In response to Vladimir Nesovs Fri 10/19/2007 5:28 AM post.
Nesov>> Edward, Nesov>> Does your estimate consider only amount of information required for *representation*, or it also includes additional processing elements required in neural setting to implement learning? EWP>> The large numbers I was discussing were for learned representation, including learned behaviors, such as many mental behaviors, but not actual pre-programmed or OS code in the traditional sense. Nesov>> I'm not sure 10^9 is far off, because much more can be required for domain-independent association/correlation catching between (subsymbolic) concepts implemented by groups of synapses(*). Gap of 10^6 is probably about right for this purpose, I can't see how it would be possible with, say, gap of only 10^2. EWP>> I dont know what you are referring to here. It is not clear whether you are supporting or attacking the notion that 10^9 bits is enough to store what the brain represents. So, if possible, please inform me in language that is a little more accessible to those not intimate with what you are referring to. EWP>> Assuming you are supporting the notion that 10^9 bits is enough, as your first clause suggests, then the only thing I can think of that might map into what you are saying is the concept of cell assemblies. The neurons in cell assemblies each take part in many cell assemblies, of say 10K neurons, and thus each of its synaptic weight is actually averaged over it roles in multiple different cell assemblies. In modeling this on a computer, this would mean a given variable in a model neuron would take part in the representation of many concepts. I have read implications that this could lead to significant representational efficiencies. EWP>> Many people seem to think such multiplexing cell assemblies are used in our brain. I have read arguments that a system with a given number of neurons can represent many more patterns using one cell assembly per pattern than if it used one neuron per pattern. (Resulting in the representational efficiencies I referred to above.) I havent seen any mathematic explanation of this. (if you know of any I would be interested in reading it). But it seems to me the more you multiplex a neuron the more cross-talk becomes an issue, particularly since different neurons in different areas of the cortex would tend to represent different connections for each of the different concepts they represent, meaning that the distribution of cross-talk between representations would not be statistically evenly spread across neurons or synapses, but could instead be unevenly concentrated at certain synapses, increasing the likelihood of cross-talk becoming a problem. EWP>> Since, many brain scientists say the brain uses this technique, and assuming it creates representational capacities which multiply the representational capacity of each synapse, then the issue is why would the cortex have at least 3x10^12 active synapses (I have heard from multiple sources that only about 1% of the average 10K synapse per neurons are built up enough to be really active, the other 99% are potential connections waiting to happen). Each synapse stores multiple variables, equal to, say, at least 2 to 8 bytes, and each, through its location in a complex connection space, represents at least a 3 to 7 byte address. So, if a synapse represents, say, 10 bytes, the brain has 3x10^13 bytes (3x10^14 bits) of representational capacity. And if there is some magic increase in representational capacity due to the use of cell assemblies, that would only further increase the number of bits the brain is arguably capable of representing. EWP>> The energy demands of the large human brain is a major evolutionary liability. It consumes more energy than any other organ. For example The average newborn's brain consumes an amazing 75-per cent of an infant's daily energy needs.1 This large liability can only be justified, in evolutionary terms, by the survival benefits the brains intelligence provides. So those roughly 3x10^13 bytes or more of representational capacity in the human cortex have had to earn their evolutionary keep. And thats not even mentioning the larger number or neurons in other parts of the brain. EWP>> Yes, the brain may be an inefficient design, but we have no strong reason to believe its efficiency would be less than one or two orders of magnitude worse than the representational scheme envisioned by the cognitive scientist who said the brain only stores 10^ 9 bits. EWP>> So, I still dont see how anyone can defend the notion that the human brain represents only 10^9 bits. Nesov>> New concepts/correlations/associations can be established between events (spikes) that are not initially aligned in any way, including different delays in time (through axonal delays and spiking sequences), so to catch regularities when and where they happen to appear, big enough amount of synapse groups should be there 'on watch'. EWP>> Again I dont know what you are referring to here. I understand that timing is important to neuronal patterns, but it seems that such added temporal complexity would only increase the number of bits required for a computer to model the information the brain holds. EWP>> 1. <http://www.eurekalert.org/pub_releases/2006-02/nsae-tsf021706.php> http://www.eurekalert.org/pub_releases/2006-02/nsae-tsf021706.php Edward W. Porter Porter & Associates 24 String Bridge S12 Exeter, NH 03833 (617) 494-1722 Fax (617) 494-1822 [EMAIL PROTECTED] -----Original Message----- From: Vladimir Nesov [mailto:[EMAIL PROTECTED] Sent: Friday, October 19, 2007 5:28 AM To: agi@v2.listbox.com Subject: Re: [agi] Poll Edward, Does your estimate consider only amount of information required for *representation*, or it also includes additional processing elements required in neural setting to implement learning? I'm not sure 10^9 is far off, because much more can be required for domain-independent association/correlation catching between (subsymbolic) concepts implemented by groups of synapses(*). Gap of 10^6 is probably about right for this purpose, I can't see how it would be possible with, say, gap of only 10^2. New concepts/correlations/associations can be established between events (spikes) that are not initially aligned in any way, including different delays in time (through axonal delays and spiking sequences), so to catch regularities when and where they happen to appear, big enough amount of synapse groups should be there 'on watch'. ----- (*) By groups of synapses I mean sets of synapses that can excite a common neuron, but single neuron can host multiple groups of synapses responsible for multiple subsymbolic concepts. It's not neurologically grounded, just a wild theoretic estimate. On 10/19/07, Edward W. Porter <[EMAIL PROTECTED]> wrote: Matt Mahoney's Thu 10/18/2007 9:15 PM post states MAHONEY>> There is possibly a 6 order of magnitude gap between the size of a cognitive model of human memory (10^9 bits) and the number of synapses in the brain (10^15), and precious little research to resolve this discrepancy. In fact, these numbers are so poorly known that we aren't even sure there is a gap. EWP>> This gap, which Matt was so correct to highlight, is an important one, and points out one of the many crippling legacy of the small hardware mindset. EWP>> I have always been a big believer in memory based reasoning, and for the last 37 years I have always assumed a human level representation of world knowledge would require something like 10^12 to 10^14 bytes, which is 10^13 to 10^15 bits. ( i.e., "within several orders of magnitude of the human brain", a phrase I have used so many times before on this list.) My recollection is that after reading Minsky's reading list in 1970 and my taking of K-line theory to heart, the number I guessed at that time for world knowledge was either 10^15 bits or bytes, I forget which. But, of course, my notions then were so primitive compared to what they are today. EWP>> Should we allow ourselves to think in terms of such big numbers? Yes. Let's take 10^13 bytes, for example. EWP>> 10^13 bytes with 2/3s of it in non-volatile memory and 10 million simple RAM opp processors, capable of performing about 20 trillion random RAM accesses/sec, and a network with a cross-sectional bandwidth of roughly 45 TBytes/sec (if you ran it hot), should be manufacturable at a marginal cost in 7 years of about $40,000, and could be profitably sold with amortization of development costs for several hundred thousand dollars if there were a market for several thousand of them -- which there almost certainly would be because of their extreme power. EWP>> Why so much more than the 10^9 bits mentioned above? EWP>> Because 10^9 bits only stores roughly 1 million atoms (nodes or links) with proper indexing and various state values. Anybody who thinks that is enough to represent human-level world knowledge in all its visual, audio, linguistic, tactile, kinesthetic, emotional, behavioral, and social complexity hasn't thought about it in sufficient depth. EWP>> For example, my foggy recollection is that Serre's representation of the hierarchical memory associated the portion of the visual cortext from V1 up to the lower level of the pre-frontal cortex (from the paper I have cited so many times on this list) has several million pattern nodes (and, as Josh has pointed out, this is just for the mainly feedforward aspect of visual modeling). This includes nothing for the vast majority of V1 and above, and nothing for audio, language, visual motion, associate cortex, prefrontal cortex, etc. EWP>> Matt, I am not in any way criticizing you for mentioning 10^9 bits, because I have read similar numbers myself, and your post pointed with very appropriate questioning to the gap between that and what the brain would appear to have the capabilility to represent. This very low number is just another manifestation of the small hardware mindset that has dominated the conventional wisdom in the AI since its beginning. If the only models one could make had to fit in the very small memories of most past machines, it is only natural that one's mind would be biased toward grossly simplified representation. EWP>> So forget the notion that 10^9 bits can represent human-level world knowledge. Correct me if I am wrong, but I think the memory required to store the representation in most current best selling video games is 10 to 40 times larger. Ed Porter P.S., Please give me feed back on whehter this technique of distinguishing original from responsive text is better than my use of all-caps, which received criticism. ----- This list is sponsored by AGIRI: <http://www.agiri.org/email> http://www.agiri.org/email To unsubscribe or change your options, please go to: <http://v2.listbox.com/member/?&> http://v2.listbox.com/member/?& _____ This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/? <http://v2.listbox.com/member/?&> & -- Vladimir Nesov mailto:[EMAIL PROTECTED] _____ This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/? <http://v2.listbox.com/member/?& > & ----- This list is sponsored by AGIRI: http://www.agiri.org/email To unsubscribe or change your options, please go to: http://v2.listbox.com/member/?member_id=8660244&id_secret=55360964-a2c1a5