Re: [agi] What Must a World Be That a Humanlike Intelligence May Develop In It?
On Sat, Jan 10, 2009 at 11:02 PM, Ben Goertzel b...@goertzel.org wrote: On a related note, even a very fine powder of very low friction feels different to water - how can you capture the sensation of water using beads and blocks of a reasonably large size? The objective of a CogDevWorld such as BlocksNBeadsWorld is explicitly **not** to precisely simulate the sensations of being in the real world. My question to you is: What important cognitive ability is drastically more easily developable given a world that contains a distinction between fluids and various sorts of bead-conglomerates? The objection is not valid in equating beads with dry powder. Certain forms of adhesion of the beads form a good approximation to fluids. You can have your hand wet with sticky beads etc. The model feels underspecified to me, but I'm OK with that, the ideas conveyed. It doesn't feel fair to insist there's no fluid dynamics modeled though ;-) Best regards. --- 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=126863270-d7b0b0 Powered by Listbox: http://www.listbox.com
Re: [agi] Direct communication between minds
Yes, the modality tapping as a target for BCI... The inner speech can be accessed this way (your phonological loop). The experiments here are less function-discriminative I think, it's rather what (classification) than how (inverse-engineering the imagination). http://scienceblogs.com/developingintelligence/2009/01/svms_decode_intentions_the_sta.php For the paper I want, I'm interested in the language aspect of all this (and especially AGI-related). What is the place for inner speech, communication of the system with itself that is amenable to episodic memory, in a cognitive architecture, its relation to thought in general? Is the inner speech language exactly the same as the language used for communication with others (e.g. in its semantics)? How to enhance the transfer of context-delineation? 2009/1/5 Abram Demski abramdem...@gmail.com: Lukasz, I think the most realistic near-term (next 10/20 years?) telepathy technology will deal with sensory modalities, rather than high-level concepts. In particular, it doesn't seem unrealistic to directly interface the phonological loop of two people, or similarly the visiospatial sketchpad. These low-level areas of the brain probably speak the same language from person to person. This would allow people to exchange any sounds or images that they imagined. I can think of some obvious questions. -How is it turned on/off, and how is it directed at particular people? -Is it possible to control which sounds/images are transferred, or would it be a dump of everything currently in working memory? -Can sounds and images coming in from the environment be easily shut out? -Would it be too unpleasant to have sounds or images forced upon one's imagination? Would current thoughts be erased, et cetera? --Abram --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] Direct communication between minds
On Tue, Jan 6, 2009 at 12:34 PM, Lukasz Stafiniak lukst...@gmail.com wrote: The experiments here are less function-discriminative I think, it's rather what (classification) than how (inverse-engineering the imagination). http://scienceblogs.com/developingintelligence/2009/01/svms_decode_intentions_the_sta.php But, the ability to transfer classifiers from a person to another without any re-training is really impressive. They suggest there are generic morphological thought patterns. For the paper I want, I'm interested in the language aspect of all this (and especially AGI-related). --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] Direct communication between minds
Hi, I plan to write a paper for a local CogSci conference, and a super-cool subject that comes to my mind is that of minds communicating through means more capable than what we're accustomed to. Is a revolution in language possible? Brainstorming: - Every now and then you hear that new developments could lead to telepathic devices (a naive post by the ever-famous Freeman Dyson here: http://www.edge.org/q2009/q09_3.html RADIOTELEPATHY, THE DIRECT COMMUNICATION OF FEELINGS AND THOUGHT FROM BRAIN TO BRAIN) - A promise that AGIs could communicate with their raw thoughts using their internal knowledge representation - A look from (AI-oriented) language semantics perspective: the dichotomy of the universal ontology approach (ontological semantics) vs. the lexicon-grounded semantics (multi-layered semantic networks) - Perhaps what I need to look at is the architecture of communication: a cascade (with feedback links) of -- the want to express / to interpret, let's stick with the sender side -- the activation of the concepts to express -- the crystallization of the message (this can occur online while expressing); it's a selection-abstraction phase that takes into account the whole effect of the message on the receiver -- the selection of proper expressive means (for example, the stream of words or gestures) - it's language everywhere: when we describe a process scientifically, we do it in some language; when we then engineer the process, it seems to us to have its true internal language; with AI, this is the KR, and the semantics are provided by mind dynamics; therefore there is an internal language to every mind, separate from the communication language, and thus the problem of expressing oneself as translation; I'm sure a whole bunch of academic philosophy deals with it (like the late Wittgenstein dismissing the translation idea) - A look at a cognitive architecture and how aspects of its operation can be communicated across different minds that instantiate it: referential meanings, complex concepts, emotions Could you share your thoughts? I'd appreciate relevant pointers. Happy New Year, Łukasz Stafiniak --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] [GoogleTech Talk] Case-based reasoning for game AI
The lecture is actually about more than just CBR. I recommend watching if you're bored, this is really entertaining :-) http://machineslikeus.com/news/video-case-based-reasoning-game-ai Bits seem similar to what Novamente is working on. Ambitious, but with engineering rather than AGI-focused spirit. --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] [Science Daily] Our Unconscious Brain Makes The Best Decisions Possible
http://www.sciencedaily.com/releases/2008/12/081224215542.htm Nothing surprising ;-) --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] Re: [OpenCog] Transfer learning
I'd like to propose two somewhat related papers: (Graph-Based Domain Mapping for Transfer Learning in General Games Gregory Kuhlmann, Peter Stone) http://www.cs.utexas.edu/~pstone/Papers/bib2html/b2hd-ECML07-rulegraphs.html Intrinsically Motivated Reinforcement Learning Nuttapong Chentanez Andrew G. Barto Satinder Singh http://www.eecs.umich.edu/~baveja/Papers/FinalNIPSIMRL.pdf On Tue, Dec 16, 2008 at 3:57 PM, Ben Goertzel b...@goertzel.org wrote: I just read an interesting (somewhat mathy) paper on transfer learning, and put the link here http://www.opencog.org/wiki/Transfer_Learning ben --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] Re: [OpenCog] Re: What is the role of MOSES in Novamente and Open Cog?-----was---- internship opportunity at Google (Mountain View, CA)
Talking on a very abstract level, MOSES could be ultimately developed so that it explores what you could call constraint relaxation strategies; combo trees are built such that they meet the constraints optimally with more-or-less random exploration of different trade-offs. Pure MOSES (current) starts with no knowledge and learns along the way, you could also call the knowledge learned constraints, especially once it would be expressed in transfer-friendly declarative way. Everything is a constraint. On Wed, Dec 17, 2008 at 4:23 PM, Ed Porter ewpor...@msn.com wrote: My point was the parallel constraint relaxation would appear able to be used much like a genetic algorithm, except that in a Novamente system it would have the ability to take advantage of much of the relevant world knowledge, and knowledge about how to best reason from world knowledge, that was contained in hypergraph, when proposing solutions. Your acknowledgement that in WebMind the hypergraph was used without MOSES, implies you agreement that the hypergraph could be used for exploring a possible solution space to various problems, and doing creative thinking. My REAL main point, was that from my reading about Combo and MOSES in your 2007 Novamente book, and from reading one of Moshe's long papers about it, MOSES seem to take to little advantage of all the rich, complex hierarchical and generalization knowledge contained in the hypergraph --- although it was clear to me that their would be ways in which it could be modified to do so. --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] internship opportunity at Google (Mountain View, CA)
Oh my, I've been very tired the other day! (as my English there shows...) I'm sorry for spamming the list. On Mon, Dec 15, 2008 at 11:41 PM, Lukasz Stafiniak lukst...@gmail.com wrote: I am initially interested but please consider other propositions as --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] internship opportunity at Google (Mountain View, CA)
Oh my, I've been very tired the other day! (as my English there shows...) I'm sorry for spamming the list. On Mon, Dec 15, 2008 at 11:41 PM, Lukasz Stafiniak lukst...@gmail.com wrote: I am initially interested but please consider other propositions as --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
Re: [agi] internship opportunity at Google (Mountain View, CA)
I am initially interested but please consider other propositions as well... I haven't decided yet (and I have some preconditions: it's overseas, I need to finish writing a draft of my PhD thesis first, etc). On Mon, Dec 15, 2008 at 7:32 PM, Moshe Looks madscie...@google.com wrote: * Learning procedural abstractions * Adapting estimation-of-distribution algorithms to program evolution * Applying plop to various interesting data sets * Adapting plop to do natural language processing or image processing * Better mechanisms for exploiting background knowledge in program evolution I'm generally interested in these topics... * Functional programming experience (esp. Lisp, but ML, Haskell, or even the functional style of C++ count too) OCaml is my programming language of choice for quite a lot of time, I have some recent Haskell experience, and quite a lot of Emacs Lisp experience. I don't know Common Lisp yet. * Experience with evolutionary computation or stochastic local search (esp. estimation-of-distribution algorithms and/or genetic programming) I'm teaching classes in Evolutionary Algorithms (and the lecturer puts emphasis on EDAs), I've taught classes in Data Mining, I'll be teaching classes in AI in the Spring (an introductory course). I have some experience with implementing GP and simulated annealing. * Open-source contributor Not much success stories or cooperation... Only: I've implemented a rich type system for Speagram, and pmwiki-mode for Emacs. http://pmwiki-mode.sourceforge.net/wiki/ (oups, the site is currently down; http://pmwiki-mode.cvs.sourceforge.net/viewvc/pmwiki-mode/pmwiki-mode/). --- 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=123753653-47f84b Powered by Listbox: http://www.listbox.com
[agi] Transfer Learning
Given the recent appearance of transfer learning here, it deserves a separate thread. Nice link: http://www.cs.utexas.edu/~lilyanam/TL/ Transfer Learning Reading Group at UT Austin Seems to be very AGI-friendly... The top read-link seems to be a good one to start: http://ftp.cs.wisc.edu/machine-learning/shavlik-group/torrey.aaai08.pdf Transfer in Reinforcement Learning via Markov Logic Networks (sounds NM/OpenCogPrime-ish) --- 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] General musings on AI, humans and empathy...
A beautiful post, Ben. Thank you. On Sun, Nov 9, 2008 at 12:44 AM, Ben Goertzel [EMAIL PROTECTED] wrote: http://multiverseaccordingtoben.blogspot.com/2008/11/in-search-of-machines-of-loving-grace.html -- Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI [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=117534816-b15a34 Powered by Listbox: http://www.listbox.com
[agi] Re: Two Remarkable Computational Competencies of the SGA
OK it's just a Compact Genetic Algorithm -- genetic drift kind of stuff. Nice read, but very simple (subsumed by any serious EDA). It says you can do simple pattern mining by just looking at the distribution, without complex statistics. On Wed, Oct 29, 2008 at 8:13 PM, Lukasz Stafiniak [EMAIL PROTECTED] wrote: Very relevant even if you don't agree. Too much rhetoric though (it's not really that earth-shaking). I haven't made up my mind yet. http://evoadaptation.wordpress.com/2008/10/18/new-manuscript-two-remarkable-computational-competencies-of-the-simple-genetic-algorithm/ --- 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=117534816-b15a34 Powered by Listbox: http://www.listbox.com
Re: [agi] a mathematical explanation of AI algorithms?
The more recent work by G. E. Hinton brought here by Ed Porter is very interesting mathematically (if you go into the details of trying to argument why it works -- probabilistic modeling a la graphical models). On Thu, Oct 9, 2008 at 12:32 AM, Ben Goertzel [EMAIL PROTECTED] wrote: For neural nets, Daniel Amit had a good book in the 80's reviewing the dynamics of attractor neural nets ... On Wed, Oct 8, 2008 at 6:25 PM, Vladimir Nesov [EMAIL PROTECTED] wrote: Read an introductory text on machine learning to get up to speed -- it's the math of AI, and there's lots of it. Statistics, information theory. It's an important perspective from which to look at less well understood hacks, to feel the underlying structure. --- 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=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
[agi] Fwd: Usage-based Computational Language Acquisition
Found this while grepping thru tons of posts on linguistlist (19.2502). Could be of interest. Date: Wed, 13 Aug 2008 13:39:46 From: Kerstin Fischer [EMAIL PROTECTED] Subject: Usage-based Computational Language Acquisition E-mail this message to a friend: http://linguistlist.org/issues/emailmessage/verification.cfm?iss=19-2502.htmlsubmissionid=186781topicid=3msgnumber=1 Full Title: Usage-based Computational Language Acquisition Date: 28-Jul-2009 - 03-Aug-2009 Location: Berkeley, CA, USA Contact Person: Kerstin Fischer Meeting Email: [EMAIL PROTECTED] Linguistic Field(s): Cognitive Science; Computational Linguistics; Language Acquisition Subject Language(s): English (eng) Call Deadline: 07-Sep-2008 Meeting Description: Usage-based models of language acquisition: computational perspectives Theme Session at ICLC 11, Berkeley, CA. Date: July 28-August 3, 2009 Organizers: Kerstin Fischer Arne Zeschel, University of Southern Denmark Call for Papers Theme Session Description: Usage-based approaches to language acquisition have not only produced many valuable insights in the field of child language studies (cf. Tomasello 2003 and Goldberg 2006 for overviews), but have also helped to corroborate important assumptions of emergentist theories of language in general (cf. Dabrowska 2005). In line with basic tenets of Cognitive Linguistics, these approaches emphasize the key role of communicative and experiential grounding in language use and language structure, and seek to explain its acquisition in terms of general (i.e., non-specialized) cognitive principles and mechanisms as far as possible. At the same time, explicit, testable models of how these principles and mechanisms are implemented in the context of grounded construction learning are only beginning to be developed (cf. Bod, to appear). The purpose of this workshop is to bring together language acquisition researchers from linguistics, psychology and computer science who work on such models in order to discuss how usage-based constructionist accounts of language acquisition can benefit from such research. Topics will include, but are not restricted to: - cognitive capacities that constitute prerequisites for normal child language acquisition (cf. Tomasello et al. 2005, Tomasello 2006) and how they can be accommodated in language learning simulations (e.g., Steels and Kaplan 2002); - the basic mechanisms and psycholinguistic plausibility of different approaches to automatic construction learning (e.g., Chang Maia 2001; Batali 2002; Steels 2004; Dominey and Boucher 2005); - the kinds of semantic representations that grounded language learning experiments or simulations should draw on (Bergen Chang 2005; Feldman 2006); - the way in which the acquisition of particular constructions may be grounded in the previous acquisition of certain other constructions (Johnson 2001; Morris, Cottrell Elman 2000; Abbot-Smith Behrens 2006); and, finally, - ways of accommodating useful notions from Cognitive Linguistics in computational models of language processing and acquisition (cf. Chang et al. 2002). The session will compare different approaches to automatic construction learning and consider the extent to which they can inform usage-based accounts of child language acquisition. In that, it seeks to bridge the gap between kindred research in Cognitive Linguistics and related areas of Cognitive Science, and to provide a forum for discussing important challenges for future research on emergentist models of language. Submission Procedure: Abstracts should be: - 500 words max - submitted in .rtf or .doc format - turned in by Sept 7th at the latest - accompanied by an e-mail specifying the title of the paper, name(s) of author(s), affiliation and a contact e-mail address - sent to [EMAIL PROTECTED] and [EMAIL PROTECTED] Please note that both the theme session proposal itself and the individual contributions will undergo independent reviewing by the ICLC program committee. --- 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=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Any further comments from lurkers??? [WAS do we need a stronger politeness code on this list?]
So far I've been in favor of mailing lists, because of their: - push rather than pull mechanism (all news in one reader) - filtering and/or threading (client-side) Arguably, threading (by subject lines), provided to me by gmail, is short-spanned, and forums share with wikis some of the top-down self-organization of knowledge which mailing lists completely lack. On Sun, Aug 3, 2008 at 8:53 PM, Ben Goertzel [EMAIL PROTECTED] wrote: The function you're describing as being carried out by an FAQ, would be served by a forum similar to the ImmInst fora, actually. ben On Sun, Aug 3, 2008 at 2:49 PM, Joseph Henry [EMAIL PROTECTED] wrote: I have seen very good and productive threads on this list, but they tend to be the exception. Hence I mostly just delete the items from the list, and follow the occasional thread that looks interesting or involves people who have posted more reasonable items in the past Yeah, that is typically what I do as well. Only a small number of threads ever make it past the 2-3 day lifespan in my inbox. I like the idea of the centralized FAQ (which I remember seeing the beginning of a while back). Not because I would point people to it, but rather I would find it useful to see others (or myself) pointed to it as I am still playing a long game of catch up. I think the FAQ should also include areas for people like Tintner to explain their theories in FULL detail to prevent any more confusion, arguments, or alienation. (Lets just put that to rest guys...) --- 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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
Re: FW: [agi] WHAT PORTION OF CORTICAL PROCESSES ARE BOUND BY THE BINDING PROBLEM?
On Tue, Jul 15, 2008 at 8:01 AM, Brad Paulsen [EMAIL PROTECTED] wrote: The terms forward-chaining and backward-chaining when used to refer to reasoning strategies have absolutely nothing to do with temporal dependencies or levels of reasoning. These two terms refer simply, and only, to the algorithms used to evaluate if/then rules in a rule base (RB). In the FWC algorithm, the if part is evaluated and, if TRUE, the then part is added to the FWC engine's output. In the BWC algorithm, the then part is evaluated and, if TRUE, the if part is added to the BWC engine's output. It is rare, but some systems use both FWC and BWC. That's it. Period. No other denotations or connotations apply. Curiously, the definition put by Abram Demski is the only one I've been aware of until yesterday (I believe it's the one used among theorem proving people). Let's see what googling says on forward chaining: 1. (Wikipedia) 2. http://www.amzi.com/ExpertSystemsInProlog/05forward.htm A large number of expert systems require the use of forward chaining, or data driven inference. [...] Data driven expert systems are different from the goal driven, or backward chaining systems seen in the previous chapters. The goal driven approach is practical when there are a reasonable number of possible final answers, as in the case of a diagnostic or identification system. The system methodically tries to prove or disprove each possible answer, gathering the needed information as it goes. The data driven approach is practical when combinatorial explosion creates a seemingly infinite number of possible right answers, such as possible configurations of a machine. 3. http://ai.eecs.umich.edu/cogarch0/common/prop/chain.html Forward-chaining implies that upon assertion of new knowledge, all relevant inductive and deductive rules are fired exhaustively, effectively making all knowledge about the current state explicit within the state. Forward chaining may be regarded as progress from a known state (the original knowledge) towards a goal state(s). Backward-chaining by an architecture means that no rules are fired upon assertion of new knowledge. When an unknown predicate about a known piece of knowledge is detected in an operator's condition list, all rules relevant to the knowledge in question are fired until the question is answered or until quiescence. Thus, backward chaining systems normally work from a goal state back to the original state. 4. http://www.ontotext.com/inference/reasoning_strategies.html * Forward-chaining: to start from the known facts and to perform the inference in an inductive fashion. This kind of reasoning can have diverse objectives, for instance: to compute the inferred closure; to answer a particular query; to infer a particular sort of knowledge (e.g. the class taxonomy); etc. * Backward-chaining: to start from a particular fact or from a query and by means of using deductive reasoning to try to verify that fact or to obtain all possible results of the query. Typically, the reasoner decomposes the fact into simpler facts that can be found in the knowledge base or transforms it into alternative facts that can be proven applying further recursive transformations. --- 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=108809214-a0d121 Powered by Listbox: http://www.listbox.com
Re: [agi] WHAT SORT OF HARDWARE $33K AND $850K BUYS TODAY FOR USE IN AGI
On Mon, Jun 30, 2008 at 8:07 AM, Terren Suydam [EMAIL PROTECTED] wrote: By the way, just wanted to point out a beautifully simple example - perhaps the simplest - of an irreducibility in complex systems. Individual molecular interactions are symmetric in time, they work the same forwards and backwards. Yet diffusion, which is nothing more than the aggregate of molecular interactions, is asymmetric. Figure that one out. This is just statistical mechanics. The interesting thing is that we make an opportunistic assumption, that any colliding particles are independent before collision (this introduces the time arrow), which is then empirically confirmed by the fact that derived properties agree with the phenomenological theory of entropy. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
Re: [agi] WHAT SORT OF HARDWARE $33K AND $850K BUYS TODAY FOR USE IN AGI
On Mon, Jun 30, 2008 at 8:07 AM, Terren Suydam [EMAIL PROTECTED] wrote: By the way, just wanted to point out a beautifully simple example - perhaps the simplest - of an irreducibility in complex systems. Individual molecular interactions are symmetric in time, they work the same forwards and backwards. Yet diffusion, which is nothing more than the aggregate of molecular interactions, is asymmetric. Figure that one out. This is just statistical mechanics. The interesting thing is that we make an opportunistic assumption, that any colliding particles are independent before collision (this introduces the time arrow), which is then empirically confirmed by the fact that derived properties agree with the phenomenological theory of entropy. P.S. The biggest issue that spoiled my joy of reading Permutation City is that you cannot simulate dynamic systems ( = solve numerically differential equations) out-of-order, you need to know time t to compute time t+1 (or, alternatively, you need to know t+2), the same goes for space, I presume you need to know x-1,x,x+1 to compute the next-step x. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=106510220-47b225 Powered by Listbox: http://www.listbox.com
[agi] Adaptivity in Hybrid Cognitive Systems Osnabruck PhD program
I'm not affiliated but I've found this interesting. They seem to have 8 positions for PhD students: http://www.cogsci.uni-osnabrueck.de/PhD/GK/ Their research program is really worth checking-out: http://www.cogsci.uni-osnabrueck.de/PhD/GK/research/body.html --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On Mon, May 26, 2008 at 1:26 PM, Mark Waser [EMAIL PROTECTED] wrote: What I'd rather do instead is see if we can get a .NET parallel track started over the next few months, see if we can get everything ported, and see the relative productivity between the two paths. That would provide a provably true answer to the debate. There are also sane languages using the C++ object model (http://felix-lang.org/). And there is Mono, though I've heard it falls behind .NET considerably in terms of efficiency. The thing is, will multi-language sourcing be encouraged? (Will every contributor be allowed to write in his language, provided it compiles with the rest?) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] More Info Please
On Sun, May 25, 2008 at 10:26 PM, Ben Goertzel [EMAIL PROTECTED] wrote: Certainly there are plenty of folks with equal software engineering experience to you, advocating the Linux/C++ route (taken in the current OpenCog version) rather than the .Net/C# route that I believe you advocate... No, I believe he advocates OCaml vs. F# ;-) (sorry for leaving-out Haskell and others) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
[agi] Re: Type inference
On Tue, May 20, 2008 at 10:05 AM, Lukasz Stafiniak [EMAIL PROTECTED] wrote: On Tue, May 20, 2008 at 5:16 AM, Stephen Reed [EMAIL PROTECTED] wrote: Code synthesis, according to my plan, should avoid the need for type inference. The AGI would know in advance the type(s) of the variable. Do you see a use for type inference in my work that I have overlooked? How would it know in advance what type of the variable does it need? Perhaps to compute some result it would need some other or more specific arguments than initially conceived? In my opinion, planning and program synthesis are closely related, and type inference is just a way of looking at some issues involved. (I repeat some of my message below, for the group.) You can avoid type inference (or something equivalent) only if you use propositional logic (in a specific sense), a logic where you cannot specify properties of objects piece-by-piece. If you can, you need to track what properties exactly a variable (a parameter, a result of some computation/action) has to have having already occurred in contexts it has occurred in, to know where it can be applied / what can be done with it further on. I call this type inference. You can't guess the program at once and then typecheck it, you need do type inference as you go. Oh, a P.S.: you use incremental parsing, interpreting a sentence word-by-word. Type inference is a (usually limited and static) form of program interpretation. It needs to be done incrementally to accompany program synthesis. And this is sometimes difficult! I don't do it incrementally yet (in the system I currently work on). The relation between type inference and program synthesis can be seen from the parallels between algorithm W (the classical type inference for the core of ML languages) and my algorithm C (for program synthesis in the same type system) [http://www.ii.uni.wroc.pl/~lukstafi/pmwiki/uploads/Main/DM_generation.pdf]. Interested people can look at the translated beginning of [http://www.ii.uni.wroc.pl/~lukstafi/pmwiki/uploads/Main/dyplom_en.pdf]. My ideas have evolved since back then. Currently I work on type inference (yeah) in a bit more expressive context. The page of my system: http://www.ii.uni.wroc.pl/~lukstafi/pmwiki/index.php?n=Infer.Infer Using weak constraints would add it more AGIsh flavor, but my work is in programming languages theory context and I needed the simplest thing possible. Currently (in addition to the glue type-trees) I have linear inequalities. I plan to add Datalog as a sublogic, then it should at least start looking a bit more like AI... Here two examples with linear arithmetic: input: newtype Bar newtype List : nat newcons LNil : List 0 newcons LCons : for all (n) : Bar * List(n) -- List(n+1) let rec split = function LNil - LNil, LNil | LCons (x, LNil) as y - y, LNil | LCons (x, LCons (y, z)) - match split z with (l1, l2) - LCons (x, l1), LCons (y, l2) output: split1 : [?gen10 = ?gen9; ?gen9 = ?gen10 + 1;] List (?gen10 + ?gen9) - List ?gen9, List ?gen10 -- what this means: split (the first thing defined by this name) is a function that takes a list of length a + b and returns a pair of lists, one of length a and the other of length b, where b = a and a = b+1, that is they are of roughly the same length. input: newtype Binary : nat newtype Carry : nat newcons Zero : Binary 0 newcons PZero : for all (n) : Binary(n) -- Binary(n+n) newcons POne : for all (n) : Binary(n) -- Binary(n+n+1) newcons CZero : Carry 0 newcons COne : Carry 1 let rec plus = function CZero - (function Zero - (fun b - b) | PZero a1 as a - (function Zero - a | PZero b1 - PZero (plus CZero a1 b1) | POne b1 - POne (plus CZero a1 b1)) | POne a1 as a - (function Zero - a | PZero b1 - POne (plus CZero a1 b1) | POne b1 - PZero (plus COne a1 b1))) | COne - (function Zero - (function Zero - POne(Zero) | PZero b1 - POne b1 | POne b1 - PZero (plus COne Zero b1)) | PZero a1 as a - (function Zero - POne a1 | PZero b1 - POne (plus CZero a1 b1) | POne b1 - PZero (plus COne a1 b1)) | POne a1 as a - (function Zero - PZero (plus COne a1 Zero) | PZero b1 - PZero (plus COne a1 b1) | POne b1 - POne (plus COne a1 b1))) output: plus1 : [?vCarry_n_12 = 1;] Carry ?vCarry_n_12 - Binary ?gen10 - Binary ?gen11 - Binary (?gen10 + ?gen11 + ?vCarry_n_12) - what this means: plus is a function that takes a value of type Carry c and binary numbers Binary a and Binary b, and returns a binary number Binary (a+b+c) where the carry token c = 1 (numbers in types are natural numbers). --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id
Re: [agi] Uninterpreted RDF terms
Word Grammar comes to my mind, where when A -R- B, and A' is-a A, then you know A' -R- B' where B' is-a B. Because I want to have lattices (partial orders) in my system anyway, and because nodes of my graph-terms might be objects of any domain (they can be nested graph-terms even), they could usually be from the partial orders domain and I could add subtyping semantics. So, inheritance relations would be specified separately from graph-terms. So, you would write: (rdf:type ?object-type ?X) (is-a ?X owl:Thing) where is-a is not part of the graph-term, it rather relates graph-terms. Or you could look at it as a special edge, similarly to Word Grammar. On Sun, May 18, 2008 at 4:17 AM, Stephen Reed [EMAIL PROTECTED] wrote: Hi Lukasz, Here is a typical Capability Description from my first set of bootstrap cases: (capability name: defineInstanceVariable description: Defines an instance variable having the given name and object type. preconditions: (rdf:type ?variable-name cyc:NonEmptyCharacterString) (rdf:type ?object-type owl:Thing) (rdf:type ?variable-comment cyc:NonEmptyCharacterString) (rdf:type ?variable-invariant-conditions cyc:Tuple) (implies (cyc:memberOfTuple ?variable-invariant-condition ?variable-invariant-conditions) (rdf:type ?variable-invariant-condition cyc:CycLFormula)) input-roles: (texai:blRole ?variable-name a variable name) (texai:blRole ?object-type a type) (texai:blRole ?variable-comment a comment) (texai:blRole ?variable-invariant-conditions some invariant conditions) output-roles: (texai:blRole ?defined-instance-variable the defined instance variable) postconditions: ;;TODO properties of the output with regard to the inputs (rdf:type ?defined-instance-variable texai:org.texai.bl.domainEntity.BLInstanceVariable) ) I think that the restriction you propose is not expressive enough to handle this case. If I am wrong please correct me. The matching is performed on the preconditions, postconditions and invariant-conditions. The latter is not illustrated in this example but consist of implications similar in form to the one found in the preconditions of this example. For the others on this list following my progress, the example is from a set of essential capability descriptions that I'll use to bootstrap the skill acquisition facility of the the Texai dialog system. The subsumption-based capability matcher is done. I'm writing Java code that implements each of these capabilities. That should be completed in a few more days, and then I'll fit that into the already completed dialog system. At that point I should be able to begin exploring what essential utterances will be needed to acquire skills by being taught, and generate Java programs to perform them. -Steve --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
[agi] (Interpreted!) RDF terms
Because I use saturation anyway, my algorithms can be parameterized with a monotonic consequence operator, which could implement an implicational theory, with rules (at least) of the form for all (...) [ Phi == exist (...) Psi ] where Phi and Psi are conjunctions of RDF atoms plus Psi could also contain equalities. I could also add negation for atoms, by restricting labels and not-labels (in a similar way to how offspring_i and offspring_j would be restricted), but observe that there would be no closed-world assumption: lack of an edge means I don't know yet. This is similar to bottom-up logic programming. On Sat, May 17, 2008 at 7:40 PM, Lukasz Stafiniak [EMAIL PROTECTED] wrote: Steve, How severe would you consider a restriction on RDF graphs that would allow at most one incoming and at most one outgoing edge with a given label, for capability descriptions? This would allow to do unification (and generalization aka. intersection) on graphs easily (not as easily as on terms, but nearly). Outside the system where it would be needed (I have automatic programming / program analysis in mind), the theory/graphs can be extended of course. For example, the parenting relation would have to be split into x offspring_i y means x is the i-th offspring of y, and we could also add outgoing and incoming restrictions, e.g. that a node cannot have incoming offspring_i and offspring_j edges for i j. Outside, we would have the implication x offspring_i y == x offspring y. Best wishes. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=103754539-40ed26 Powered by Listbox: http://www.listbox.com
Re: [agi] AGI-08 videos
On Tue, May 6, 2008 at 4:07 PM, Mark Waser [EMAIL PROTECTED] wrote: Note: Most of these complaints do *NOT* apply to Texai (except possibly the two to five level complaint -- except that Texai is actually starting at what I would call one of the middle levels and looks like it has reasonable plans for branching out. Texai has the added value of freshness, but the challenge Steve is facing now is perhaps bigger than the ones he has conquered already: to reflect on the system's state and to represent, learn and reason about actions. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Interesting approach to controlling animated characters...
IMHO, Euphoria shows that pure GA approaches are lame. More details here: http://aigamedev.com/editorial/naturalmotion-euphoria On Thu, May 1, 2008 at 5:39 PM, Ben Goertzel [EMAIL PROTECTED] wrote: Now this looks like a fairly AGI-friendly approach to controlling animated characters ... unfortunately it's closed-source and proprietary though... http://en.wikipedia.org/wiki/Euphoria_%28software%29 ben --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Interesting approach to controlling animated characters...
If I was paid to get a good animation, I would cheat: I would use -- mixed forward/inverse dynamics instead of pure (forward) simulation, -- motion capture data mining, -- hand-crafted parameterized models instead of generic NNs On Thu, May 1, 2008 at 8:19 PM, Ben Goertzel [EMAIL PROTECTED] wrote: Actually, it seems their technique is tailor-made for imitative learning If you gathered data about how people move in a certain context, using motion capture, then you could use their GA/NN stuff to induce a program that would generate data similar to the motion-captured data. This would then be more generalizable than using the raw motion-capture data -- Ben On Thu, May 1, 2008 at 2:11 PM, Lukasz Stafiniak [EMAIL PROTECTED] wrote: IMHO, Euphoria shows that pure GA approaches are lame. More details here: http://aigamedev.com/editorial/naturalmotion-euphoria On Thu, May 1, 2008 at 5:39 PM, Ben Goertzel [EMAIL PROTECTED] wrote: Now this looks like a fairly AGI-friendly approach to controlling animated characters ... unfortunately it's closed-source and proprietary though... http://en.wikipedia.org/wiki/Euphoria_%28software%29 ben --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Between logical semantics and linguistic semantics
On Mon, Apr 14, 2008 at 5:14 PM, Stephen Reed [EMAIL PROTECTED] wrote: My first solution to this problem is to postpone it by employing a controlled English, in which such constructions will be avoided if possible. Secondly, Jerry Ball demonstrated his solution in Double R Grammar at the 2007 AAAI Fall Symposium, Cognitive Approaches to NLP. His slide presentation is here, which I think fully addresses your issues. To summarize Dr. Ball's ideas, which I will ultimately adopt for Texai: Thanks, very interesting slides. I think he forgets to mention Dynamic Syntax (Ruth Kempson, Dov Gabbay). Serial processing [word by word parsing] with algorithmic backtracking has no hope for on-line processing in real-time in a large coverage NLP system. I think that Double R accomodation approach can be approximated by incremental right-to-left parsing. Something along the lines of http://www.speagram.org/wiki/Grammar/ChartParser but still needs much work (the approach was developed when I've been in computational semantics phase, it ignores cognitive linguistics, and is too fragmented: only categorical semantics (and agreement, by use of variables in types) are processed, with relational and referential semantics postponed to latter stages). The up side is that it can handle general Context Free Grammars. I didn't know that Microsoft uses some kind of right-to-left parsing, I thought it is my invention :-) I regret that some aspects of my implementation are difficult to follow because I am using Jerry Ball's Double R Grammar, but not his ACT-R Lisp engine, using instead my own incremental, cognitively plausible, version of Luc Steel's Fluid Construction Grammar engine. I combined these two systems because Jerry Ball's engine is not reversible, Luc Steel's grammar is not a good coverage of English, and the otherwise excellent Fluid Construction Grammar engine is not incremental. -Steve Perhaps you could get some linguist to capitalize on your work with a publication? Best Regards, Łukasz --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Between logical semantics and linguistic semantics
2008/4/14 Lukasz Stafiniak [EMAIL PROTECTED]: On Mon, Apr 14, 2008 at 5:14 PM, Stephen Reed [EMAIL PROTECTED] wrote: Serial processing [word by word parsing] with algorithmic backtracking has no hope for on-line processing in real-time in a large coverage NLP system. I think that Double R accomodation approach can be approximated by incremental right-to-left parsing. Something along the lines of http://www.speagram.org/wiki/Grammar/ChartParser but still needs much If you're confused by the equations, increments are left-to-right, and between increments, there's right-to-left accomodation-like stage. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=101455710-f059c4 Powered by Listbox: http://www.listbox.com
Re: [agi] Between logical semantics and linguistic semantics
On Wed, Apr 9, 2008 at 6:03 AM, Stephen Reed [EMAIL PROTECTED] wrote: I would be interested in your comments on my adoption of Fluid Construction Grammar as a solution to the NL to semantics mapping problem. (1) Word Grammar (WG) is a construction-free version of your approach. It is based solely on spreading activation. It doesn't have a sharp separation of syntax and semantics: there's only one net. Nodes representing subgraphs corresponding to constructions can be organized into inheritance hierarchies (extensibility). But pure WG makes things very awkward logics-wise, making it work would be a lot of research (the WG book doesn't discuss utterance generation IIRC, but reversing parsing-interpretation seems quite direct: select the most activated word which doesn't have a left landmark, introduce a word-instance node for it, include spreading its activation through a right-landmark (ignoring direction of the landmark) edge). Texai is impure by its very nature, perhaps it could be made more (than just sharing the spreading activation idea) of a mix WG*FCG. (2) FCG is closer to traditional apporaches a la computational linguistics than WG. (3) One could give up some FCG features to simplify it, for example by assuming one-to-one correspondence between constructions and atomic predicates. (4) I'm interested in how do you handle backtracking: giving up on application of a construction when it leads to inconsistency. Chart-based unification parsing can be optimized to share applications of constructions which are parallel, and this can be extended to operators which are (like unification) monotonic, e.g. cannot make unsatisfiable/inconsistent state a satisfiable/consistent one. Merging conjuncts new facts to old ones so it is monotonic in monotonic logics. (Default/defeasible logics are nonmonotonic.) (4a) Does the fact that your parser is incremental mean that you do early commitment to constructions? (Double R Grammar seems to support early commitment when there is choice, but backtracking is still needed to get an interpretation when there are only ones without it.) I will get to studying your sources when I'll have some time... --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
Re: [agi] Between logical semantics and linguistic semantics
On Thu, Apr 10, 2008 at 2:49 AM, Pei Wang [EMAIL PROTECTED] wrote: Lukasz, Thanks! To me, your logical semantics and linguistic semantics correspond to meaning of concepts and meaning of words, respectively, and the latter is a subset of the former, as far as an individual is concerned. Now I think that my short description was wrong, I don't intend this reading but I'm a bit lost right now. Words are empirically evident, while some (e.g. see the beginning of the Double R Grammar paper) deny the existence of concepts (for some psychological or psycholinguistic theory of what concepts are). Linguists study languages, which are easier for empirical exploration and systematization than the mind. (And linguists do also use introspection, introspection is an empirical method I think.) Some random comments on the Multinet material: *. Principal requirements for a KRS -- completeness: it is easy to show that a KRS is incomplete, but hard (if possible) to show/prove it is complete. Only some approximation to completeness is required. *. CD theory: Though the idea in very intuitive, it is fundamentally wrong, for two major reasons: (1) Unlike the situation of chemistry, where everything can be analyzed as compound formed by 100+ chemical elements, meaning of a concept/word cannot be reduced into semantic primitives. Though we can use simpler concepts to define complicated ones, such a definition never capture the full meaning of the latter, but can only approximate it to various degrees. This critique is agreed upon and accounted for by Multinet. (2) The meaning of a concept/word is not a constant, but a variable. Of course, any description of it will be constant, but it is just a snapshot taken as a moment, and the semantic theory must allow meaning to change, rather than attempt to specify the real or true meaning, or to converge to such an attractor. The QAS of Multinet should do better at accounting for change, and should be _more than a QAS_ to provide the use-theoretic semantics they are aiming at. The meaning in Multinet can change: a multinet can change, the restrictive knowledge limits the amount of allowed change. *. KB and the world: still depends on Tarskian semantics, with KR aiming at to describe the world as it is. No, Multinet is very critical about model-theoretic (or truth-theoretic) semantics. It has the division into intensional level which doesn't refer to the outside world, and pre-extensional level, which I think can also have internalized reading (the world objects as they are meant). (We need balance of course, it is useful to know the world as it is. I might add, it is also useful to know the world as it should be...) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
Re: [agi] Between logical semantics and linguistic semantics
On Wed, Apr 9, 2008 at 6:03 AM, Stephen Reed [EMAIL PROTECTED] wrote: Thanks for the compliment Lukasz. I am reading your slides and here are my comments: (1) I had seven years experience with the Cyc project. Would you agree that Cyc aspires to be a KRS as you define it? Well, as Hermann Helbig defines it. Yes, Cyc gets its highlight in the Multinet book. (I've only written a bit of text on the slides, most of them, the diagrams, come from the Multinet book.) (2) Sadly, Cyc lacks procedural methods as first class KB objects. Also known as codelets, these are pieces of procedural code that can be fired as the consequent of a rule. Cyc has a facility to do this but the procedures themselves are semantically opaque, being calls into the Cyc runtime engine. In my own work I want to fully represent procedures, using Cyc action scripts as the starting point, so that the system can do things in addition to answering questions. Multinet doesn't help here much: QAS is not an agent; Multinet helps by providing the semantic framework for situations and actions. We need to add the representation of self grounded in that semantics. (4) Conceptual Dependency Theory (CD) - This is somewhat like Cyc in that Doug Lenat is a mathematician and was strongly attracted to symbolic representations that are independent of natural language. The glaring problem with this approach is that coverage of commonsense phenomena is harder without guidance from natural language sources. To illustrate my point, rather than start with an English encyclopedia and represent it entirely, the Cyc project began with some commonsense situations, (e.g. one day in the life of Fred) and represented them from first philosophical/mathematical principles. In my own work, I want to extend the Cyc ontology to cover all the concepts mentioned in the glosses (definitions) of WordNet, and ultimately the propositional content of Wikipedia articles. Well, perhaps Cyc falls short on both fronts: it is too broad to be CD: it represents much more meaning than can be built from CD's-like atoms. But the representation is provided ad-hoc by knowledge ingeneers, it is not grounded in the general net of meaning, which in Multinet is provided by the NL semantics. But perhaps these are just false slogans and Cyc knowledge is dense enough. (5) Sorts and Features - To me these are Cyc-like, except that Cyc made the decision to represent appropriate features as class membership (e.g. the property cyc:mainColorOfObject is a sub-property of cyc:isa / rdf:type). Supposedly, this representation is faster for Cyc deductive inference. On the red petal slide you see that Multinet is flexible about how things are represented (e.g. general rules transform between PROP and ATTR-VAL, similarily rules relate features and their corresponding concept nodes; well I don't know enough about features in Multinet). (6) Knowledge types - Multinet appears more expressive in this respect than the Cyc ontology, although the Cyc KR language CycL allows meta assertions so I believe that MultiNet could be encoded in a Cyc KB. I believe Cyc is well worked-out by the many man-years of development. (I still need to get my hands dirty.) (7) Conceptual capsule - interesting, Cyc has the supporting assertions but not the notion of what assertions uniquely define an object. This seems important to the object/concept-centeredness and intensionality of Multinet. (8) How does Multinet address connectionism or probabilistic inference (e.g. Bayes)? Did I miss where a probability may be associated with an assertion, or with an argument place in an assertion? The book only mentions that the underlying logic should have levels of truthworthiness. Multinet doesn't represent probabilities because neither does language: Multinet has modal modifiers, e.g. (very) probable, (very) unlikely etc. and intensional generalized quantifiers, like some, most. (8) lexicon - need more examples for me to comment. I would be interested in your comments on my adoption of Fluid Construction Grammar as a solution to the NL to semantics mapping problem. I'll try to find time for some more in-depth comments here. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
Re: [agi] Between logical semantics and linguistic semantics
Steve, I'm just on the 7th page of the Double R Grammar paper so I'm rushing ideas here, but it is interesting to see how Multinet, while taking roots in Conceptual Dependency Theory / Case Grammar, and taking the concepts it talks about as mental realities, lands quite close to the philosophy of Double R Grammar (which defines itself in opposition to the above) by insisting on grounding concepts in lexicon. On Wed, Apr 9, 2008 at 4:02 PM, Lukasz Stafiniak [EMAIL PROTECTED] wrote: On Wed, Apr 9, 2008 at 6:03 AM, Stephen Reed [EMAIL PROTECTED] wrote: (4) Conceptual Dependency Theory (CD) - This is somewhat like Cyc in that Doug Lenat is a mathematician and was strongly attracted to symbolic representations that are independent of natural language. The glaring problem with this approach is that coverage of commonsense phenomena is harder without guidance from natural language sources. To illustrate my point, rather than start with an English encyclopedia and represent it entirely, the Cyc project began with some commonsense situations, (e.g. one day in the life of Fred) and represented them from first philosophical/mathematical principles. In my own work, I want to extend the Cyc ontology to cover all the concepts mentioned in the glosses (definitions) of WordNet, and ultimately the propositional content of Wikipedia articles. Well, perhaps Cyc falls short on both fronts: it is too broad to be CD: it represents much more meaning than can be built from CD's-like atoms. But the representation is provided ad-hoc by knowledge ingeneers, it is not grounded in the general net of meaning, which in Multinet is provided by the NL semantics. But perhaps these are just false slogans and Cyc knowledge is dense enough. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
[agi] Between logical semantics and linguistic semantics
I have recently polished my copy-and-paste slides on Multinet: http://www.ii.uni.wroc.pl/~lukstafi/pmwiki/uploads/AGI/Multinet.pdf Pei Wang also gives an interesting chapter about semantics in AGI-Curriculum. By logical semantics I mean the meaning of the contents of mind, and by linguistic semantics the meaning of the contents of communication. What AGI-importance do you assign to capturing the semantics of natural language? (And NL-semantics' impact on logical semantics, as opposed to letting the computer build the representation for itself, out of some elementary thought mechanics.) P.S. Thanks to Pei Wang for the interesting curriculum and to Stephen Reed for the great work on Texai. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
Re: [agi] Instead of an AGI Textbook
On Fri, Mar 28, 2008 at 9:29 PM, Robert Wensman [EMAIL PROTECTED] wrote: A few things come to my mind: 1. To what extent is learning and reasoning a sub topic of cognitive architectures? Is learning and reasoning a plugin to a cognitive architecture, or is in fact the whole cognitive architecture about learning and reasoning. If cognitive architectures department of AGI research is to be usefully delineated, then these are not its subtopics. But neither they are plug-ins. It is in this chapter I introduce you to the overall structure of the system. From other chapters you know that... 2. I would like a special topic on AGI goal representation. More specifically, a topic that discusses how a goal specified by any human designer, can be related to the world model and actions that an AGI system creates? For example, how can the human specified goal, be related to a knowledge representation that is constantly developed by the AGI system? Yes, more work needed on lifelong goal structures, Pollock's master plans, integration with motivational system (which in the primitive form is spreading activation). 3. Why do AI/AGI researchers always talk about knowledge representation. It gives such a strong bias towards static or useless knowledge bases. Why not talk more about World modelling. Because of the more active meaning of the word modelling as opposed to representation, it implies that things such as inference etc. need to be considered. Since the word modelling is also used to denote the process of creating a model, it also implies that we need mechanisms for learning. I really think we should consider if not knowledge representation is a concept straightly borrowed from dumb-narrow AI, or if it really is a key concept for AGI. Sure enough, there will always be knowledge representation, but the question is whether it is an important/relevant/sufficient/misleading concept for AGI. Agreed. I think that knowledge representation label should not be abandoned, but should be grown towards how the system accomodates the sophisticated semantics of natural language and/or its formative domain where formative domain can be social environment, programmistic environment etc. 4. In fact. I would suggest that AGI researchers start to distinguish themselves from narrow AGI by replacing the over ambiguous concepts from AI, one by one. For example: I neither agree nor disagree with your suggestion, I just thank for clarifying your ideas here considerably :-) --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=98558129-0bdb63 Powered by Listbox: http://www.listbox.com
Re: [agi] would anyone want to use a commonsense KB?
On Feb 19, 2008 2:41 PM, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: I think resolution theorem proving provides a way to answer yes/no queries in a KB. I take it as a starting point, and try to think of ways to speed it up and to expand its abilities (answering what/where/when/who/how queries). Oh my, resolution answers wh-questions as well as decision questions in FOL. You just record the answer substitution. (BTW, Prolog is a positive resolution.) We need to be more technical here. --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: [agi] would anyone want to use a commonsense KB?
On Feb 17, 2008 2:11 PM, Russell Wallace [EMAIL PROTECTED] wrote: On Feb 17, 2008 9:56 AM, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: I'm planning to collect commonsense knowledge into a large KB, in the form of first-order logic, probably very close to CycL. Before you embark on such a project, it might be worth first looking closely at the question of why Cyc hasn't been useful, so that you don't end up making the same mistakes. This is perhaps a good opportunity to poll you on why do you think Cyc KB hasn't been useful / successful, I'm interested in grounded opinions (Stephen?), and not about Cyc as an AGI but about Cyc KB as what it was supposed to be (e.g. a universal backbone so that expert systems didn't fall off the knowledge cliff). --- agi Archives: http://www.listbox.com/member/archive/303/=now RSS Feed: http://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: http://www.listbox.com/member/?member_id=8660244id_secret=95818715-a78a9b Powered by Listbox: http://www.listbox.com
Re: Singularity Outcomes [WAS Re: [agi] OpenMind, MindPixel founders both commit suicide
On Jan 29, 2008 12:35 AM, Matt Mahoney [EMAIL PROTECTED] wrote: --- Vladimir Nesov [EMAIL PROTECTED] wrote: Exactly. That's why it can't hack provably correct programs. Which is useless because you can't write provably correct programs that aren't extremely simple. *All* nontrivial properties of programs are undecidable. http://en.wikipedia.org/wiki/Rice%27s_theorem This is false. You can write nontrivial programs for which you can prove nontrivial properties. Rice's theorem tells that you cannot prove nontrivial properties for programs written in Turing-complete languages and given unbounded resources and handed to you by an adversary. And good luck translating human goals expressed in ambiguous and incomplete natural language into provably correct formal specifications. This is true. - 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=8660244id_secret=90871958-149830
Re: [agi] Logical Satisfiability
On Jan 15, 2008 10:49 PM, Jim Bromer [EMAIL PROTECTED] wrote: At any rate, I should have a better idea if the idea will work or not by the end of the year. Lucky you, last time I proved P=NP it only lasted two days ;-) Some resources for people caught by this off-topic thread: - old year's celebrity: PCP Theorem by Gap Amplification http://www.cs.huji.ac.il/~dinuri/mypapers/combpcp.pdf - Introduction to Complexity Theory (Lecture Notes): http://www.wisdom.weizmann.ac.il/~oded/cc-sum.html - complexities do collapse at times (L=SL, 2004): Undirected ST-connectivity in Log-Space http://www.wisdom.weizmann.ac.il/~reingold/publications/sl.ps - 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=8660244id_secret=86291535-6f34ee
Re: [agi] AGI and Deity
Under this thread, I'd like to bring your attention to Serial Experiments: Lain, an interesting pre-Matrix (1998) anime. - 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=8660244id_secret=75885762-854b15
[agi] Case-Based Reasoning
Perhaps you could find this interesting: Case-Based Approximate Reasoning Hüllermeier, Eyke On May 14, 2007 2:50 PM, Mark Waser [EMAIL PROTECTED] wrote: Also anything you can find on case-based reasoning, tho it is woefully rare. Having done a lot of case-based reasoning almost 23 years ago . . . . Case-based reasoning is effectively analogous to weighted nearest neighbor in multi-dimensional space. If you (or the system) can define the dimensions and scale and weight them, it's an awesome method -- this is equivalent to the logic-based/expert-system approach to CBR. The other alternative, which most people don't realize is exactly equivalent to CBR, is to just use neural networks (since they just effectively map the multi-d space -- complete with scaling and weighting). Having used both methods, I would say that, until they both scale themselves fairly quickly into oblivion, the neural network method is more accurate while CBR provides much better explanations. The unfortunate thing is that as you add more and more dimensions, both methods falter pretty quickly. - Original Message - From: J Storrs Hall, PhD [EMAIL PROTECTED] To: agi@v2.listbox.com Sent: Monday, May 14, 2007 7:51 AM Subject: Re: [agi] Tommy On Saturday 12 May 2007 10:24:03 pm Lukasz Stafiniak wrote: Do you have some interesting links about imitation? I've found these, not all of them interesting, I'm just showing what I have: Thanks -- some of those look interesting. I don't have any good links, but I'd reccomend Hurley Chater, eds, Perspectives on Imitation (in 2 vols). Also anything you can find on case-based reasoning, tho it is woefully rare. Josh - 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/?; 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/?; - 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=8660244id_secret=74317004-ffbd21
Re: [agi] Funding AGI research
it reminds me that old joke about three kinds of mathematicians ;-) On Nov 19, 2007 5:25 AM, Benjamin Goertzel [EMAIL PROTECTED] wrote: On Nov 18, 2007 11:24 PM, Benjamin Goertzel [EMAIL PROTECTED] wrote: There are a lot of worthwhile points in your post, and a number of things I don't fully agree with, but I don't have time to argue them all right now... Instead I'll just pick two points: er, looks like that was three ;-) 1) The Babbages and Leibnizes of a given historical period are often visible only in HINDSIGHT. You can't say that there are no Babbages or Leibnizes of AGI around right now ... there could be some on this very list, unrecognized by you, but who will be recognized by all a few decades from now... 2) I don't think it's true that Babbage's or Leibniz's machines were specced out so much better than, say, Novamente. Relative to the technology of their time, plenty of details were left unspecified -- it just seems obvious to us now, in hindsight, how to fill in those details. It wasn't obvious to all their contemporaries. And while, in hindsight, the workability of their machines seems obvious to us, to their contemporaries it must have seemed like the workability of their machines required a huge leap of intuition. They had no rigorous mathematical proof of the workability of their machines, nor did they have working prototypes. They had conceptual arguments that pushed the boundaries of the science of their times, and seemed like nonsense to many of their contemporaries. 3) I don't agree that AGI is primarily a computer science problem, any more than, say, building a car is primarily a metalworking problem. AGI requires computer science problems to be solved as part of its solution; but IMO the essence of AGI-creation is not computer science. This seems to be a genuine difference of scientific intuition btw the two of us. Plenty of others whom I respect appear to share the same opinion as you. -- Ben G On Nov 18, 2007 11:04 PM, J. Andrew Rogers [EMAIL PROTECTED] wrote: On Nov 18, 2007, at 7:06 PM, Benjamin Goertzel wrote: Navigating complex social and business situations requires a quite different set of capabilities than creating AGI. Potentially they could be combined in the same person, but one certainly can't assume that would be the case. I completely agree. But if we are to assume that AGI requires some respectable amount of funding, as seems to be posited by many people, then it seems that it will require a person with broader skills than the stereotypical computer science nerd. In that case, maybe AGI is not accessible to someone who is unwilling or unable to be anything but a computer science nerd. As if the pool of viable AGI researchers was not small enough already. And, I don't think it's fair to say that if you're smart enough to solve AGI, you should be able to quickly make a pile of money doing some kind of more marketable technical-computer-science, and fund the AGI yourself. This assumes a lot of things, for instance that AGI is the same sort of problem as technical-computer-science problems, so that if someone can do AGI better than others, they must be able to do technical- computer-science better than others too. But I actually don't think this is true; I think that AGI demands a different sort of thinking. I'm not so sure about this. All hard problems seem to receive similar sentiments until they are actually solved. I do think that AGI is a relatively hard problem even among the hard problems, but there are other computer science problems that had thousands of pages of literature devoted to them without much progress that when they were solved by someone turned out to be relatively simple. That 20/20 hindsight thing. To the extent that there is any special sauce in AGI, I expect it will look like one of these cases. Solving computer science problems is a pretty general skill, in part because it is a pretty shallow field in most important respects. To use AI research as an example, it is composed of only a handful of fundamental ideas from which a myriad of derivatives and mashups have been created. Most other problems in computer science have the same feature, and when problems get solved it is because someone looked at the handful of fundamentals and ignored the vast bodies of derivative products which add nothing new. Vast quantities of research does not equate to a significant quantity of ideas. AI is a little more complex than some other topics, but is still far simpler at the level of fundamentals than some people make it out to be. People are incapable of solving AGI for the same reason they are incapable of solving any of the other interesting computer
Re: [agi] What best evidence for fast AI?
I think that there are two basic directions to better the Novamente architecture: the one Mark talks about more integration of MOSES with PLN and RL theory On 11/13/07, Edward W. Porter [EMAIL PROTECTED] wrote: Response to Mark Waser Mon 11/12/2007 2:42 PM post. MARK Remember that the brain is *massively* parallel. Novamente and any other linear (or minorly-parallel) system is *not* going to work in the same fashion as the brain. Novamente can be parallelized to some degree but *not* to anywhere near the same degree as the brain. I love your speculation and agree with it -- but it doesn't match near-term reality. We aren't going to have brain-equivalent parallelism anytime in the near future. ED I think in five to ten years there could be computers capable of providing every bit as much parallelism as the brain at prices that will allow thousands or hundreds of thousands of them to be sold. But it is not going to happen overnight. Until then the lack of brain level hardware is going to limit AGI. But there are still a lot of high value system that could be built on say $100K to $10M of hardware. You claim we really need experience with computing and controlling activation over large atom tables. I would argue that obtaining such experience should be a top priority for government funders. MARK The node/link architecture is very generic and can be used for virtually anything. There is no rational way to attack it. It is, I believe, going to be the foundation for any system since any system can easily be translated into it. Attacking the node/link architecture is like attacking assembly language or machine code. Now -- are you going to write your AGI in assembly language? If you're still at the level of arguing node/link, we're not communicating well. ED nodes and links are what patterns are made of, and each static pattern can have an identifying node associated with it as well as the nodes and links representing its sub-patterns, elements, the compositions of which it is part, it associations, etc. The system automatically organize patterns into a gen/comp hierarchy. So, I am not just dealing at a node and link level, but they are the basic building blocks. MARK ... I *AM* saying that the necessity of using probabilistic reasoning for day-to-day decision-making is vastly over-rated and has been a horrendous side-road for many/most projects because they are attempting to do it in situations where it is NOT appropriate. The increased, almost ubiquitous adaptation of probabilistic methods is the herd mentality in action (not to mention the fact that it is directly orthogonal to work thirty years older). Most of the time, most projects are using probabilistic methods to calculate a tenth place decimal of a truth value when their data isn't even sufficient for one. If you've got a heavy-duty discovery system, probabilistic methods are ideal. If you're trying to derive probabilities from a small number of English statements (like this raven is white and most ravens are black), you're seriously on the wrong track. If you go on and on about how humans don't understand Bayesian reasoning, you're both correct and clueless in not recognizing that your very statement points out how little Bayesian reasoning has to do with most general intelligence. Note, however, that I *do* believe that probabilistic methods *are* going to be critically important for activation for attention, etc. ED I agree that many approaches accord too much importance to the numerical accuracy and Bayesian purity of their approach, and not enough importance on the justification for the Bayesian formulations they use. I know of one case where I suggested using information that would almost certainly have improved a perception process and the suggestion was refused because it would not fit within the system's probabilistic framework. At an AAAI conference in 1997 I talked to a programmer for a big defense contractor who said he as a fan of fuzzy logic system; that they were so much more simple to get up an running because you didn't have to worry about probabilistic purity. He said his group that used fuzzy logic was getting things out the door that worked faster than the more probability limited competition. So obviously there is something to say for not letting probabilistic purity get in the way of more reasonable approaches. But I still think probabilities are darn important. Even your this raven is white and most ravens are black example involves notions of probability. We attribute probabilities to such statements based on experience with the source of such statements or similar sources of information, and the concept most is a probabilistic one. The reason we humans are so good at reasoning from small data is based on our ability to estimate rough probabilities from similar or generic patterns. MARK The
Re: [agi] What best evidence for fast AI?
On Nov 14, 2007 3:48 PM, Edward W. Porter [EMAIL PROTECTED] wrote: Lukasz, Which of the multiple issues that Mark listed is one of the two basic directions you were referring to. Ed Porter (First of all, I'm sorry for attaching my general remark as a reply: I was writing from a cell-phone which limited navigation.) I think, that it would be a more fleshed-out knowledge representation (but without limiting the representation-building flexibility of Novamente). -Original Message- From: Lukasz Stafiniak [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 14, 2007 9:15 AM To: agi@v2.listbox.com Subject: Re: [agi] What best evidence for fast AI? I think that there are two basic directions to better the Novamente architecture: the one Mark talks about more integration of MOSES with PLN and RL theory - 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=8660244id_secret=64970556-f74c23
Re: [agi] What best evidence for fast AI?
On Nov 12, 2007 10:34 PM, Linas Vepstas [EMAIL PROTECTED] wrote: I can easily imagine that next-years grand challenge, or the one thereafter, will explicitly require ability to deal with cyclists, motorcyclists, pedestrians, children and dogs. Exactly how they'd test this, however, I don't know ... DARPA seems to be winding up the car challenges :-( (anyone knows anything to the contrary?) - 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=8660244id_secret=64332374-2a763e
Re: AIXItl; Wolfram's hypothesis (was Re: [agi] How valuable is Solmononoff Induction for real world AGI?)
On Nov 10, 2007 4:47 PM, Tim Freeman [EMAIL PROTECTED] wrote: From: Lukasz Stafiniak [EMAIL PROTECTED] The programs are generally required to exactly match in AIXI (but not in AIXItl I think). I'm pretty sure AIXItl wants an exact match too. There isn't anything there that lets the theoretical AI guess probability distributions and then get scored based on how probable the actual world is according to that distribution -- each hypothesis is either right or wrong, and wrong hypotheses are discarded. I agree that I misinterpreted the meaning of exact match. AIXItl uses strategies whose outputs do not need to agree with history. - 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=8660244id_secret=63846012-5d1170
[agi] Re: Solomonoff Machines – up close and personal
On Nov 10, 2007 11:42 PM, Edward W. Porter [EMAIL PROTECTED] wrote: You say there is no magic in AIXI. Is it just make believe Let X be the best way to solve problems. Use X, or does it say something of value to those like me who want to see real AGI's built? Some observations that come to me from reading Marcus Hutter, you can see their worth: (1) he gives rates of convergence for a posterior distribution to a real distribution, assuming that the real distribution has a non-zero prior probability (2) he shows that a weighted (=mean taking) predictor converges possibly exponentially faster than maximum likelihood predictor (3) he shows that the expectimax algorithm is optimal given unbound resources = you can view things functionally: best possible policy (program, strategy), or iteratively (expectimax) (4) he discusses the issue of choosing horizon, reinforcement learning (=RL) work usually uses geometric discounting, Hutter shows that it gives (ln 0.5 / ln d) effective horizon (where d is the discount rate), (and it would be theoretically justified when the agent has probability d of surviving to next cycle) (5) he discusses RL from a general stance, e.g. classes of environments, application to learning frameworks more specific than RL (supervised learning, optimization) (but only theoretically) (6) he discusses the issues of dividing computation resources into using currently best strategy and searching for new strategy (in his time-optimal algorithm for all well-specified problems) (7) his computational AIXItl model, assuming that it's the best out there since Marcus didn't come with something better :-), justifies some practical approaches of letting the competing policies estimate their expected utility: AIXItl only allows policies for which it can find a proof that the policy doesn't overestimate its utility, Eric Baum's market economy uses the policies' claims as a currency (cheating policies go bankrupt), accuracy-based Michigan-style learning classifier systems XCS use (evolutionary) selection pressure on the accuracy of the policies' claims (8) the Kolmogorov-complexity-inspired distribution over programs is related to new better than genetic programming approaches (Schmidhuber's OOPS, MOSES) (but perhaps only distantly) - 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=8660244id_secret=63908199-d1781a
Re: [agi] How valuable is Solmononoff Induction for real world AGI?
On Nov 9, 2007 5:26 AM, Edward W. Porter [EMAIL PROTECTED] wrote: ED ## what is the value or advantage of conditional complexities relative to conditional probabilities? Kolmogorov complexity is universal. For probabilities, you need to specify the probability space and initial distribution over this space. ED ## What's a TM? (Turing Machine, or a code for a universal Turing Machine = a program...) Also are you saying that the system would develop programs for matching patterns, and then patterns for modifying those patterns, etc, So that similar patterns would be matched by programs that called a routine for a common pattern, but then other patterns to modify them to fit different perceptions? Yes, these programs will be compact description od data when enough data gets collected, so their (posterior) probability will grow with time. But the most probable programs will be very cryptic, without redundancy to make the structure evident. So are the programs just used for computing Kolmogorov complexity or are they also used for generating and matching patterns. It is difficult to say: in AIXI, the direct operation is governed by the expectimax algorithm, but the algorithm works in future (is derived from the Solomonoff predictor). Hutter mentions alternative model AIXI_alt, which models actions the same way as the environment... Does it require that the programs exactly match a current pattern being received, or does it know when a match is good enough that it can be relied upon as having some significance? It is automatic: when you have a program with a good enough match, then you can parameterize it over the difference and apply twice, thus saving the code. Remember that the programs need to represent the whole history. Can the programs learn that similar but different patterns are different views of the same thing? Can they learn a generalizational and compositional hierarchy of patterns? With an egzegetic enough interpretation... I will comment on further questions in a few hours. - 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=8660244id_secret=63453551-e3704c
Re: [agi] How valuable is Solmononoff Induction for real world AGI?
On Nov 9, 2007 5:26 AM, Edward W. Porter [EMAIL PROTECTED] wrote: So are the programs just used for computing Kolmogorov complexity or are they also used for generating and matching patterns. The programs do not compute K complexity, they (their length) _are_ (a variant of) Kolmogorov complexity. The programs compute (predict) the environment. Does it require that the programs exactly match a current pattern being received, or does it know when a match is good enough that it can be relied upon as having some significance? The programs are generally required to exactly match in AIXI (but not in AIXItl I think). But the significance is provided by the compression on representation of similar things, which favors the same sort of similarity in the future. Can they run on massively parallel processing. I think they can... In AIXI, you would build a summation tree for the posterior probability. The Hutters expectimax tree appears to alternate levels of selection and evaluation. Can the Expectimax tree run in reverse and in parallel, with information coming up from low sensory levels, and then being selected based on their relative probability, and then having the selected lower level patterns being fed as inputs into higher level patterns and then repeating that process. That would be a hierarchy that alternates matching and then selecting the best scoring match at alternate levels of the hierarchy as is shown in the Serre article I have cited so many times before on this list. To be optimal, the expectimax must be performed chronologically from the end of the horizon (dynamic programming principle: close to the end of the time horizon, you have smaller planning problems -- less opportunities; from smaller solutions to smaller problems you build bigger solutions backwards in time). But the probabilities are conditional on all current history including low sensory levels. (Generally, your comment above doesn't make much sense in the AIXI context.) ED## are these short codes sort of like Wolfram little codelettes, that can hopefully represent complex patterns out of very little code, or do they pretty much represent subsets of visual patterns as small bit maps. It depends on reality, whether the reality supports Wolfram's hypothesis. Best Regards. - 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=8660244id_secret=63539823-b308a9
Re: [agi] How valuable is Solmononoff Induction for real world AGI?
On 11/8/07, Edward W. Porter [EMAIL PROTECTED] wrote: HOW VALUABLE IS SOLMONONOFF INDUCTION FOR REAL WORLD AGI? I will use the opportunity to advertise my equation extraction of the Marcus Hutter UAI book. And there is a section at the end about Juergen Schmidhuber's ideas, from the older AGI'06 book. (Sorry biblio not generated yet.) http://www.ii.uni.wroc.pl/~lukstafi/pmwiki/uploads/AGI/UAI.pdf - 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=8660244id_secret=63153472-64b600
Re: [agi] The Grounding of Maths
What you describe is not a visualization, but the silent inner speech. http://en.wikipedia.org/wiki/Lev_Vygotsky#Thinking_and_Speaking On 10/12/07, a [EMAIL PROTECTED] wrote: Benjamin Goertzel wrote: Well, it's hard to put into words what I do in my head when I do mathematics... it probably does use visual cortex in some way, but's not visually manipulating mathematical expressions nor using visual metaphors... I can completely describe. I completely do mathematics by visually manipulating and visually replacing symbols with other symbols. I also do mathematical reasoning and theorem proving with that. Mathematicians commonly have high visuospatial intelligence, that's why they have high IQs. - 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=8660244id_secret=53098509-58aae0
Re: [agi] The Grounding of Maths
For those interested in higher dimensions, I've just grabbed a link from wikipedia: * Christopher E. Heil, A basis theory primer, 1997. http://www.math.gatech.edu/~heil/papers/bases.pdf Well, a mathematician needs to _understand_ (as opposed to what I would call a knowledge base - inference engine disconnect), and visualization is a metaphor for understanding, not the understanding itself. What visualization actually often means, is an _unsound reduction_ of a general sophisticated notion to a simple model, without ever realizing that the model contradicts the notion, but instead supplementing this heuristical model with additional mental discipline at rough corners. On 10/12/07, Eliezer S. Yudkowsky [EMAIL PROTECTED] wrote: Benjamin Goertzel wrote: Well ... going beyond imaginary numbers... how do *you* do mathematics in quaternionic and octonionic algebras? Via visualization? Personally, I can sorta visualize 4D, but I I suck at visualizing 8-dimensional space, so I tend to reason more abstractly when thinking about such things... Just visualize it in N-dimensional space, then let N go to 8. - 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=8660244id_secret=53080937-a6329b
Re: Self-improvement is not a special case (was Re: [agi] Religion-free technical content)
On 10/12/07, Eliezer S. Yudkowsky [EMAIL PROTECTED] wrote: some of us are much impressed by it. Anyone with even a surface grasp of the basic concept on a math level will realize that there's no difference between self-modifying and writing an outside copy of yourself, but *either one* involves the sort of issues I've been calling reflective. Well, this could be at least a definition of self-modifying ;-) - 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=8660244id_secret=53163374-01d6ba
[agi] Learning Classifier Systems vs. evolutionary economy systems by Eric Baum
Hi, Has anyone done in-depth (i.e. experimental or theoretical) comparison of accuracy-based LCSs (XCS) and Eric Baum's economy? Eric only mentions superiority over ZCS. But XCS are closer to Eric's systems, fitness of rules is based on their prediction of reward (compare to making bids). I wonder if economics could be explained by RL theory. - 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=8660244id_secret=51379467-c3f951
Re: [agi] Do the inference rules of categorical logic make sense?
When looking at it through a crisp glass, the relation is a preorder, not a (partial) order. And priming is essential. For example, in certain contexts, we think that an animal is a human (anthropomorphism). On 10/9/07, Mark Waser [EMAIL PROTECTED] wrote: Ack! Let me rephrase. Despite the fact that Pei always uses the words of inheritance (and is technically correct), what he means is quite different from what most people assume that he means. You are stuck on the common meanings of the terms is an ancestor of and is a descendant of and it's impeding your understanding. - 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=8660244id_secret=51437008-630e6a
Re: [agi] Do the inference rules of categorical logic make sense?
Major premise and minor premise in a syllogism are not interchangeable. Read the derivation of truth tables for abduction and induction from the semantics of NAL to learn that different ordering of premises results in different truth values. Thus while both orderings are applicable, one will usually give more confident result which will dominate the other. On 10/6/07, Edward W. Porter [EMAIL PROTECTED] wrote: But I don't understand the rules for induction and abduction which are as following: ABDUCTION INFERENCE RULE: Given S -- M and P -- M, this implies S -- P to some degree INDUCTION INFERENCE RULE: Given M -- S and M -- P, this implies S -- P to some degree The problem I have is that in both the abduction and induction rule -- unlike in the deduction rule -- the roles of S and P appear to be semantically identical, i.e., they could be switched in the two premises with no apparent change in meaning, and yet in the conclusion switching S and P would change in meaning. Thus, it appears that from premises which appear to make no distinctions between S and P a conclusion is drawn that does make such a distinction. At least to me, with my current limited knowledge of the subject, this seems illogical. - 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=8660244id_secret=50749379-2a7926
[agi] Highlights: Ute Schmid, Inductive Synthesis of Functional Programs
Ute Schmid publications: http://www.cogsys.wiai.uni-bamberg.de/schmid/publications.html About this book Because of its promise to support human programmers in developing correct and efficient program code and in reasoning about programs, automatic program synthesis has attracted the attention of researchers and professionals since the 1970s. This book focusses on inductive program synthesis, and especially on the induction of recursive functions; it is organized into three parts on planning, inductive program synthesis, and analogical problem solving and learning. Besides methodological issues in inductive program synthesis, emphasis is placed on its applications to control rule learning for planning. Furthermore, relations to problem solving and learning in cognitive psychology are discussed. http://www.springer.com/west/home?SGWID=4-102-22-6954766-0changeHeader=truereferer=www.springeronline.comSHORTCUT=www.springer.com/3-540-40174-1 - 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=8660244id_secret=48074982-038f88
Re: [agi] Re: HTM vs. IHDR
On 6/29/07, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: I've talked to John Weng many times before, and I found that his AGI has some problems but he wasn't very eager to talk about them. For example, it could only recognize pre-trained objects (eg, a certain doll) but not general object classes like dolls, cups or cars. It seems intuitive that bottom-up approach is better at generalization. HTM is much more sophisticated, conditional probabilities, and the learning in context of sequences, must really be helpful. (IHDR can have time-chunking but this is not that useful at categorization.) It seems that the advantages of IHDR are limited to quick learning and one-instance learning (HTM cannot do one-instance learning, which is simple for IHDR). I'm not sure if HTM could learn online. - 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=231415user_secret=e9e40a7e
Re: [agi] Re: HTM vs. IHDR
On 6/29/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: It seems intuitive that bottom-up approach is better at generalization. HTM is much more sophisticated, conditional probabilities, and the learning in context of sequences, must really be helpful. (IHDR can have time-chunking but this is not that useful at categorization.) It seems that the advantages of IHDR are limited to quick learning and one-instance learning (HTM cannot do one-instance learning, which is simple for IHDR). I'm not sure if HTM could learn online. And, IHDR would still be better at quick and one-instance learning than HTM naively augmented to do that. - 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=231415user_secret=e9e40a7e
Re: [agi] Re: HTM vs. IHDR
BTW, has HTM been seriously tried at medical images understanding? - 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=231415user_secret=e9e40a7e
Re: [agi] Reinforcement Learning: An Introduction Richard S. Sutton and Andrew G. Barto
On 6/27/07, Vladimir Nesov [EMAIL PROTECTED] wrote: I think AI books are not particularly helpful, not at first (if you know enough about algorithms, programming and math, generally). AI provides technical answers to well-formulated questions, but with AGI right questions is what's lacking. So, my current reading is The Cambridge Handbook of Thinking and Reasoning. http://www.cambridge.org/uk/catalogue/catalogue.asp?isbn=0521531012 I guess I'll finish reading Bayesian Approach to Imitation in RL before they put that Cambridge book online ;-) - 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=231415user_secret=e9e40a7e
[agi] HTM vs. IHDR
I'm starting to learn about Numenta's HTM, but perhaps someone would like to share in advance: what are the essential differences between HTM and Yuang Weng's IHDR augmented with Observation-driven MDPs? - 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=231415user_secret=e9e40a7e
[agi] Re: HTM vs. IHDR
Ouch, they differ more than I thought... Good :-) (HTM based more on Bayes nets) On 6/24/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: I'm starting to learn about Numenta's HTM, but perhaps someone would like to share in advance: what are the essential differences between HTM and Yuang Weng's IHDR augmented with Observation-driven MDPs? - 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=231415user_secret=e9e40a7e
Re: [agi] Reinforcement Learning: An Introduction Richard S. Sutton and Andrew G. Barto
On 6/24/07, Bob Mottram [EMAIL PROTECTED] wrote: I have one of Richard Sutton's books, and RL methods are useful but I also have some reservations about them. Often in this sort of approach a strict behaviorist position is adopted where the system is simply trying to find an appropriate function mapping inputs to outputs. The internals of the system are usually treated as a black box with a homogenous structure, and it's this zero architecture or trivial architecture approach which can make the learning problem exceptionally hard. But they don't need to be, there is always place to accomodate knowledge. You can use structured value function approximators, use off-policy methods for supervised learning, etc. BTW, has anyone tried value estimates with an uncertainty dimension, and with a prior that favors more certain estimates but degrades smoothly for less certain estimates, for both action selection and value backup, or at least for action selection? This kind of action selection seems to be smarter than e-greedy strategies. (More certain value estimate is less optimistic, smaller, but is based more experience.) - 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=231415user_secret=e9e40a7e
Re: [agi] Re: HTM vs. IHDR
The obvious observation is that HTM is bottom-up and IHDR is top-down. HTM builds hierarchy by merging fixed, topologically-organized, coordinate-system-based subspaces: tilings, where IHDR builds hierarchy by splitting input space by adaptively learned Gaussian features. - 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=231415user_secret=e9e40a7e
Re: [agi] AGI introduction
On 6/22/07, Pei Wang [EMAIL PROTECTED] wrote: I put a brief introduction to AGI at http://nars.wang.googlepages.com/AGI-Intro.htm , including an AGI Overview followed by Representative AGI Projects. I think that hybrid and integrated descriptions are useful, especially when seeing AGI in the broader context of agent systems, but they need to be further elaborated (I posted about TouringMachines hoping to bring that up). For me, now, they seem almost co-extensive. As for the meaning, to me, hybrid means integrated at the level of engineering, and integrative means integrated, (rather by synthesis than dominance), at the conceptual level. - 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=231415user_secret=e9e40a7e
Re: [agi] AGI introduction
On 6/23/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: I think that hybrid and integrated descriptions are useful, especially when seeing AGI in the broader context of agent systems, but they need to be further elaborated (I posted about TouringMachines hoping to bring that up). For me, now, they seem almost co-extensive. As for the meaning, to me, hybrid means integrated at the level of engineering, and integrative means integrated, (rather by synthesis than dominance), at the conceptual level. For example, the RL book shows how to integrate planning and reactive reinforcement at the conceptual level. - 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=231415user_secret=e9e40a7e
Re: [agi] AGI introduction
On 6/22/07, Pei Wang [EMAIL PROTECTED] wrote: Hi, I put a brief introduction to AGI at http://nars.wang.googlepages.com/AGI-Intro.htm , including an AGI Overview followed by Representative AGI Projects. Thanks! As a first note, SAIL seems to me a better replacement for Cog, because SAIL has much generality and some theoretical accomplishment where Cog is (AFAIK) hand-crafted engineering. - 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=231415user_secret=e9e40a7e
[agi] Association for Uncertainty in Artificial Intelligence
Looking through Wikipedia articles I stumbled upon a probably very interesting place: http://www.auai.org/ Association for Uncertainty in Artificial Intelligence - 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=231415user_secret=e9e40a7e
[agi] Reinforcement Learning: An Introduction Richard S. Sutton and Andrew G. Barto
Obligatory reading: http://www.cs.ualberta.ca/~sutton/book/ebook/the-book.html Cheers. - 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=231415user_secret=e9e40a7e
Re: [agi] Reinforcement Learning: An Introduction Richard S. Sutton and Andrew G. Barto
On 6/23/07, Bo Morgan [EMAIL PROTECTED] wrote: Reinforcement learning is a simple theory that only solves problems for which we can design value functions. But it is good for AGI newbies like me to start with :-) - 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=231415user_secret=e9e40a7e
[agi] Autonomous Training
Hello, Have you worked on or thought about autonomous training? An AGI, before engaging into a critical mission, has to prepare herself for that, so she has to learn and simulate the domain of the mission. - 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=231415user_secret=e9e40a7e
Re: [agi] Pure reason is a disease.
On 6/14/07, Matt Mahoney [EMAIL PROTECTED] wrote: I don't believe this addresses the issue of machine pain. Ethics is a complex function which evolves to increase the reproductive success of a society, for example, by banning sexual practices that don't lead to reproduction. Ethics also evolves to ban harm to other members of the group, but not to non-members (e.g. war is allowed), and not to other species (hunting is allowed), except to the extent that such actions would harm the group. There is no precedent for ethics with regard to machines. We protect machines only to the extent that harming them harms the owner. Nevertheless, I think your argument about pain being related to complexity relates to the more general principle of protecting that which resembles a human, even if that resemblance is superficial or based on emotion. I was reminded of this when I was playing Grand Theft Auto III. Besides carjacking, murder, and assorted mayhem, the game allows you to pick up prostitutes. Afterwards, the game gives you the option of getting your money back by beating her to death, but I declined. I felt empathy for a video game character. http://www.goertzel.org/books/spirit/uni3.htm -- VIRTUAL ETHICS - 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=231415user_secret=e9e40a7e
Re: [agi] Pure reason is a disease.
On 6/13/07, Matt Mahoney [EMAIL PROTECTED] wrote: If yes, then how do you define pain in a machine? A pain in a machine is the state in the machine that a person empathizing with the machine would avoid putting the machine into, other things being equal (that is, when there is no higher goal in going through the pain). - 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=231415user_secret=e9e40a7e
Re: [agi] Pure reason is a disease.
On 6/13/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: On 6/13/07, Matt Mahoney [EMAIL PROTECTED] wrote: If yes, then how do you define pain in a machine? A pain in a machine is the state in the machine that a person empathizing with the machine would avoid putting the machine into, other things being equal (that is, when there is no higher goal in going through the pain). To clarify: (1) there exists a person empathizing with that machine (2) this person would avoid putting the machine into the state of pain - 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=231415user_secret=e9e40a7e
Re: [agi] Pure reason is a disease.
On 6/14/07, Matt Mahoney [EMAIL PROTECTED] wrote: I would avoid deleting all the files on my hard disk, but it has nothing to do with pain or empathy. Let us separate the questions of pain and ethics. There are two independent questions. 1. What mental or computational states correspond to pain? 2. When is it ethical to cause a state of pain? There is a gradation: - pain as negative reinforcement - pain as an emotion - pain as a feeling When you ask if something feels pain, then you don't ask if pain is adequate description of some aspect in that thing or person X, but whether X can be attributed as feeling. And this is related to the comlexity of X, and this complexity is related with ethics. - 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=231415user_secret=e9e40a7e
Re: [agi] Symbol Grounding
On 6/12/07, Mark Waser [EMAIL PROTECTED] wrote: a question is whether a software program could tractably learn language without such associations, by relying solely on statistical associations within texts. Isn't there an alternative (or middle ground) of starting the software program with a seed of initial structure and then letting it grow from there (rather than relying only on statistical associations -- which I believe will be intractable for quite some time). It is at least conceivable. The idea is that you give the system reasonable means to build models (= simulations). The initial structure lets the system build approximate models to at least some minimal but not isolated amount of texts. The system then should have some explorative means to build new more complicated models (the hard part). The model extension should be guided by parts of (partially or approxiamtely) interpretable texts that do not quite fit (e.g. have uninterpreted words). The extensions are then evaluated by their predictive characteristics (how much new text can be consistently interpreted in them). Also, have a look at Polyscheme, etc. - 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=231415user_secret=e9e40a7e
Re: [agi] Symbol Grounding
On 6/12/07, Derek Zahn [EMAIL PROTECTED] wrote: Some people, especially those espousing a modular software-engineering type of approach seem to think that a perceptual system basically should spit out a token for chair when it sees a chair, and then a reasoning system can take over to reason about chairs and what you might do with them -- and further it is thought that the reasoning about chairs part is really the essence of intelligence, whereas chair detection is just discardable pre-processing. My personal intuition says that by the time you have taken experience and boiled it down to a token labeled chair you have discarded almost everything important about the experience and all that is left is something that can be used by our logical inference systems. Assume that the inference systems do well. Therefore, not *that* much information is discarded. Therefore, the inference systems have found a workaround to collect the information about a particular chair that is not directly accessible through a single token (e.g by a subtle context of a myriad of other tokens). - 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=231415user_secret=e9e40a7e
Re: [agi] about AGI designers
On 6/6/07, Peter Voss [EMAIL PROTECTED] wrote: 'fraid not. Have to look after our investors' interests… (and, like Ben, I'm not keen for AGI technology to be generally available) But at least Novamente makes a convinceable amount of their ideas available IMHO. P.S. Probabilistic Logic Networks is coming no later than early 2008 I hope? :-) - 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=231415user_secret=e9e40a7e
Re: [agi] Books
I've ended up with the following list. What do you think? * Ming Li and Paul Vitanyi, An Introduction to Kolmogorov Complexity and Its Applications, Springer Verlag 1997 * Marcus Hutter, Universal Artificial Intelligence: Sequential Decisions Based On Algorithmic Probability, Springer Verlag 2004 * Vladimir Vapnik, Statistical Learning Theory, Wiley-Interscience 1998 * Pedro Larrañaga, José A. Lozano (Editors), Estimation of Distribution Algorithms: A New Tool for Evolutionary Computation, Springer 2001 * Ben Goertzel, Cassio Pennachin (Editors), Artificial General Intelligence (Cognitive Technologies), Springer 2007 * Pei Wang, Rigid Flexibility: The Logic of Intelligence, Springer 2006 * Ben Goertzel, Matt Ikle', Izabela Goertzel, Ari Heljakka Probabilistic Logic Networks, in preparation * Juyang Weng et al., SAIL and Dav Developmental Robot Projects: the Developmental Approach to Machine Intelligence, publication list * Ralf Herbrich, Learning Kernel Classifiers: Theory and Algorithms, MIT Press 2001 * Eric Baum, What is Thought?, MIT Press 2004 * Marvin Minsky, The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Simon Schuster 2006 * Ben Goertzel, The Hidden Pattern: A Patternist Philosophy of Mind, Brown Walker Press 2006 * Ronald Brachman, Hector Levesque, Knowledge Representation and Reasoning, Morgan Kaufmann 2004 * Peter Gärdenfors, Conceptual Spaces: The Geometry of Thought, MIT Press 2004 * Wayne D. Gray (Editor), Integrated Models of Cognitive Systems, Oxford University Press 2007 * Logica Universalis, Birkhäuser Basel, January 2007 - 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=231415user_secret=e9e40a7e
Re: [agi] Books
On 6/9/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: I've ended up with the following list. What do you think? I would like to add Locus Solum by Girard to this list, and then is seems to collapse into a black hole... Don't care? * Ming Li and Paul Vitanyi, An Introduction to Kolmogorov Complexity and Its Applications, Springer Verlag 1997 * Marcus Hutter, Universal Artificial Intelligence: Sequential Decisions Based On Algorithmic Probability, Springer Verlag 2004 * Vladimir Vapnik, Statistical Learning Theory, Wiley-Interscience 1998 * Pedro Larrañaga, José A. Lozano (Editors), Estimation of Distribution Algorithms: A New Tool for Evolutionary Computation, Springer 2001 * Ben Goertzel, Cassio Pennachin (Editors), Artificial General Intelligence (Cognitive Technologies), Springer 2007 * Pei Wang, Rigid Flexibility: The Logic of Intelligence, Springer 2006 * Ben Goertzel, Matt Ikle', Izabela Goertzel, Ari Heljakka Probabilistic Logic Networks, in preparation * Juyang Weng et al., SAIL and Dav Developmental Robot Projects: the Developmental Approach to Machine Intelligence, publication list * Ralf Herbrich, Learning Kernel Classifiers: Theory and Algorithms, MIT Press 2001 * Eric Baum, What is Thought?, MIT Press 2004 * Marvin Minsky, The Emotion Machine: Commonsense Thinking, Artificial Intelligence, and the Future of the Human Mind, Simon Schuster 2006 * Ben Goertzel, The Hidden Pattern: A Patternist Philosophy of Mind, Brown Walker Press 2006 * Ronald Brachman, Hector Levesque, Knowledge Representation and Reasoning, Morgan Kaufmann 2004 * Peter Gärdenfors, Conceptual Spaces: The Geometry of Thought, MIT Press 2004 * Wayne D. Gray (Editor), Integrated Models of Cognitive Systems, Oxford University Press 2007 * Logica Universalis, Birkhäuser Basel, January 2007 - 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=231415user_secret=e9e40a7e
Re: [agi] Books
On 6/9/07, YKY (Yan King Yin) [EMAIL PROTECTED] wrote: I'm not aware of any book on pattern recognition with a view on AGI, except The Pattern Recognition Basis of Artificial Intelligence by Don Tveter (1998): http://www.dontveter.com/basisofai/basisofai.html You may look at The Cambridge Hankbook of Thinking and Reasoning first, especially the chapters on similarity and analogy. Thanks, it's interesting. - 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=231415user_secret=e9e40a7e
Re: [agi] AGI Consortium
On 6/8/07, Mark Waser [EMAIL PROTECTED] wrote: You are never going to see a painting by committee that is a great painting. And he's right. This was Sterling's indictment of Wikipedia–and to the wisdom of crowds fad sweeping the Web 2.0 pitch sessions of Silicon Valley–but it's also a fair assessment of what holds most (not all) open source enterprises back: Lack of vision. Every project has some developers recruitment policy; a smart mind is an integrated mind. The ideological divide goes between Open Knowledge and Source, and Closed Knowledge and Source. - 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=231415user_secret=e9e40a7e
Re: [agi] AGI Consortium
On 6/8/07, J Storrs Hall, PhD [EMAIL PROTECTED] wrote: This is basically right. There are plenty of innovative Open Source programs out there, but they are typically some academic's thesis work. Being Open Source can allow them to be turned into solid usable applications, but it can't create them in the first place. Being Closed Source can't create them neither (just a note for the sake of completeness). - 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=231415user_secret=e9e40a7e
[agi] Books
Which books would you recommend? For which there is a better replacement? My results of quick amazon.com browsing: Body Language: Representation in Action (Bradford Books) (Hardcover) by Mark Rowlands (Author) http://www.amazon.com/Body-Language-Representation-Action-Bradford/dp/0262182556/ref=pd_bxgy_b_text_b/104-8541523-0043944?ie=UTF8qid=1181214688sr=1-91 Integrated Models of Cognitive Systems (Cognitive Models and Architectures) (Hardcover) by Wayne D. Gray (Editor) http://www.amazon.com/Integrated-Models-Cognitive-Systems-Architectures/dp/0195189191/ref=sr_1_168/104-8541523-0043944?ie=UTF8s=booksqid=1181215075sr=1-168 Reasoning about Uncertainty (Paperback) by Joseph Y. Halpern http://www.amazon.com/Reasoning-about-Uncertainty-Joseph-Halpern/dp/0262582597/ref=sr_1_132/104-8541523-0043944?ie=UTF8s=booksqid=1181214901sr=1-132 Pattern Recognition, Third Edition (Hardcover) by Sergios Theodoridis (Author), Konstantinos Koutroumbas (Author) http://www.amazon.com/Pattern-Recognition-Third-Sergios-Theodoridis/dp/0123695317/ref=sr_1_110/104-8541523-0043944?ie=UTF8s=booksqid=1181214779sr=1-110 Knowledge Representation and Reasoning (The Morgan Kaufmann Series in Artificial Intelligence) by Ronald Brachman (Author), Hector Levesque (Author) http://www.amazon.com/Knowledge-Representation-Reasoning-Artificial-Intelligence/dp/1558609326/ref=sr_1_47/104-8541523-0043944?ie=UTF8s=booksqid=1181214465sr=1-47 Pattern Recognition and Machine Learning (Information Science and Statistics) (Hardcover) by Christopher M. Bishop (Author) http://www.amazon.com/Pattern-Recognition-Learning-Information-Statistics/dp/0387310738/ref=sr_1_6/104-8541523-0043944?ie=UTF8s=booksqid=1181214349sr=1-6 Learning Kernel Classifiers: Theory and Algorithms (Adaptive Computation and Machine Learning) (Hardcover) by Ralf Herbrich (Author) http://www.amazon.com/Learning-Kernel-Classifiers-Algorithms-Computation/dp/026208306X/ref=sr_1_184/104-8541523-0043944?ie=UTF8s=booksqid=1181214108sr=1-184 - 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=231415user_secret=e9e40a7e
[agi] Re: Books
OK, (1) Which book on pattern recognition is the most AGIsh? (Vapnik comes in his own right) ((2) - (3) as before), (4) When will Probabilistic Logic Networks be out? - 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=231415user_secret=e9e40a7e
Re: [agi] Vectorianism and a2i2
It's a far better answer than I asked for :-) On 6/6/07, J Storrs Hall, PhD [EMAIL PROTECTED] wrote: Norm of our vectors, known of old-- Lord of our far-flung number line Beneath whose measured length we hold Dominion over quad and spline-- Memory trace, be with us yet, Lest we forget - lest we forget! Thy referents to meanings bind; The captains and the kings are nil: Still stand Thine ancient symbols still And theory of the pattern mind. Oh R of N, be with us yet, Lest we forget - lest we forget! Far-called, our transforms melt away; On sine and cosine sinks the fire: Lo, all our linearity Is one with Nineveh and Tyre! Space of all Phases, spare us yet, Lest we forget - lest we forget! If, drunk with powersets, we loos'd Mere discrete symbol sequences-- Such boasting as Eliza used Or lesser breeds of expert sys-- Eigenvector, keep us yet, Lest we forget - lest we forget! For heathen heart that puts its trust In vacuum tube and FET-- All valiant dust that builds on dust, And logic which can't count to 3: For foolish boast and pie in sky, Have mercy on Thine AGI! * * * Oh, wait, you said VECtorian ... I thought you said VICtorian... Never mind! Josh - 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=231415user_secret=e9e40a7e
Re: [agi] Vectorianism and a2i2
On 6/6/07, Benjamin Goertzel [EMAIL PROTECTED] wrote: I believe: The practice of representing knowlege using high-dimensional numerical vectors ;-) I've misspelled from vectorialism, see: Churchland on connectionism http://www-cse.ucsd.edu/users/gary/pubs/laakso-church-chap.pdf (vectorialism as opposed to propositionalism) - 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=231415user_secret=e9e40a7e
Re: [agi] about AGI designers
On 6/6/07, Derek Zahn [EMAIL PROTECTED] wrote: D. There are no consortiums to join. I see talk about joining Novamente, but are they hiring? It might be possible to volunteer to work on peripheral things like AGISIM, but I sort of doubt that Ben is eager to train volunteers on the AGI-type code itself. On average, the cost/benefit of that would probably be quite poor. I see that AdaptiveAI has an opening for a programmer. We don't talk about them much, probably because they have chosen not to make much information availableabout what they're up to, beyond Peter Voss's vague overview paper. I think, that with understanding of what major projects are up to, a new startup should aim into complementary space (and to interoperate at some stage). Otherwise, I would insist on joining an existing project, unless they really are over the threshold with manpower. - 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=231415user_secret=e9e40a7e
Re: [agi] about AGI designers
On 6/7/07, Benjamin Goertzel [EMAIL PROTECTED] wrote: Sure, but the nature of AGI is that wizzy demos are likely to come fairly late in the development process. All of us actively working in the field understand this What about LIDA? Even if she is not very general she is more cognitive than Numenta, and has some nice HALish activity in the wild. :-) And for LIDA, I guess she is already quite mature by development (filogeny). - 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=231415user_secret=e9e40a7e
Re: [agi] analogy, blending, and creativity
On 6/2/07, J Storrs Hall, PhD [EMAIL PROTECTED] wrote: And many scientists refer to potential energy surfaces and the like. There's a core of enormous representational capability with quite a few well-developed intellectual tools. Another Grand Unification theory: Estimation of Distribution Algorithms behind Bayesian Nets, Genetic Programming and unsupervised Neural Networks. - 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=231415user_secret=e9e40a7e
[agi] Re: PolyContextural Logics
One more bite: Locus Solum: From the rules of logic to the logic of rules by Jean-Yves Girard, 2000. http://lambda-the-ultimate.org/node/1994 On 6/5/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: Speaking of logical approaches to AGI... :-) http://www.thinkartlab.com/pkl/ - 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=231415user_secret=e9e40a7e
Re: [agi] analogy, blending, and creativity
On 5/17/07, J Storrs Hall, PhD [EMAIL PROTECTED] wrote: On Wednesday 16 May 2007 04:47:53 pm Mike Tintner wrote: Josh . If you'd read the archives, you'd see that I've advocated constructive solid geometry in Hilbert spaces as the basic representational primitive. Would you like to say more re your representational primitives? Sounds interesting. The archives have no reference to constructive solid geometry in Hilbert spaces in any form. Personally, I think it's a plot. MOOO ha ha ha! It's all in your mind :-) Actually, I can't find it either but (and this is apropos to the subject) we rarely remember the exact words we said or heard; we remember more abstract representations. Chances I used CSG and/or vector spaces. Hilbert space is a rhetorical flourish anyway -- they may need it to describe quantum mechanics precisely but we'll never implement it... Many engineering departments make the mistake of never mentioning the term Hilbert space and calling it all signal analysis. - 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=231415user_secret=e9e40a7e
Re: [agi] Is the Digital Processor Really Analog?
In the name of Church-Turing-von Neumann, don't follow that heresy. Quantum computers are kind-of Heglian synthesis of analog-digital. There are quirks going on inside computers, like error correction on memory retrieval, not to mess up with your (or the computer user) symbols. If you read analog as generating its own symbols, then computers are not meant to do so. [ just a bit of loose crap from me ;-) ] - 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=231415user_secret=e9e40a7e
Re: Slavery (was Re: [agi] Opensource Business Model)
On 6/2/07, Mark Waser [EMAIL PROTECTED] wrote: By some measures Google is more intelligent than any human. Should it have human rights? If not, then what measure should be used as criteria? Google is not conscious. It does not need rights. Sufficiently complex consciousness (or even just the appearance of such) is the criteria that should be used. Google has its rights. No crazy totalitarian government tells Google what to do. - 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=231415user_secret=e9e40a7e
Re: Slavery (was Re: [agi] Opensource Business Model)
On 6/2/07, Lukasz Stafiniak [EMAIL PROTECTED] wrote: Google has its rights. No crazy totalitarian government tells Google what to do. (perhaps it should go: Google struggles for its rights, sometimes making moral compromises) - 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=231415user_secret=e9e40a7e
Re: [agi] Open AGI Consortium
On 6/2/07, Derek Zahn [EMAIL PROTECTED] wrote: For a for-profit AGI project I suggest the following definition of intelligence: The ability to create information-based objects of economic value. What about: The ability to create information-based objects generating income. This is less ambiguous and more demanding. - 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=231415user_secret=e9e40a7e
Re: [agi] Donations to support open-source AGI research (SIAI Research Program)
On 6/2/07, Benjamin Goertzel [EMAIL PROTECTED] wrote: http://www.singinst.org/research/summary [see menu to the left] embodying some research I think will be extremely valuable for pushing forward toward AGI, and that I think is well- pursued in an open, nonprofit context. * Research Area 5: Design and Creation of Safe Software Infrastructure has enough support in the mainstream (industry and academia) IMHO. Microsoft Research works on it; several little companies are in this bussiness; every CompSci university department has someone working on it. Feeling lucky with Google gives: http://www.cs.utexas.edu/users/moore/publications/zurich-05/talk.html Citing: However, currently, there is no programming language that both supports proof-based program correctness checking, and is sufficiently efficient in terms of execution to be usable for pragmatic AGI purposes. This is not how I would state things. The burden is of course efficiency of program development (proving program properties or deriving program from specification) which is not much correlated (if at all) with efficiency of the extracted/verified program. For example, you can prove properties of assembly code (like in dependently typed assembly languages). - 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=231415user_secret=e9e40a7e