[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference heron::euro_swas_ai

Title:Europe-Swas-Artificial-Intelligence
Moderator:HERON::BUCHANAN
Created:Fri Jun 03 1988
Last Modified:Thu Aug 04 1994
Last Successful Update:Fri Jun 06 1997
Number of topics:442
Total number of notes:1429

371.0. "IJCAI-91 Report. from Frank Lynch" by CHEVIE::FITZGIBBON (Joe Fitzgibbon, Valbonne EIC) Mon Sep 30 1991 13:44


                       Report on IJCAI 91, Sydney, Australia
         
                                   by Frank Lynch
         
 	 About 1500 people attended the conference in Sydney, Australia, 
	 held 24-30 August.
         
            o Sponsors 
            o Exhibits
            o Other Interesting Exhibits
            o IMKA Interest
            o Technical Program
            
         
         Sponsors
         
         One of the dominating aspects of IJCAI was the presence of IBM. 
         Not from a technical sense, per se, but from sheer presence.  IBM 
         Australia was the Principal Corporate Sponsor.  There were two 
         lesser sponsor levels:  "Major Sponsor" and "Sponsor."  Major 
         Sponsors included Andersen Consulting, Department of Industry, 
         Technology and Commerce [DITAC], and Telecom Australia.  Sponsors 
         included Digital and six other organizations.
         
         
         Exhibits
         
         There were over 20 exhibitors featuring products, publications, 
         and research activities in Australia.
         
         The most notable exhibit areas were by IBM and Telecom Australia, 
         which dominated the exhibit area.  SUN Microsystems had a fairly 
         substantial booth also.  In all cases I thought content left a bit 
         to be desired.   IBM had a similar display to what they had at 
         AAAI91, featuring applications they have built for customers.
         
         SUN had the usual hardware and tools demo.  Telecom was focused on 
         how they are using AI and (mostly) other information systems 
         technology internally.  Their objective was recruiting; part of 
         their display concerned a student competition they ran for AI 
         applications and described the winning entries.  
         
         I think Digital made the right decision.  Rather than compete 
         against the large industrial exhibits of IBM and Telecom, they 
         opted to sponsor an exhibit of AI Research in Australia.  The 
         exhibit consisted of a half hour film about the top 12 research 
         efforts in Australia.  It was quite informative and I asked for a 
         copy for AITC.  In addition, they had eight workstations (mostly 
         non-Digital hardware) with some of the researchers giving demos 
         throughout the week.   The exhibit had a good location, midway 
         along the side facing the rest of the exhibits, with the 
         prominently displayed Digital logo, "DIGITAL, and DITAC, Proudly 
         Presents AI Research in Australia."  The booth was a class act and 
         most likely gained excellent PR for Digital.
         
         Some of the interesting demos in the Digital exhibit area 
         pertained to intelligent information management (e.g. Hyper-
         information).  In particular, the work now being done at CSIRO by 
         Bob Janssen was very interesting and something we should follow up 
         on.  He uses a number of models of information that are inter-
         related to form the basis for intelligent acquisition of infor-
         mation from text (well beyond normal hypertext).  This work com-
         bined with some of the MAYA metaphors could be extremely powerful.   
         The work could benefit from IMKA technology for the underlying 
         representation and models.  The work is based on Janssen's PhD 
         thesis, which will be published shortly both in book form and 
         hypermedia form using the technology developed in the thesis.
         
         
         Other Interesting Exhibits 
         
         The Computer Power Group is a 3000 person worldwide computer 
         services group, with 2000 people in the U.S.  They exhibited some 
         very interesting text retrieval technology/products.  In 
         particular, Artificially Intelligent Document Analyser (AIDA), 
         concentrates on the author's language and style (surface features) 
         rather than contents.  AIDA, with no detailed knowledge of the 
         subject matter, can analyze a 20,000 word document and effectively 
         reduce it to a 2000 word summary in approximately 10 minutes. 
         
         I had a lengthy discussion with their chief scientist Robert 
         Jones.  He will be visiting the States in October and plans to 
         drop by the AITC for a visit.  This could be an interesting 
         technology to add to our intelligent information management tool 
         set.  They are looking for technology partners and licensees.  
         
         BHP is a large Australian energy company (and a large Digital 
         customer).  It looks like they are exploring getting into the 
         systems integration type business.  Their corporate development 
         group demonstrated a number of pieces of AI software tools.  One 
         tool, TABLEAUX, had a similar philosophy to VDE.
         
         Fujitsu had a fairly substantial exhibit showing some AI tools 
         they developed as well as their internal applications regarding 
         spoken language translation.
         
         The ICOT exhibit was the first public exhibit of the results of 
         the Japanese Fifth Generation Project.  It demonstrated a number 
         of applications (VLSI routing, legal reasoning, genetic 
         information processing, and the game Go).  These applications run 
         on a combination of their parallel operating system with 
         relational database system, and logic programming language running 
         on their Parallel Inference Machine.  I personally didn't get a 
         lot out of their exhibit.
         
         Gensym had an interesting exhibit, just from the point of view of 
         seeing a process control application running in Australia.  Their 
         CEO from Massachusetts was there, and told me that a local systems 
         company (tech control) in Sydney, Australia picked Gensym to do 
         the process control on a highly automated margarine plant (Edible 
         Oils Industries, a division of Unilever).  He thought it was more 
         fully automated than any plant in the U.S. using Gensym.  The 
         system runs on VMS with G2, Basestar, and RdB.
                                                                            
         Other exhibitors included: AI Corp, Alcatel Austria-Elin Research 
         Center, Australian AI Institute, Halequin, Hitachi, and  
         Interactive Engineering.
         
         
         IMKA Interest
         
         Through discussions with a number of companies, I found a number 
         of people interested in IMKA.  Their knowledge of IMKA ranged from 
         none, to having heard of it, to having acquired the V1 spec.   
         CSIRO (an Australian government research organization) information 
         technology division was the most interested party in IMKA.  They 
         had a presentation at Digital in the U.S. and one of their 
         scientists had acquired the V1 spec.  BHP is also interested.  
         Telecom Australia seemed to have some level of interest.
         
         
         Technical Program 
         
         The Technical Program consisted of invited speakers, panels, "AI- 
         On-Line," and papers.  There were also workshops and tutorials.  I 
         didn't attend any of the paper sessions, workshops or tutorials, 
         but do have the proceedings containing all the papers and briefs 
         on the panels and a few of the invited papers.  They will be 
         placed in the AITC Library.
         
         The following are comments on parts of the Technical Program:
         
         "AI In Telecom" panel.  I found this extremely disappointing.  I 
         thought the U.S. representatives presented old hat stuff (John 
         Wright of AT&T and Adam Irgon of Bellcore).  
         
         The third panelist was Akira Kurematsu of ATR in Japan (Interp-
         reting Telephone Labs).  He presented a very ambitious cross 
         language communication architecture with translation, interpreta-
         tion and synthesis for spoken conversation.  He said they expect 
         to have some level of "product" in 10 years.  They have been fund-
         ed by the Japanese government and 140 industries since 1986 and 
         have to report out in 1993 on their initial seven year project.
                                                                                
         "Massively Parallel AI" panel.  The panel had David Walz of 
         Thinking Machines Corp., James Hendler of the University of 
         Maryland, and Tetsuya Higuchi of Electrotechnical Laboratory 
         Japan.  Walz stressed that the case-based reasoning type work on 
         the massively parallel systems is here now and is a big deal!   
         Jim Hendler talked about their work on massively parallel 
         processing of frames/semantic networks at Database Performance.  
         
         "AI In Design" panel.  This panel hashed over some of the standard 
         talk about why we need concurrent engineering and the need to 
         capture design intent and help in conceptual design etc.  Marty 
         Tenenbaum of Enterprise Integration Technologies Co. and Stanford 
         University gave his pitch for his MKS system he developed at CIS 
         in Stanford, plus a number of future ideas on collaborative work, 
         communications, and personal notebooks for engineers.  Penny Nii 
         of Stanford University talked about software design, and the need 
         for redefined CASE:  Knowledge Assisted SW Engineering (KASE).  
         The panelists referred to a DARPA design workshop earlier this 
         year that a small number of invited persons attended.
         
         "AI-On-Line: Information Services."   This panel session discussed 
         general information systems applications of AI.  Robert Jones of 
         Computer Power Group talked about the AIDA system I described 
         above in the Exhibit section.  Peter Johnson of Softlaw, a firm 
         specializing in knowledge based assistance to the legal profession 
         told about hypertext stuff for preparing contracts, etc.  Leslie 
         Lazarus a pathologist from St. Vincent Hospital in Sydney talked 
         about a system for pathology advice that is maintained and 
         knowledge is added by the user physicians and technicians.  Shaun 
         Gray, an engineer with the Highway Department in New South Wales, 
         talked about a system he built for resource scheduling and 
         assignment for engineering projects.  Last, but not least, Frank 
         Lynch talked about going beyond isolated knowledge-based systems 
         and managing knowledge assets in an enterprise.
         
         "Invited Speaker." Shigeru Sato, head of AI at Fujitsu 
         Laboratories Ltd, Japan, gave a very interesting talk from several 
         perspectives.  Dr. Sato was introduced as coming from the second 
         largest computer company!  He described Fujitsu's computer 
         business as ~US$12b, 50000 people and 1% AI business.   He stated 
         that their AI R&D group has 400 people, and there is a total of 
         ~1000 people working in AI if one includes systems engineers 
         knowledgeable about AI.  He stated that the big challenge for AI 
         is knowledge acquisition, dealing with large knowledge bases, and 
         methodology.  He went on to describe a number of applications they 
         have done for customers in the banking industry.  Dai-Ichi Kangyo
         Japan's largest bank, has over 10 AI systems in use or under 
         development.  A lot of them use the Fujitsu AI shell that has been 
         used for many customer applications.  One new system for Dai-Ichi 
         Kangyo is a neural net application for forecasting bond rates that 
         supposedly saved DK over US$1M in its first three months of use.
                                        ###
         
         		     Frank Lynch is the AI Technology Manager for 
                             the AITC.  He can be reached at 
                             STEPS1::FLYNCH, DTN 296-5040.
         ==================================================================                                                                            

                           AITE Offers Four New Programs
                                          
         The AI Training and Education (AITE) Group is offering four new 
         programs.  Two are accelerated programs: a 22-day program called 
         FASTRAK and a 10-day Knowledge Engineering Methodologies program.  
         The third program is called Emerging Technologies.  The fourth is 
         called Imaging: Implementing Scalable Solutions for Sales Support.
         
           **************************************************************
           NOTE: Members of the AITC can sign up to take any course 
                 within a program.                                                                                            
         
                 For more information contact:
                 Suegene Levin: AIADM::LEVIN, DTN 291-8600
           **************************************************************
         
         1.  FASTRAK - 22 days - 
         
         Target audiences for FASTRAK are experienced programmers, software 
         engineers, and technical managers interested in applying 
         knowledge-based systems technology to their business problems.  
         The program provides an overview of knowledge engineering 
         information.  It provides the hands-on application training 
         required to practically apply the technology.  
         
         The first session begins on 30 September and includes: 
           o Introducing and Transferring AI Technology
           o Knowledge Acquisition
           o Knowledge Representation
           o Emerging Technologies
           o Designing, Integrating and Testing Knowledge-Based Systems
           o Knowledge-Based Programming (choice of KBMS or NEXPERT)
           o Prototyping
         
         2. KNOWLEDGE ENGINEERING METHODOLOGIES - 10 days -
         
         The Knowledge Engineering Methodologies program is an intensive 
         set of courses geared to the needs of the Knowledge Engineer.  It 
         includes:
           o Introducing and Transferring AI Technology
           o Knowledge Acquisition
           o Knowledge Representation
           o Emerging Technologies 
           o Designing, Integrating and Testing Knowledge Based Systems
         
         3. EMERGING TECHNOLOGIES  - 1 day -
         
         This is a one-day overview course that includes Case-based 
         Reasoning, Genetic Algorithms, and Neural Nets.
         
         It will be delivered in AITE's full program for the first time
         on September 25, and will be offered again on October 9, 1991.
         
         4.  IMAGING: IMPLEMENTING SCALABLE SOLUTIONS - 2 days -
         
         This is a two-day seminar providing an overview of imaging 
         technology and skills to identify and qualify customers needing 
         this technology.  The material is geared to Sales Support 
         personnel involved in selling large scale, high performance 
         imaging solutions. It is a non-product specific course aimed at 
         solutions providing for hundreds of seats, handling tens of 
         thousands of documents per day.
         
         ==================================================================                                                                              
         
                             Digital Internal Use Only
         
         
T.RTitleUserPersonal
Name
DateLines