siping@cathedral.cerc.wvu.wvnet.edu (Siping Liu) (08/08/90)
I met with a few acquaintances from other organizations in AAAI-90 and we talked about our impression on this conference. They have been to nearly all of the previous AAAI annual conferences and they told me it is no longer a great event as it used to be from the numbers of attendants and the papers submitted. I expected a larger exibition but it turned our that many main vendors were missing. Does anyone care to comment on this? or maybe I got a wrong impression? Thanks in advance. siping@cerc.wvu.wvnet.edu
marcel@ADS.COM (Marcel Schoppers) (08/08/90)
I attended the annual American Control Conference in San Diego earlier this year. In that conference there were usually 10 parallel sessions and the proceedings weighed over 20 pounds. When I commented on that, I was told that people just wouldn't come if their papers weren't published. I have often wondered why the AAAI proceedings seem to be the same size every year. Second comment: for much of the outside world (as I hear it from the company my wife works for), the one and only good thing ever done by AI was the invention of expert systems. Since then (it is said) we have merely been soaking up money and have not much to show for it. As my own interpretation of that viewpoint, let me paraphrase a comment that was made (by a control systems person) in the workshop I convened just before the conference: research is not about inventing techniques, it is about solving problems. That comment reminded me of my own occasional perception that AI research is less about solving problems than it is about inventing techniques, especially techniques with memorable names. (How many papers in this year's proceedings even mention an application, let alone a successful one? Leave it to someone else to fail in the attempt to migrate the ideas into practice!) Perceptions such as these, whether or not they are quite justified, will draw neither much attendance nor much funding. I raise the above as hypotheses, however implausible, and leave it to general discussion to do the evidence accrual. Marcel Schoppers
tusveld@hcsrnd.UUCP (Fred Tusveld) (08/08/90)
marcel@ADS.COM (Marcel Schoppers) writes: Status: R In comp.ai you write: >Second comment: for much of the outside world (as I hear it from the >company my wife works for), the one and only good thing ever done by AI >was the invention of expert systems. Since then (it is said) we have >merely been soaking up money and have not much to show for it. As my own >interpretation of that viewpoint, let me paraphrase a comment that was made >(by a control systems person) in the workshop I convened just before the >conference: research is not about inventing techniques, it is about solving >problems. That comment reminded me of my own occasional perception that AI >research is less about solving problems than it is about inventing techniques, >especially techniques with memorable names. (How many papers in this year's >proceedings even mention an application, let alone a successful one? Leave >it to someone else to fail in the attempt to migrate the ideas into practice!) >Perceptions such as these, whether or not they are quite justified, will draw >neither much attendance nor much funding. >I raise the above as hypotheses, however implausible, and leave it to >general discussion to do the evidence accrual. >Marcel Schoppers I agree on the fact that the AAAI this year was particularly theoretical. A few years ago, there used to be an Applications track in the conference. The theoretic flavour of the AAAI is partly due to the success of AI in industry. Distinct conferences addressing AI applications (eg. IAAI, IEEE CAIA) have emerged and receive a lot of attention. Also, in subfields of AI more conferences are being held. A disadvantage, however, is that the cross-fertilization that industry and academia may provide by confronting each other is hard to find these days. Fred TusveldHCS Technology, Industrial Automation, R&D Landdrostlaan 51, 7302 HA Apeldoorn The Netherlands tel. +31 55 498600 tusveld@hcsrnd.uucp.nl or hcsrnd!tusveld@relay.EU.net
sir@PacBell.COM (Sheldon Rothenberg) (08/09/90)
In article <665@babcock.cerc.wvu.wvnet.edu>, siping@cathedral.cerc.wvu.wvnet.edu (Siping Liu) writes: > I met with a few acquaintances from other organizations in AAAI-90 > and we talked about our impression on this conference. They have > been to nearly all of the previous AAAI annual conferences > and they told me it is no longer a great event as it used to be from > the numbers of attendants and the papers submitted. I expected a > larger exibition but it turned our that many main vendors were missing. > Does anyone care to comment on this? I am still sorting my opinions out on this conference, having been to only one other (IJCAI '89). Ronald Brachman, in his talk on the Future of Knowledge Representation talked about the increasing specialization of presentations and bemoaned that this increasing narrowness of effort was excluding many fine papers which bordered on several areas, e.g. planning and representation or machine learning. The attempt to put together the AI on-line sessions to draw those interested in real-world applications got strong criticism from the attendees to whom I talked other than the ART presented "Six Companies reveal the Results Of Expert Systems Successes." The commercial part of the show was much less impressive. People are spending less to promote their products. There appears to be a pervasive feeling that AAAI attracts many faculty and students who do not have money to spend, so why push too hard at this vendor show. DEC was the slickest, Bolesian the most sexist, with a model in a toga handing out roses (this display, whose backdrop had columns and was supposed to remind you of a Roman temple) to indicate its classic nature. IBM was the most underplayed, trying to show off applications rather than sell hardware or their tools. DEC's focus, as was Apollo's and Gensym's and Nexpert's to some degree was to show off applications, some of which were very impressive. DEC had a stunning environmental system which it developed for the minister of the interior of the German state of Baden-Wurtenberg (sic). Andersen Consulting showed off some flashy demos in KEE and Gensym's G2. The German system was written, by the way, in Lisp and Mercury, a European shell. The new products which garnered the most attention were by Intellicorp: Kappa, the PC shell developed by Megaknowledge and ProKappa, a product they developed for the workstation market. It looks like ProKappa will give Nexpert a strong run for the money and will eventually supplant Kee. It is object oriented, is fast, looks slick, and avoids the Lisp syntax that kept KEE from mass usage. This is getting long so I'll cut here and may post some more. Also note the relative absence of connectionists, who now have their own conference. Suspect the same process is already underway with robotics and vision folks. I too would like to hear the impressions of others... Shelley
vt_ai@abo.fi (08/09/90)
In article <|~0$!}_@ads.com>, marcel@ADS.COM (Marcel Schoppers) writes: > > I attended the annual American Control Conference in San Diego earlier > this year. In that conference there were usually 10 parallel sessions > and the proceedings weighed over 20 pounds. When I commented on that, > I was told that people just wouldn't come if their papers weren't published. > I have often wondered why the AAAI proceedings seem to be the same size > every year. We too take that attitude. It is easier to get the expenses for the conference only if a paper or a poster is being presented. > (How many papers in this year's > proceedings even mention an application, let alone a successful one? Leave > it to someone else to fail in the attempt to migrate the ideas into practice!) That is the work of the engineers more than that of computer scientists. The level is often lower, since the engineers have been doing things other than computer science also. It is therefore a little more difficult for them to get papers accepted to computer science conferences. AI research group V{rmeteknik, ]bo Akademi
reiter@forth.ed.ac.uk (Ehud Reiter) (08/10/90)
In article <6888.26c16bb2@abo.fi> vt_ai@abo.fi writes: >In article <|~0$!}_@ads.com>, marcel@ADS.COM (Marcel Schoppers) writes: >> >> I attended the annual American Control Conference in San Diego earlier >> this year. In that conference there were usually 10 parallel sessions >> and the proceedings weighed over 20 pounds. When I commented on that, >> I was told that people just wouldn't come if their papers weren't published. >> I have often wondered why the AAAI proceedings seem to be the same size >> every year. > >We too take that attitude. It is easier to get the expenses for the conference >only if a paper or a poster is being presented. I met a lot of people at AAAI who attended one of the (numerous) workshops, and used that as a funding justification for attending the conference. AAAI workshops have fairly open attendance, and I suspect that many of the workshop papers had larger audiences than many of the papers presented at technical sessions (many of the technical sessions I went to only had 50-100 people in the audience). I certainly enjoyed the workshop I went to, and wish I had gone to more. So, next time AAAI comes around, I suggest signing up for one or more workshops, and using that as a justification for travel money. Ehud Reiter (e.reiter@edinburgh.ac.uk) Ehud Reiter (e.reiter@edinburgh.ac.uk)
narayan@tut.cis.ohio-state.edu (N Hari Narayanan) (08/15/90)
I think one of the main reasons for AAAI no longer representing all subareas of AI adequately is that researchers in those fields now publish in their own conferences...Cognitive Science, Machine Learning, Knowledge Acquisition, NIPS, IEEE CAIA etc. On one hand this indicates a diversification within the field which is a sign of health. On the other, it reduces opportunities for cross-fertilization not just among the attendees but also for someone who is trying to stay abreast of AI through proceedings. It is no longer sufficient to look at one or two major proceedings. This is a danger of too many conferences that are too narrowly focussed. As an aside, I wonder if a dissatisfaction among researchers that their subarea was not getting adequate representation in AAAI was a prime cause for this diversification. Until recently, AAAI used to be the main AI conference in this country and intentionally or unintentionally the collective bias of the program committee and reviewers showed up in the proceedings (which of course can be said of other conferences as well). However, when there is only one major conference in an area, it implicitly determines what is (is not) important in the field. So research considered non-traditional (connectionism, for instance, during its incipience) might have found it hard to be published in AAAI which provided an impetus to look for other new avenues. This is all conjecture, of course! Anyone know if this is one cause for the so-called "decline" of AAAI? ------------------------------- Hari Narayanan Computer & Information Science Ohio State University narayan@cis.ohio-state.edu --------------------------------------