EFTA00775409.pdf

From: on behalf of Ben Goertzel <Milla>1 I TO:Jeffreyfipst6ncjeevaeationftynailx:orn> Subject: funding sought for Human-Level Al Roadmapping workshop at U. Tennessee Knoxville Date: Thu, 04 Jun 2009 02:41:21 +0000 Attachments: HLAIR.pdf; Goertzel_NeuroSymbolic_2009.pdf Hi Jeffrey, Itamar Arel (a prof at UT Knoxville) and I are organizing a workshop for the fall, which is aimed at laying out a detailed roadmap toward human-level AI. I think I sent you our paper on the "AGI roadmap" idea before, but in case you missed it, it's attached to this email. At the end of this email you will find the plan for the workshop. Notice the carefully structured plan, but also the list of names of folks who are interested to attend. (E.g. it's not just the usual suspects ... Dan Dennett and Barbara Hayes-Roth are AAAI Fellows and big shots, so roping them into something like this is nice.) The core idea here is similar to the workshops that John Laird and Pat Langley organized in Nov 08 and Feb 09 and I attended ... and I think vaguely similar to a workshop that you held with Minsky and others some years ago. However, I think this time we are poised to do it right, because we have a careful plan going in, and because at least two of us (me and Itamar) have well-fleshed out, thorough AGI designs. The Laird/Langley workshops gave me some powerful intuitions about what *not* to do in such a workshop, and we have a couple other veterans of those workshops here (Matthias Scheutz and Joscha Bach). The convention with such workshops is to pay participants' travel, meals and lodging and so I'm wondering if you might be able to help with that. We talked informally to NSF about getting funding for this, but, they basically will only do it if one agrees to invite their people to the workshop, and to have people give talks at the workshop. I don't want this to be another conference where people give talks, I want to actually have a working workshop where we make an AGI roadmap together. My rough estimate of the cost of airfare, meals and lodging is $1000 per person. If we have 10 participants (there are 13 on the list right now, but 2 are from government agencies and hence don't need funding; and I assume some will opt out at the last minute due to scheduling conflicts) then I suppose your arithmetic is adequate to calculate the cost at $10K ;-p So, I wonder if you'd be interested to contribute toward this. I really think this workshop and the paper that will result from it can make a significant difference in pushing the AI community along a meaningful path toward AGI. Of course, we will then present the conclusions of the workshop at AGI-10 ... though we will publish the paper coming out of the workshop in some high-quality journal rather than the conference proceedings, for maximum exposure. I have more to say about other topics but will save it for other emails Thanks Ben P.S. Also, Itamar and I worked out a nifty way to hybridize his neural-net based AI architecture with my OpenCog system ... I attach a draft conference paper on that ... I think this can really accelerate our progress toward AGI ... but that is orthogonal to the AGI Roadmapping EFTA00775409 workshop... ** * WORKSHOP: Creating a Roadmap Toward Human-Level Artificial General Intelligence LOCATION: University of Tennessee, Knoxville DATES: Fall 2009, specific dates TBD, but it will be a 2-day workshop over some weekend FORMAT: This will be a small-group "working workshop" aimed at collaboratively outlining a roadmap toward Human-Level AI (HLAI) based on certain assumptions. Participants will not present papers, although they may distribute papers via email to be read by other participants beforehand. If all goes well, the findings of the workshop will be published afterwards as a paper co-authored by the participants. We will also schedule several conference calls prior to the actual workshop, intended to allow participants to share views regarding conceptual and preliminary issues related to the roadmapping effort (the purpose here is to avoid the workshop itself being bogged down with preliminary "framing" discussions -- we want to get these out of the way beforehand!). PARTICIPANTS: The total number of participants will be small, most likely in the range 10-12. Participants are currently being recruited. The list of currently confirmed participants is: -- Sam Adams, IBM Research - Itamar Arel, Department of Electrical Engineering and Computer Science, University of Tennessee at Knoxville - Joscha Bach, Institute of Cognitive Science, University of Osnabruck, Germany -- Paul Bello, Office of Naval Research (ONR) -- Daniel Dennett, Tufts University -- Stan Franklin, Institute for Intelligent Systems, University of Memphis -- Ben Goertzel, Novamente LLC -- J. Storrs Hall, Foresight Institute -- Barbara Hayes-Roth, Lifelike Solutions Inc. - Alexei Samsonovich, Krasnow Institute for Advanced Study, George Mason University -- Matthias Scheutz, Cognitive Science Program, Indiana University, Bloomington -- Mat Schlesinger , Department of Psychology, Southern Illinois University -- Jun Zhang, Air Force Office of Scientific Research (AFOSR) SCIENTIFIC DETAILS: For information on the assumptions and ideas underlying this roadmapping project, see the whitepaper by Arel, Goertzel and Scheutz at http://www.agi-roadmap.org/images/HLAIR.pdf. The short version is: We intend to use ideas and data regarding human cognitive development to articulate a set of milestones constituting an incremental path leading from the current state of the art to human-level artificial general intelligence. In order to create a concrete series of milestones we have chosen to concentrate on paths to HLAI that involve either virtual-world or robotic embodiment. The intention is that the set of milestones should be independent of particular choices regarding cognitive architecture, learning algorithms, and so forth. Ben Goertzel, PhD CEO, Novamente LLC and Biomind LLC Director of Research, SIAI EFTA00775410 "Truth is a pathless land" -- Jiddu Krishnamurti EFTA00775411