (someone can correct me if wrong) methylation is only applicable to coding regions (methyl group only added to specific DNA sequences that mark the gene). That's not to say another switching mechanism on non-coding regions could not also be heritable (i.e., reproduced in the copied DNA strand). Using DNA switches (such as methylation) is more tractable than DNA rewriting, but again, the amount of information storage is the limiting factor. Indeed, switching on and off sections of DNA implies a big reduction in information capacity (as compared to DNA rewriting), since gene switching applies to sections of DNA. I wonder how much memory would you expect to be able to pass on through this mechanism? Also, you would need to propose the mechanism by which this form of storage would be read. Since junk DNA by definition doesn't code for anything, by what mechanism would these switches have an effect on cellular, neural, or otherwise cognitive processes? Terren --- On Wed, 12/3/08, Richard Loosemore [EMAIL PROTECTED] wrote: Ah, hang on folks: what I was meaning was that the *state* of the junk DNA was being used, not the code. I am referring to the stuff that is dynamically interacting, as a result of which genes are switched on and off all over the place so this is a gigantic network of switches. I wouldn't suggest that something is snipping and recombining the actual code of the junk DNA, only that the state of the switches is being used to code for something. Question is: can the state of the switches be preserved during reproduction? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Seeking CYC critiques
Steve, Based on your attached response, How about this alternative approach: Send (one of) them an email pointing out http://www.dreliza.com/standards.php which will obviously usurp their own efforts if they fail to participate, and offer them an opportunity to suggest amendments these standards to incorporate (some of) their own capabilities. Seeing that Dr. Eliza's approach is quite different, they should then figure out that their only choices are to join or die. I wonder how they would respond? You know these guys. How would YOU play this hand? Any thoughts? Steve Richfield On 12/2/08, Stephen Reed [EMAIL PROTECTED] wrote: Steve Richfield said: If I understand you correctly, Cycorp's code should be *public domain*, and as such, I should be able to simply mine for the features that I am looking for. It sounds like Cycorp doesn't have a useful product (yet) whereas it looks like I do, so it is probably I who should be doing this, not Cycorp. Regretfully, the KRAKEN source code is not public domain, despite the fact that US tax dollars paid for it. While at Cycorp, John DeOliveira and I lobbied for an open-source version of Cyc, that one of us dubbed OpenCyc. Doug Lenat saw the advantages of releasing a limited form of Cyc technology, especially to preclude some other possible ontology from becoming the de facto standard ontology, e.g. for the Semantic Web. However, Cycorp is bedeviled by its own traditional, proprietary nature and Lenat did not want to release the source code for the object store, lisp runtime, inference engine, applications and utilities. The first release of OpenCyc that I prepared contained many, but not all, of the full Cyc concept terms, and their defining assertions. No rules, nor numerous other commonsense assertions about these concepts were released. The provided OpenCyc runtime was binary only, without source code, and with its HTML browser as its sole released application. A Java API to Cyc, that I wrote, was also released with its source code under the Apache License. The KRAKEN application is not provided with OpenCyc, and it was growing stale from lack of maintenance when I was let go from the company in August 2006. -Steve Stephen L. Reed Artificial Intelligence Researcher http://texai.org/blog http://texai.org 3008 Oak Crest Ave. Austin, Texas, USA 78704 512.791.7860 -- *From:* Steve Richfield [EMAIL PROTECTED] *To:* agi@v2.listbox.com *Sent:* Monday, December 1, 2008 10:22:37 PM *Subject:* Re: [agi] Seeking CYC critiques Steve, If I understand you correctly, Cycorp's code should be *public domain*, and as such, I should be able to simply mine for the features that I am looking for. It sounds like Cycorp doesn't have a useful product (yet) whereas it looks like I do, so it is probably I who should be doing this, not Cycorp. Any thoughts? Who should I ask for code from? Steve Richfield == On 12/1/08, Stephen Reed [EMAIL PROTECTED] wrote: Steve Richfield said: KRAKEN contains lots of good ideas, several of which were already on my wish list for Dr. Eliza sometime in the future. I suspect that a merger of technologies might be a world-beater. I wonder if the folks at Cycorp would be interested in such an effort? If you can find a sponsor for the effort and then solicit Cycorp to join in collaboration, I believe that they would be interested. The Cycorp business model as I knew it back in 2006, depended mostly upon government research sponsorship to (1) accomplish the research that the sponsor wanted, e.g. produce deliverables for the DARPA Rapid Knowledge Formation project, and (2) incrementally add more facts and rules to the Cyc KB, write more supporting code for Cyc. Cycorp, did not then, and likely even now does not have internal funding for non-sponsored enhancements. -Steve Stephen L. Reed Artificial Intelligence Researcher http://texai.org/blog http://texai.org 3008 Oak Crest Ave. Austin, Texas, USA 78704 512.791.7860 -- *From:* Steve Richfield [EMAIL PROTECTED] *To:* agi@v2.listbox.com *Sent:* Monday, December 1, 2008 3:19:37 PM *Subject:* Re: [agi] Seeking CYC critiques Steve, The KRAKEN paper was quite interesting, and has a LOT in common with my own Dr. Eliza. However, I saw no mention of Dr. Eliza's secret sauce, that boosts it from answering questions to solving problems given symptoms. The secret sauce has two primary ingredients: 1. The syntax of differential symptom statements - how people state a symptom that separates it from similar symptoms of other conditions. 2. Questions, the answers to which will probably carry #1 above recognizable differential symptom statements. Both of the above seem to require domain *experienced* people to code, as book learning doesn't seem to convey what people typically say, or what you have to say to them
Re: [agi] Seeking CYC critiques
Steve, Frankly, I find it rather unlikely that Cyc would consider it obvious that your project will usurp their own efforts if they fail to participate. Two points 1) it's not as though you have any really awesome demonstrated results with Dr. Eliza, that would compel them 2) even if you did, my impression as an outsider is that the Cyc crew is extremely fixated on their own approach and more resistant to change than most AI organizations. They *have* changed their approach in significant ways over time (e.g. introducing context, introducing uncertainty etc. into their representation) ... but, still, I think my point holds... Please note that the vast majority of the AI community thinks Cyc should change what they're doing in various significant ways, but they are staying true to their course, which is either noble persistence or foolish stubbornness depending on your perspective ;-) ben g On Wed, Dec 3, 2008 at 3:55 PM, Steve Richfield [EMAIL PROTECTED] wrote: Steve, Based on your attached response, How about this alternative approach: Send (one of) them an email pointing out http://www.dreliza.com/standards.php which will obviously usurp their own efforts if they fail to participate, and offer them an opportunity to suggest amendments these standards to incorporate (some of) their own capabilities. Seeing that Dr. Eliza's approach is quite different, they should then figure out that their only choices are to join or die. I wonder how they would respond? You know these guys. How would YOU play this hand? Any thoughts? Steve Richfield On 12/2/08, Stephen Reed [EMAIL PROTECTED] wrote: Steve Richfield said: If I understand you correctly, Cycorp's code should be public domain, and as such, I should be able to simply mine for the features that I am looking for. It sounds like Cycorp doesn't have a useful product (yet) whereas it looks like I do, so it is probably I who should be doing this, not Cycorp. Regretfully, the KRAKEN source code is not public domain, despite the fact that US tax dollars paid for it. While at Cycorp, John DeOliveira and I lobbied for an open-source version of Cyc, that one of us dubbed OpenCyc. Doug Lenat saw the advantages of releasing a limited form of Cyc technology, especially to preclude some other possible ontology from becoming the de facto standard ontology, e.g. for the Semantic Web. However, Cycorp is bedeviled by its own traditional, proprietary nature and Lenat did not want to release the source code for the object store, lisp runtime, inference engine, applications and utilities. The first release of OpenCyc that I prepared contained many, but not all, of the full Cyc concept terms, and their defining assertions. No rules, nor numerous other commonsense assertions about these concepts were released. The provided OpenCyc runtime was binary only, without source code, and with its HTML browser as its sole released application. A Java API to Cyc, that I wrote, was also released with its source code under the Apache License. The KRAKEN application is not provided with OpenCyc, and it was growing stale from lack of maintenance when I was let go from the company in August 2006. -Steve Stephen L. Reed Artificial Intelligence Researcher http://texai.org/blog http://texai.org 3008 Oak Crest Ave. Austin, Texas, USA 78704 512.791.7860 From: Steve Richfield [EMAIL PROTECTED] To: agi@v2.listbox.com Sent: Monday, December 1, 2008 10:22:37 PM Subject: Re: [agi] Seeking CYC critiques Steve, If I understand you correctly, Cycorp's code should be public domain, and as such, I should be able to simply mine for the features that I am looking for. It sounds like Cycorp doesn't have a useful product (yet) whereas it looks like I do, so it is probably I who should be doing this, not Cycorp. Any thoughts? Who should I ask for code from? Steve Richfield == On 12/1/08, Stephen Reed [EMAIL PROTECTED] wrote: Steve Richfield said: KRAKEN contains lots of good ideas, several of which were already on my wish list for Dr. Eliza sometime in the future. I suspect that a merger of technologies might be a world-beater. I wonder if the folks at Cycorp would be interested in such an effort? If you can find a sponsor for the effort and then solicit Cycorp to join in collaboration, I believe that they would be interested. The Cycorp business model as I knew it back in 2006, depended mostly upon government research sponsorship to (1) accomplish the research that the sponsor wanted, e.g. produce deliverables for the DARPA Rapid Knowledge Formation project, and (2) incrementally add more facts and rules to the Cyc KB, write more supporting code for Cyc. Cycorp, did not then, and likely even now does not have internal funding for non-sponsored enhancements. -Steve Stephen L. Reed Artificial Intelligence Researcher
Re: [agi] Lamarck Lives!(?)
Junk DNA doesn't code for protein, but it seems to carry out various control functions over the protein synthesis and interaction processes, no? ben g On Wed, Dec 3, 2008 at 4:02 PM, Terren Suydam [EMAIL PROTECTED] wrote: Possibly... it has been shown with methylation. But I think the mechanism you're proposing could not involve methylation because (someone can correct me if wrong) methylation is only applicable to coding regions (methyl group only added to specific DNA sequences that mark the gene). That's not to say another switching mechanism on non-coding regions could not also be heritable (i.e., reproduced in the copied DNA strand). Using DNA switches (such as methylation) is more tractable than DNA rewriting, but again, the amount of information storage is the limiting factor. Indeed, switching on and off sections of DNA implies a big reduction in information capacity (as compared to DNA rewriting), since gene switching applies to sections of DNA. I wonder how much memory would you expect to be able to pass on through this mechanism? Also, you would need to propose the mechanism by which this form of storage would be read. Since junk DNA by definition doesn't code for anything, by what mechanism would these switches have an effect on cellular, neural, or otherwise cognitive processes? Terren --- On Wed, 12/3/08, Richard Loosemore [EMAIL PROTECTED] wrote: Ah, hang on folks: what I was meaning was that the *state* of the junk DNA was being used, not the code. I am referring to the stuff that is dynamically interacting, as a result of which genes are switched on and off all over the place so this is a gigantic network of switches. I wouldn't suggest that something is snipping and recombining the actual code of the junk DNA, only that the state of the switches is being used to code for something. Question is: can the state of the switches be preserved during reproduction? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] I intend to live forever, or die trying. -- Groucho Marx --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
Ben, I basically agree. There many things going in the human brain. There are all the different neuro- chemicals, receptors, and blockers, some of which are not only effective across individual synapses, but often across broader distances. There is the fact that neuron branches can apparently grow in directions guided by chemical gradients. There are synchronies and brain waves, and the way in which they might spatially encode or decode information. And so on. So I admit the brain is much more complicated than most neural net models. But I have not seen any explanation of how changes in gene expression in a neuron's nucleus would store memories, even given the knowledge that the epigenome can store information. If there is such an explanation, either now or in the future, I would welcome hearing it. Ed Porter -Original Message- From: Ben Goertzel [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 3:24 PM To: agi@v2.listbox.com Subject: Re: [agi] Lamarck Lives!(?) On Wed, Dec 3, 2008 at 3:19 PM, Ed Porter [EMAIL PROTECTED] wrote: Terry and Ben, I never implied anything that could be considered a memory at a conscious level is stored at just one synapse, but all the discussions I have heard of learning in various brain science books and lectures imply synaptic weights are the main place of our memories are stored. Nevertheless, although it's an oft-repeated and well-spread meme, the available biological evidence shows only that **this is one aspect of the biological basis of memory in organisms with complex brains** There certainly is data about long-term potentiation and its relationship to memory ... but the available data comes nowhere near to justifying the sorts of assumptions made in setting up formal neural net models, in which synaptic modification is assumed as the sole basis of learning/memory... ben g --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
http://en.wikipedia.org/wiki/Epigenetic_inheritance#DNA_methylation_and_chromatin_remodeling The DNA sites where methylation can occur are rare, except in the regions where gene transcription occurs... which generally supports what I was saying about coding regions. However it is certainly possible that a different (as yet undiscovered) enzyme could methylate a different section of DNA that has no correlation at all with transcription. The key point is that it's certainly possible in principle to have some kind of signaling mechanism that uses junk DNA as a substrate, and which can be inherited epigenetically. It doesn't seem likely that methylation (as we know it) fits the bill, so probably Richard would require an as yet unknown mechanism for switching junk DNA. --- On Wed, 12/3/08, Ben Goertzel [EMAIL PROTECTED] wrote: Junk DNA doesn't code for protein, but it seems to carry out various control functions over the protein synthesis and interaction processes, no? ben g On Wed, Dec 3, 2008 at 4:02 PM, Terren Suydam [EMAIL PROTECTED] wrote: Possibly... it has been shown with methylation. But I think the mechanism you're proposing could not involve methylation because (someone can correct me if wrong) methylation is only applicable to coding regions (methyl group only added to specific DNA sequences that mark the gene). That's not to say another switching mechanism on non-coding regions could not also be heritable (i.e., reproduced in the copied DNA strand). Using DNA switches (such as methylation) is more tractable than DNA rewriting, but again, the amount of information storage is the limiting factor. Indeed, switching on and off sections of DNA implies a big reduction in information capacity (as compared to DNA rewriting), since gene switching applies to sections of DNA. I wonder how much memory would you expect to be able to pass on through this mechanism? Also, you would need to propose the mechanism by which this form of storage would be read. Since junk DNA by definition doesn't code for anything, by what mechanism would these switches have an effect on cellular, neural, or otherwise cognitive processes? Terren --- On Wed, 12/3/08, Richard Loosemore [EMAIL PROTECTED] wrote: Ah, hang on folks: what I was meaning was that the *state* of the junk DNA was being used, not the code. I am referring to the stuff that is dynamically interacting, as a result of which genes are switched on and off all over the place so this is a gigantic network of switches. I wouldn't suggest that something is snipping and recombining the actual code of the junk DNA, only that the state of the switches is being used to code for something. Question is: can the state of the switches be preserved during reproduction? Richard Loosemore --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] I intend to live forever, or die trying. -- Groucho Marx --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
RE: [agi] Lamarck Lives!(?)
I think the key is to see the gene switching not as an information store per se but as part of a larger dynamic process (which might be similar in principle to simulated annealing), in which the contributions of whole neurons (e.g., the outputs) are switched in some way meaningful to the dynamic. --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Ben, I basically agree. There many things going in the human brain. There are all the different neuro- chemicals, receptors, and blockers, some of which are not only effective across individual synapses, but often across broader distances. There is the fact that neuron branches can apparently grow in directions guided by chemical gradients. There are synchronies and brain waves, and the way in which they might spatially encode or decode information. And so on. So I admit the brain is much more complicated than most neural net models. But I have not seen any explanation of how changes in gene expression in a neuron's nucleus would store memories, even given the knowledge that the epigenome can store information. If there is such an explanation, either now or in the future, I would welcome hearing it. Ed Porter -Original Message- From: Ben Goertzel [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 3:24 PM To: agi@v2.listbox.com Subject: Re: [agi] Lamarck Lives!(?) On Wed, Dec 3, 2008 at 3:19 PM, Ed Porter [EMAIL PROTECTED] wrote: Terry and Ben, I never implied anything that could be considered a memory at a conscious level is stored at just one synapse, but all the discussions I have heard of learning in various brain science books and lectures imply synaptic weights are the main place of our memories are stored. Nevertheless, although it's an oft-repeated and well-spread meme, the available biological evidence shows only that **this is one aspect of the biological basis of memory in organisms with complex brains** There certainly is data about long-term potentiation and its relationship to memory ... but the available data comes nowhere near to justifying the sorts of assumptions made in setting up formal neural net models, in which synaptic modification is assumed as the sole basis of learning/memory... ben g --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com agi | Archives | Modify Your Subscription --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Yes. Ed etc., what comes to mind is Eugene Ishikevich (sp?) 's nonlinear dynamics models of fast and slow dynamics in neurons, which are based on ion channel models similar to (but more sophisticated in some cases than) the classic Hodgkin-Huxley equations Potentially the gene switching under discussion could affect some of the parameters in Ishikevich's equations, thus modifying the behavior of neurons. In Ishikevich's equations, classic LTP would take the form of modifying certain of the equational parameters, whereas this gene switching could take the form of modifying others. Note, I'm not selling Ishikevich's stuff as a final and correct model of neurodynamics -- just pointing it out as one interesting brain model that potentially would explain how these gene switches could conceivable affect learning via affecting holistic brain dynamics ... attractors and all that fun stuff... -- Ben G On Wed, Dec 3, 2008 at 5:12 PM, Terren Suydam [EMAIL PROTECTED] wrote: I think the key is to see the gene switching not as an information store per se but as part of a larger dynamic process (which might be similar in principle to simulated annealing), in which the contributions of whole neurons (e.g., the outputs) are switched in some way meaningful to the dynamic. --- On Wed, 12/3/08, Ed Porter [EMAIL PROTECTED] wrote: Ben, I basically agree. There many things going in the human brain. There are all the different neuro- chemicals, receptors, and blockers, some of which are not only effective across individual synapses, but often across broader distances. There is the fact that neuron branches can apparently grow in directions guided by chemical gradients. There are synchronies and brain waves, and the way in which they might spatially encode or decode information. And so on. So I admit the brain is much more complicated than most neural net models. But I have not seen any explanation of how changes in gene expression in a neuron's nucleus would store memories, even given the knowledge that the epigenome can store information. If there is such an explanation, either now or in the future, I would welcome hearing it. Ed Porter -Original Message- From: Ben Goertzel [mailto:[EMAIL PROTECTED] Sent: Wednesday, December 03, 2008 3:24 PM To: agi@v2.listbox.com Subject: Re: [agi] Lamarck Lives!(?) On Wed, Dec 3, 2008 at 3:19 PM, Ed Porter [EMAIL PROTECTED] wrote: Terry and Ben, I never implied anything that could be considered a memory at a conscious level is stored at just one synapse, but all the discussions I have heard of learning in various brain science books and lectures imply synaptic weights are the main place of our memories are stored. Nevertheless, although it's an oft-repeated and well-spread meme, the available biological evidence shows only that **this is one aspect of the biological basis of memory in organisms with complex brains** There certainly is data about long-term potentiation and its relationship to memory ... but the available data comes nowhere near to justifying the sorts of assumptions made in setting up formal neural net models, in which synaptic modification is assumed as the sole basis of learning/memory... ben g --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com agi | Archives | Modify Your Subscription agi | Archives | Modify Your Subscription -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] I intend to live forever, or die trying. -- Groucho Marx --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html Actually, this makes sense. It explains most of the discrepancy between the 10^9 bits of human long term memory estimated by Landauer and the 10^15 synapses in the human brain. If memory is stored in neurons (by gene regulation to control activation threshold), then you have only 10^11 bits of storage, or 1 bit per neuron. Here is how it could work. Imagine a neural network with fixed, randomly weighted synapses. Then insert a neuron at each synapse with one input and one output. Then you could apply Hebbian learning by modifying the conductivity of the middle neuron. If the input and output neurons fire at the same time, then the middle neuron would lower its threshold if both weights are the same, or raise it if the weights have opposite sign. In other words, instead of A - B with a variable weight, you have A - M - B with a middle neuron M of variable conductivity and two fixed weights. Of course real neurons have thousands of inputs and outputs. This means that there are thousands of neurons between A and B, and these middle neurons connect to thousands of others. If these connections are random, then Hebbian learning applied to these thousands of middle neurons would correlate only with AB and create minor noise for other neurons. -- Matt Mahoney, [EMAIL PROTECTED] --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Lamarck Lives!(?)
Well, LTP is definitely real ... and I'm quite sure the scheme you describe is *not* how learning works in the brain ;-) ,,, but I'm equally sure that the full story has not yet been uncovered... ben On Wed, Dec 3, 2008 at 5:25 PM, Matt Mahoney [EMAIL PROTECTED] wrote: http://www.newscientist.com/article/mg20026845.000-memories-may-be-stored-on-your-dna.html Actually, this makes sense. It explains most of the discrepancy between the 10^9 bits of human long term memory estimated by Landauer and the 10^15 synapses in the human brain. If memory is stored in neurons (by gene regulation to control activation threshold), then you have only 10^11 bits of storage, or 1 bit per neuron. Here is how it could work. Imagine a neural network with fixed, randomly weighted synapses. Then insert a neuron at each synapse with one input and one output. Then you could apply Hebbian learning by modifying the conductivity of the middle neuron. If the input and output neurons fire at the same time, then the middle neuron would lower its threshold if both weights are the same, or raise it if the weights have opposite sign. In other words, instead of A - B with a variable weight, you have A - M - B with a middle neuron M of variable conductivity and two fixed weights. Of course real neurons have thousands of inputs and outputs. This means that there are thousands of neurons between A and B, and these middle neurons connect to thousands of others. If these connections are random, then Hebbian learning applied to these thousands of middle neurons would correlate only with AB and create minor noise for other neurons. -- Matt Mahoney, [EMAIL PROTECTED] --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [EMAIL PROTECTED] I intend to live forever, or die trying. -- Groucho Marx --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=120640061-aded06 Powered by Listbox: http://www.listbox.com
Re: [agi] Seeking CYC critiques
Steve Richfield said: Seeing that Dr. Eliza's approach is quite different, they should then figure out that their only choices are to join or die. I wonder how they would respond? You know these guys. How would YOU play this hand? I believe that Larry Lefkowitz is still the marketing director at Cycorp. The company phone number is (512) 342-4000. You can mention that you and I have had an email discussion, by way of introduction. I suggest that you walk Larry through how Dr. Eliza works, taking care at the very first step to distance your software greatly from the Weizenbaum Eliza program of the same name. Remember that Larry's job is to sell Cyc, and he will be much more interested if he believes that Cycorp can make money as a result of talking to you. I would downplay email because it could be too easily ignored or misunderstood by Cycorp. FYI, here is Larry Lefkowitz's profile on Linkedin. Regards, -Steve Stephen L. Reed Artificial Intelligence Researcher http://texai.org/blog http://texai.org 3008 Oak Crest Ave. Austin, Texas, USA 78704 512.791.7860 From: Steve Richfield [EMAIL PROTECTED] To: agi@v2.listbox.com Sent: Wednesday, December 3, 2008 2:55:04 PM Subject: Re: [agi] Seeking CYC critiques Steve, Based on your attached response, How about this alternative approach: Send (one of) them an email pointing out http://www.dreliza.com/standards.php which will obviously usurp their own efforts if they fail to participate, and offer them an opportunity to suggest amendments these standards to incorporate (some of) their own capabilities. Seeing that Dr. Eliza's approach is quite different, they should then figure out that their only choices are to join or die. I wonder how they would respond? You know these guys. How would YOU play this hand? Any thoughts? Steve Richfield On 12/2/08, Stephen Reed [EMAIL PROTECTED] wrote: Steve Richfield said: If I understand you correctly, Cycorp's code should be public domain, and as such, I should be able to simply mine for the features that I am looking for. It sounds like Cycorp doesn't have a useful product (yet) whereas it looks like I do, so it is probably I who should be doing this, not Cycorp. Regretfully, the KRAKEN source code is not public domain, despite the fact that US tax dollars paid for it. While at Cycorp, John DeOliveira and I lobbied for an open-source version of Cyc, that one of us dubbed OpenCyc. Doug Lenat saw the advantages of releasing a limited form of Cyc technology, especially to preclude some other possible ontology from becoming the de facto standard ontology, e.g. for the Semantic Web. However, Cycorp is bedeviled by its own traditional, proprietary nature and Lenat did not want to release the source code for the object store, lisp runtime, inference engine, applications and utilities. The first release of OpenCyc that I prepared contained many, but not all, of the full Cyc concept terms, and their defining assertions. No rules, nor numerous other commonsense assertions about these concepts were released. The provided OpenCyc runtime was binary only, without source code, and with its HTML browser as its sole released application. A Java API to Cyc, that I wrote, was also released with its source code under the Apache License. The KRAKEN application is not provided with OpenCyc, and it was growing stale from lack of maintenance when I was let go from the company in August 2006. -Steve Stephen L. Reed Artificial Intelligence Researcher http://texai.org/blog http://texai.org 3008 Oak Crest Ave. Austin, Texas, USA 78704 512.791.7860 From: Steve Richfield [EMAIL PROTECTED] To: agi@v2.listbox.com Sent: Monday, December 1, 2008 10:22:37 PM Subject: Re: [agi] Seeking CYC critiques Steve, If I understand you correctly, Cycorp's code should be public domain, and as such, I should be able to simply mine for the features that I am looking for. It sounds like Cycorp doesn't have a useful product (yet) whereas it looks like I do, so it is probably I who should be doing this, not Cycorp. Any thoughts? Who should I ask for code from? Steve Richfield == On 12/1/08, Stephen Reed [EMAIL PROTECTED] wrote: Steve Richfield said: KRAKEN contains lots of good ideas, several of which were already on my wish list for Dr. Eliza sometime in the future. I suspect that a merger of technologies might be a world-beater. I wonder if the folks at Cycorp would be interested in such an effort? If you can find a sponsor for the effort and then solicit Cycorp to join in collaboration, I believe that they would be interested. The Cycorp business model as I knew it back in 2006, depended mostly upon government research sponsorship to (1) accomplish the research that the sponsor wanted, e.g. produce deliverables for the DARPA Rapid Knowledge Formation project,
Re: [agi] Seeking CYC critiques
Ben, Oops, I have made my usual mistake of presuming what is obvious to me. OK, to correct my error... I appears obvious to me that the first person who proposes the following things together as a workable standard, will own the future 'web. This because the world will enter the metadata faster than anyone is going to build a semantic web or anything like it without these items. In short, this is a sort of calculated retrograde step to get the goodies NOW and not sometime in the future. 1. Simple HTML meta commands to enter ontological meta information, so that information appearing on the web page can be properly filed without AI analysis of content. This completely avoids all of the usual disambiguation and machine understanding errors, so that ~99% accuracy should be achievable, ESPECIALLY when the agent that processes them is able to communicate any difficulties for humans to correct, and 2. #1 that works beyond searching and question-answering to function here and now to solve at least some interesting problems. Perhaps in years to come, people can omit some/all of this metadata and future AI interfaces to the web will still work, bit I simply see no reason to wait until then to smarten the 'web. Once the metadata is in place, any bright programmer can implement theInternet Singularity by simply populating his tables based on the metadata. I'll certify right up front that my proposed standard probably has its share of holes, and specifically, that it is completely oriented to problem solving and NOT to question answering, which Dr. Eliza presently doesn't even attempt. However, once stated (as I did on the hyperlinked site in my previous posting), the challenge is on the table for others to either do better, or to sit back and eat (or have shoved down their throats) a defacto standard that may not be at all to their liking. Stephen had already alluded to the probable intractability of Cyc, which you have certainly reinforced. THAT is why I sought to put them on notice to speak now, or forever hold your peace. Perhaps I should serve notice, complete with a certified mail requiring an official signature, that a standards effort is in process that they can either supplement, challenge, or ignore at their peril. Continuing with your comments... On 12/3/08, Ben Goertzel [EMAIL PROTECTED] wrote: Steve, Frankly, I find it rather unlikely that Cyc would consider it obvious that your project will usurp their own efforts if they fail to participate. Two points 1) it's not as though you have any really awesome demonstrated results with Dr. Eliza, that would compel them Yea, most people can't even talk realistically about chronic illnesses, so whatever its real-world value, it makes a really crappy demo. Consequently, I have added a new section about bad teeth. There, you can describe a problem that your dentist says is bad enough to have a tooth extracted, and Dr. Eliza will discuss your tooth, often finding some unorthodox way of saving it. Plans are to have this up and running on http://www.DrEliza.com by next Monday. 2) even if you did, my impression as an outsider is that the Cyc crew is extremely fixated on their own approach and more resistant to change than most AI organizations. The key to my approach is that they needn't change at all to bend my proposed standard. Of course, without considerable change, they'll remain stuck in question answering. No matter what happens, I win... 1. If they ignore me, I can simply point out that they chose not to participate in a standards effort. Even if Dr. Eliza fails in the market, the standard could still perpetuate. This would leave their ontological construction high and dry forever. 2. If they participate, then the standard would be stronger for everyone. 3. They might challenge with a standard of their own. If this fails to support problem solving, then they have limited both their standard and their product, and I would simply ignore them as would everyone else due to the limitations. If they incorporate problem solving, then they would have to expand their product and KB accordingly, which would be a BIG effort considering the size of their KB, and the fact that *experienced* domain experts are needed to code this stuff. Either way, this leave them out on a limb. They *have* changed their approach in significant ways over time (e.g. introducing context, introducing uncertainty etc. into their representation) ... but, still, I think my point holds... I have no reason to suspect that it doesn't. Please note that the vast majority of the AI community thinks Cyc should change what they're doing in various significant ways, but they are staying true to their course, which is either noble persistence or foolish stubbornness depending on your perspective ;-) Once we have kicked my thoughts around here and I run in whatever direction looks best, then I suspect that the issue of persistence vs. foolishness will be