Feedback from Attendees: Institute on Intelligent Systems: Biological and Computational Perspectives

August 1998



I felt the "Summer Institute on Intelligent Systems" was one of the best small meetings I have been to in many years.

The physical arrangements were outstanding. This was one of the best and most efficiently organized meetings I have ever seen. The entire meeting packed up and moved to Friday Harbor in the middle, an amazing logistical feat. There seemed to be no perceptible glitches and everything ran smoothly. I have helped organize meetings and I appreciate how difficult this must have been to do. Impressive! Congratulations!

The intellectual content was equally good.

The key intellectual decision that colored the entire conference was to stress invertebrate neurobiology as the neurobiology component. Given the way things turned out, this was the right thing to do but it must not have been obvious at the time. Given the conference stress on intelligence, a focus on vertebrate, or even primate neurobiology would have been the more common decision. Such a focus would probably have lead to a great deal of diffuse neurotheorizing from the usual suspects. At this point, primate neurobiology is not sure about much of anything. There are provocative ideas, dazzling and ingenious theories, suggestive experiments, but almost nothing actually nailed down. We really don't have more than a glimmer of how mammalian cerebral cortex is organized. (My bias is that most neuroscience is looking at the wrong level, but that is an idiosyncratic opinion.)

The invertebrate work, on the other hand, is clear. My feeling was that we actually are close to understanding a significant amount of the functionality and organzation of some important invertebrate systems. I felt some important principles may becoming apparent. The thing that most impressed me was the efficient and economical "engineering" found in the invertebrate systems. Nothing is wasted, everything has its function, everything relates to the task, often in an elegant and understandable way.

Some general ideas about modularity and hierarchy were floating around under the surface, but their presence did not seem obvious in the experimental material we saw. There are obviously specialized systems, but the notion of encapsulated modularity as an organizing principle for information processing in the nervous system was not obviously supported. The same comment holds for "hierarchies."

Distinct "modes" of behavior (i.e. feeding, escape, sex, etc.) seemed to be the only obvious high level organizing principle present. Even here, there were complex interactions between drives.

...I predict the beautiful invertebrate work described at the conference is unfamiliar to most systems neuroscientists, especially the large group interested in the high level function of the mammalian cerebral cortex...

-- James Anderson, Professor and Chair of Cognitive & Linguistic Sciences, Brown University


Jeez, what can I say..... that was the best scientific interchange in which I've participated in over a decade..... and the work by you and Chris was everywhere evident. Many thanks for all the blood sweat and tears... the long nights, the aggravation of no-shows, continuing concern about the quality of the discussions. I appreciate also the fact that it was a gamble for you both.... no enterprise (I'm sure MSR and CS are not exceptions) will invest so heavily in a concept like this meeting without putting its proponents into a measure of jeopardy. I hope that the outcome of the Institute will impress all colleagues sufficiently to assure a full appreciation of what you did.

I will follow up with ideas for the future interactions.... the timing couldn't be better in terms of (for instance) 'building in' an assured long-term relationship with the Whiteley Center/MSR/UW-FHL. I'm mulling over ways we could make this compelling/structural, and not just our imaginations (we get busy and things change!).

Eric, I am sure that we can arrange a working visit for you that crosses with Bill Frost's time here next year. The trick will be to get dates worked out with Bill and you that work, then let me know specifically so that we can reserve suitable housing on campus. In following years, the Whiteley Center would provide a suitable venue and housing.

-- Dennis Willows, Director, Friday Harbor Laboratories, University of Washington


Eric,

I've just plowed through several screens of email, and, happily found yours...which permitted me to muse again on the great time in Seattle/FHL. I found the meeting a mind-expanding experience...I just had to shake off the issues at the forefront of my thoughts, listen and free associate. Good for me, I'm sure. I especially appreciated the cross-fertilization of scientific cultures that is rarely possible. It was easy to talk with participants about their work, their biases, and their insights. And, without a doubt, discussion of the Evolution of Intelligent Systems permitted me to sprout a few new synapses...great fun. Your boundless enthusiasm was the spark! Thank you for including me...

-- Martha Gillette, Departments of Cell & Structural Biology, Molecular & Integrative Physiology, the Neuroscience Program and College of Medicine University of Illinois at Urbana-Champaign


Dear Eric,

As I told you before leaving Friday Harbor, the meeting turned out to be an important event for me. Most useful were the contacts that I developed with people interested in neural nets, especially Platt, Bishop, Anderson, and Becker. I have flirted with neural net work for some time, and now that I know people who I can bounce ideas off of and ask for questions of, I think I may go beyoynd flirting. I was also gratified that the crayfish story on local vs distributed modes of escape mediation fired Jim Anderson up considerably; he will now use this as an example in his course lectures. Finally, as a direct result of the conference I became familiar with Minsky's ideas on representation, which provided some significant grist for my mill.

So all in all it was a great success from my point of view. The social aspects were also delightful, a point with which Sally and Darcy concur.

I should add that I was also pleased by your interest in the evolution of the crayfish escape circuit. It occurred to me after I left that the tail end of a chapter I wrote several years had some evolution-related stories that you might get a kick out of; I will fax it too you shortly after sending this message.

-- Frank Krasne, Department of Psychology, UCLA


Eric, Chris, Dennis,

Thank you for organizing such a memorable event! I really enjoyed the meeting and was very stimulated by many of the talks. I was particularly glad to get exposed to invertebrate (vs. vertebrate) neural systems which I had only brief familiarity with before.

The setting and arrangements for the meeting were of course spectacular. Maybe they helped to imprint the talks into our brains better than the standard urban hotel setting! Also it was a great idea to bring families along! This is hard to do at normal meetings at which I always feel that I'm not doing justice to either meeting OR family.

I'm hoping to start a new project as a result of the meeting - I'll keep you posted.

-- Blake Hannaford, Department of EE/CSE, University of Washington


Eric,

I thought the meeting went well, and I certainly benefitted from it. Thanks for your work organizing, both the meeting and the other events.

Probably the most useful thing for me was informal discussions, either in the working groups or otherwise. Knowing some of the current thinking from control theory from decision theory could turn out to be useful for our stuff. I think one thing that might be useful in getting people on the same page from the outset would be to have a series of wish list -- e.g. what do the neural network people want from the neuroscientists, and vice-versa. At least opening that line of thinking would be great.

I'll let you know if anything else comes to mind. It was good to meet you - we'll have to get you over here sometime.

-- Fred Rieke, Program in Physiology and Biophysics, University of Washington


Eric,

First, let me thank you for the fantastic workshop. I learned a great deal and felt it went very well...

As for the conference parts that I attended (all but the last day) here are my reactions (you should discount the importance of these as I lost out on the last break out session):

(1) The general goal of undertanding natural systems that are highly integrative, plastic, adaptive and robust and using that understanding to affect design in artificial systems is both timely and appropriate. The difficulty with this is developing a common language among engineers, computer scientists and biologists. Here the conference excelled by requiring low lingo levels in the talks. This was further reinforced with the breakout groups where all parties tried to maximize communication and minimize jargon.

(2) My personal takes on the outcome are that there is a very important need to understand (a) how "learning" and "adaptation" differ, (b) how natural systems use vast amounts of data in their control dynamics, (c) the dynamics and capacity of large-scale networks (and networks of networks).

(3) I think that there are a lot of fundamental research/technological opportunities at the interface of information theory/dynamical sytstems/and biology. The conference highlighted possible directions (e.g. modelling real networks, understanding dynamical systems that have large data flow, and understanding problems with vast parameter spaces...

Once again many many thanks for the great conference.

-- Tom Daniel, Department of Biology, University of Washington (MacArthur Prize winning neurobiologist)


Dear Eric,

Just back from the International Congress of Neuroethology in San Diego. I believe Bill Kristan sent you a copy of the abstracts, many of which deal with topics highly relevant to the workshop.

The Microsoft/UW workshop was very useful in a number of ways. It helped to define issues of great interest to several different constituencies interested in intelligent behavior. The particular group you assembled was remarkable in the level of motivation to find common ground with kindred spirits from other disciplines. The discussion of what intelligence means in the context of biological systems, robotics and software simulations was leading in a very fruitful direction. Models and simulations were omnipresent at the neuroethology meeting and for the first time a vendor was present selling small robots for exploring navigational control systems in biology. One version of the VersaBot has "Output signals representative of EEGs and single units in frequency and amplitude"-this is a remarkable development in systems neuroscience giving graphic evidence of convergent evolution among intellectual communities with divergent tools but common interests. The Microsoft workshop clearly tapped into an emerging trend toward synthesis of different approaches and recognition among neurobiologists that important insights into biological control systems and dynamics emerge from disciplines outside biology.

I hope vey much that you and your colleagues will find a way to build on the momentum generated by our Workshop. There are several ways to do this. One way is to develop a short course at Friday Harbor which would involve both simulation and modeling exercises on workstations and data collection from real nervous systems. Another way is to have another workshop with significant overlap of goals but with many new participants carefully selected to achieve the same productive group dynamics evidenced at our workshop. I would be happy to contribute to futher definition of either of these if you would find that useful.

I am most grateful to have been included in a truely exilerating intellectual exercise.

-- Alan Gelperin, Bell Labs, Lucent Technologies.


Dear Eric,

The Summer Institute on Intelligent Systems was a spectacular success, for many reasons. The diverse mix of neurobiologists, network modellers, computer scientists, engineers and mathematicians, all with shared interest in understanding intelligent computation, assured that everyone would be exposed to new and stimulating perspectives. The many exotic excursions and exquisite meals created a delightful and unforgettable experience. To my mind, however, the most valuable aspect was the ample opportunity for in-depth one-on-one discussions with colleagues, particularly those that one might not otherwise encounter. Many of the contacts made here for the first time will continue to provide valuable and fruitful interactions.

The program was as intense and full as possible, with a very successful range of topics. As we discussed, there are additional areas that would also be appropriate for a symposium on intelligent systems. In particular, current neuroscience research on cortical mechanisms of higher functions by neural recordings in trained monkeys. Also, brain imaging in human subjects performing cognitive tasks. These might be directions to consider for some future meeting.

I thoroughly enjoyed this meeting and want to thank you and Dennis and Chris for your work in organizing it and congratulate you on its success. Also many thanks to Microsoft for underwriting this extremely valuable symposium.

-- Eberhard Fetz, Program in Physiology and Biophysics, University of Washington


Dear Eric,

I just wanted to say how much I enjoyed the meeting that you organised in Seattle. I found it very stimulating and can see many areas of synergy between biology and computational neuroscience as a result. I will certainly be pursuing discussions with Chris Bishop here in Cambridge and hope that other collaborations will emerge with computer scientists elsewhere. I am more convinced than ever of the help that computer scientists can give to biologists in helping to understand the brain, and of the insights that biologists can give to computer scientists in the problems that they are attempting to solve.

Thanks once again for all that you did to make this meeting such a success and for inviting me to participate.

With best wishes,

-- Malcolm Burrows, Department of Zoology, Cambridge University


Eric,

Very many thanks for organising the workshop. It was tremendous fun.

Overall I found the workshop to be very stimulating with some excellent talks. I think it is important to back away from the chalk face from time to time and contemplate the broader issues in our field, and in this respect the workshop worked well. My only disappointment was the generally poor intellectual level of the one hour discussion session - I think a transcript would be quite amusing - some of the comments were totally content free! Otherwise it was an excellent meeting.

Interestingly I think there is real fruit to be found here, not so much from new learning algorithms for artificial models nor from new insights into biological intelligence gained from learning theory (either of which would be ambitious) but rather from the application of sophisticated data analyses techniques to biological data - most of the biology talks showed very elegant experimental methods coupled with rather simple data analysis (the ICA talk was the only exception). If time permits I hope to talk to Malcom Burrows further about some possible collaboration along these lines.

-- Chris Bishop, Microsoft Research--Cambridge Lab


Eric,

I thought all in all it was a terrific meeting. Not being immersed in either the low-level biology or machine learning communities, I learned quite a lot. I'm slightly less positive about the working group sessions. I guess ours didn't go particularly well. We chucked all your questions and spent a good deal of time formulating our own, then worrying about presenting something coherent to the larger group rather than actually making significant progress on any one topic. So that is probably the fault of our group. Maybe if we'd had more time and one single project to work on it would have been more fruitful. It would have been fun to actually come up with a model of one of the biologist's data as a joint project. Thanks again for inviting me, it was truly a highlight of my summer.

-- Sue Becker, Department of Psychology and Center for Neural Systems, Department of Computer Science and Systems, McMaster University