Day 1 - Welcome to the workshop (Giacomo, Saray, Stan, Rodney, Andre, Tobi)

 

The 2026 workshop kicked off on a beautiful sunny clear morning with light winds. The participants were welcomed by Giacomo Indiveri who presented the history of the workshop since its inception in 2007, 20 years ago. There have been 18 CCNW since then not counting 2 year break for COVID.

Giacomo, Stan Kerstjens, and Saray Soldado reiterated that the morning events are discussions, led by people in certain directions but not presentations. No slides are allowed, not even pre-preparation of the flipchart boards. No applause is allowed, just as in a discussion around a table. Chiara Bartolozzi explained that the projects can start at CCNW, continue  at Telluride and also BNEW (Bangalore Neuromorphic Engineering Workshop) or in the home labs. Participants were reminded to record all contributors to the projects for later possible publication of the results.

Rodney Douglas explained the history of the Telluride workshops. His own experience at Oxford doing neurophysiology and how he was invited by Christof Koch to spend time at Caltech. 

He invited Tobi explain how it got going at Caltech. "Rodney came into picture when I was student in Carver's lab, he was visiting regularly, the Caltech CNS program had recently been formed, it came  from Hopfield". 

Mead was doing digital, had finished his textbook. Around 1968, his father Max asked Carver, "Some people say neuron channels are like transistors, is that true?" Carver started studying a membrane channel. Set up Subterranean Group. Worked out exponential dependency, stuff like that in fungus-produced channel Alamethicin. 

Caltech hired Hopfield as solid state physicist, had not done neuroscience. Invited by Francis O. Schmidt (later developed electron microscopy, made sliding fiber theory of muscles). Coined name neuroscience. Wanted physicist on board, recruited Hopfield 1977. "I don't know about neuroscience", "That's ok, just be a physicist". Was working on theory of electron transport in molecules, but thought it would be interesting for Caltech to start looking at how mind arises from brain. 

Sent note to Murph Goldberger, response was it's a no brainer, let's set it up.  After much Caltech politicking, CNS program got set up. 

Feynman Mead Hopfield were talking about transistors, neurons, in the Physics of Computation course in early 1980s. Read Hopfield's 2013 recounting of navigating Caltech's politics and administration that was required to set up Caltech's CNS program. 

John Allman, famous for finding  Necker cube cells in area MT, was through Misha, Carver's entre to neuroscience, Misha had taken the John/Dick Feynman/Carver course on physics of computation, started Carver working on system level things, suggested silicon retina. 

Carver got grant from aerospace slush fund the System Development Foundation, enough to hire 6 students for 5 years. Arnold Beckman had told Carver about this possibility. Carver's PCMP lab started making chips, an incredible 900 chips over 9 years. Design was informal. Fabrication was free because Carver was involved in MOSIS. Very primitive simulation tools, but a chip could (at absolute minimum) be designed overnight and come back a couple of months later. This is how everything got going.

Rodney Douglas found himself in Koch's group meeting. He asserted XXX, then Misha Mahowald piped up and asked belligerently, "What the hell do you expect?" They started working on the HH neuron model  that got the Nature cover. 

Then Terry Sejnowski got involved and invited all to the seminal meeting in a rented house in Idyllwild. The NSF was very interested in funding a workshop and Rodney, Terry and Christof wrote the grant for Telluride, which was chosen for Christof  dogs' happiness. 

Then similar thing happened for  CCNW when it looked like NSF would stop Telluride funding (which they re-continued after threat of moving to  Europe). Rodney told story of how the original CCNW location was found, based on suggestion from Giacomo who was born there, after an entire-Sardinia search, literally 200m from the original event that took place at inland research station.

After this historical introduction, we all gathered on the tennis court for a wonderful pre-coffee break activity to gauge the interests of the participants, where participants were invited to arrange themselves on the court according to how they considered themselves placed on either 1 or 2 axes.

After the coffee break, workgroups announced themselves, including

• The Brainscales projects
• Jamie Knight: Flight of the bumblebee, measuring insect wingbeats, particularly bumblebees, based on dataset from 5 bees of various belligerence. 2nd: FPGA FENN (FPGA Enhanced Neural Networks). Concrete idea: Make SSM
• Alejandro: TouchAndMotion: XY plotter plus tactile PSD to close sensory motor loop.
• Taekwang and Victoria: SpeckTracker:  Drone plus DVS plus SNN to hand control drone by gestures.
• Marcello: SelfHealingCircuit: Cells that are LUT specified, if LUTs are corrupted by e.g. radiation, how can it heal itself?
• Marian Stacche: Neuroskin (SkinnyNeurons): Optical fiber embedded in silicone skin, can measure shape changes
• Rui Graca: NextGenNVS: DG on next gen neuromorphic vision sensors. WG: Next Gen NVS pixels.
• Miriam and Mohamed: EyeMovementsPanTilt: Pantilt with DVS, what can we do with eye movements?
• Lorenzo and Roel: OnDeviceLearning: FPGA SNN accelerator tied to DVS and interested in learning features and classes online using local clustering.
• Maciej: Camera with Manchester Dudek SCAMP 3 stage retina modeling on and possible FPGA vision transformer on SERIOS, bit serial processing
• Fillipo: AdaptiveController:  leveraging world models in Mujoco
• Carmen: IF neurons: can they generate sine waves with online learning?
• Moritz: DendriticCompute and Small 8 neuron mixed chip with modulatory control, fresh off fab. Simple neuromophic control, 1st is possible dendritic computation such as tapped delay or sequence detection, 2nd is how to deal with mismatch
• Thomas Novotny: DG: SpikeBasedComputation: Learning with spiking neurons, gradients or not? 
• Ricardo: DG: Dynamical Deep Learning and Distributed Learning: Predictive coding on low-d manifolds where simple rules can be effective for learning
• Olympia: WG: Spike coding of a Neurapixel recording dataset. Focus on decoding.
• Ania: WG: flies navigate high contrast arena, there is correlation between particular area activity and straightness of behavior
• Yesin: Cool embedded SNN system
• Abdul: WG: Hyperparameter controlled network with multiple functions noise robust
• Alejandro: WG: ?? Eye movement saccades and learning with pantilt camera. DG: are bioinspired solutions more robust, e,g, can fly connectome result in more robust performance?
• Mirco: Can we extract patterns, dynamics, etc from mixed SNN chip, e.g. DYNAP?
• Dan Goodman: DG: Heterogeneity: Reln between neuro and engineering.
• Leonardo: WG: Mismatch:  Can we use mismatch of activity and timescales (eg on DYNAP) and their distributions to implement mixture of Gaussians.
• Federico: MorseVibrations: Tactile sensing on metal plate using Morse code of vibrations, sensor collects vibrations.
• Giacomo: DYNAP projects, also new Neurocore with on chip learning
• Tobi: Argo the robot sailboat

Stan led a post-it note reorganization of the workgroups and discussion groups for the first 3 days of the workshop.

The rest of the afternoon and evening was occupied by several tutorials, sports (soccer and beach volleyball), snorkeling, and finally, after dinner, a popular discussion group and poker tutorial in the hotel bar lobby. That concluded the first day around midnight.