The CapoCaccia Workshops toward Neuromorphic Intelligence

History of the workshop

The Capo Caccia workshops have a long heritage.

In the late 1980's a group of neuromorphic engineers grew up around Carver Mead at Caltech. That group group included neuroscientists from the Caltech CNS program, the MRC Anatomical Neuropharmcology Unit in Oxford, and the Salk Institute. Their common goal was to understand the principles of the nervous system at a level which could be implemented in electronic systems embedded in CMOS VLSI. This group created early silicon sensors such as retinae and cochleae; network processors for resolving stereo correspondence; conductance-based neurons, etc. Soon the question of system integration arose, and various ideas for the construction of a "Silicon Cortex" were being discussed. In order to get focussed on a specification for such a system, the group decided to meet for a week-end retreat. The retreat took place at a large house in Idylwylld, in the hills east of San Diego. At that time PBS was shooting a documentary about the brilliant and charismatic Misha Mahowald, a student of Carver's who had collaborated with him in the construction of the early silicon retina, and had also made other seminal contributions to the emerging field. A short cut from that film appears below.

The Idylwylld meeting was so exciting and innovative that the groups determined to hold further collaborative meetings. By good fortune, at about that time the US National Science Foundation decided to establish a program that would encourage the interface between Neuroscience and Engineering.

At the recommendation of the Brain Working Group at NSF, a one-day meeting was held in Washington on August 27, 1993. The recommendations of this group were for NSF to organize and fund a “hands-on” workshop that would draw together an international group of scientists interested in exploring neuromorphic engineering. The goal of the workshop would be that scientists collaborate to design and fabricate artificial neural systems, such as vision systems, head-eye systems, auditory systems and autonomous robots, whose architecture and design principles are based on those of biological nervous systems. At that time, very few groups were pursuing research, the individual scientists were drawn from a variety of disciplines, and the academic and industrial research had few common meeting points.

Sejnowski, Koch, Ballard, Mahowald and Douglas proposed such a workshop that would be highly interactive, practical, and educational. They proposed that the workshop have a duration of at least 2-3 weeks, and be held in some attractive remote location, which would encourage creativity and collaboration. Their choice of location was Telluride, then a small town at 9000 feet high in Southwest Colorado, and about seven hours by car from Denver (see the workshop's website).

That workshop is still held annually — and is by now the one of longest running and  most successful of NSF's scientific workshops. Over the years nearly 1000 participants  have attended the workshops. They are drawn from academia, government laboratories, and industry. Their backgrounds have spanned physics, robotics, computer science, neurophysiology, psychophysics, electrical engineering, and computational neuroscience. The Workshop provided background material in basic cellular and systems neuroscience as well as practical tutorials covering all aspects of analog VLSI design, simulation, layout, and testing as well as sensory-motor integration and active vision systems. Working groups were established in robotics (focused on the use of Koalas and legged robots) — this included Central Pattern Generators, sensory-motor interactions, inter-chip communications, active vision, audition, and spatial localization using vision and audition together, attention and selection, locomotion and industrial applications.

The first week covered transistors and simple circuit design, basic biological concepts and systems, introductions to the programming environments needed for the systems brought by the various course participants as well as an introduction to wheeled and bipedal robots. The second week focused on design frames for silicon retinas, from the silicon compilation and layout of on-chip video scanners, to building the peripheral boards necessary for interfacing analog retinas to video output monitors and scanners. The third week featured sessions on floating gates, including lectures on the physics of tunneling/injection and on robotics, as well as extensive experimentation with test chips and the various robots. Most recently we have been combining these systems to begin working toward more sophisticated behavioral outcomes. Projects carried out by subgroups during the workshops included active vision, audition, olfaction, motor control, central pattern generators, robotics, multi-chip communication, analog VLSI, and learning. The results of these projects have been recorded in the Annual Reports of the workshops submitted to NSF. Many lasting collaborations were established during these projects and many publications arose out of work initiated there.

With the New Millenium the organizing committee of Telluride decided to shift the focus of the workshop up from chip and small system level towards the more general task of building plausibly cognitive systems. At the same time the European contingent of Telluride decided to launch a similar workshop series in Europe, that would serve the growing interests in the field there. In particular, the European workshop would  serve as a meeting point for scientists collaborating under EU's vigorous Future and Emerging Technologies program. Rodney Douglas and Giacomo Indiveri of INI were the founders of this new workshop and they selected as a matching location to Telluride the beautiful and remote Capo Caccia peninsula in northwest Sardinia. That workshop was named the Capo Caccia Workshops for Neuromorphic Cognition. The workshops would last 2 weeks rather than 3; place a greater emphasis than Telluride on a free-format program; and encourage social interactions by having all participants housed in one full-board hotel. In addition the CC neuromorphic workshop would be self-funding, so relieving the organizers and participants from the complications and overhead of grant-writing and reporting.

As with Telluride, the CC meetings have grown in popularity and effect, despite the inconvenience of travelling to this location.

Recently the organizers have decided to shift the long term goal of the series once again. As of 2020 the workshop will be named The Capo Caccia Workshops toward Neuromorphic Intelligence.