Exactly Cyber/Hammo, the feedback wasn't as glowing or feverish as I'd have hoped, but it's unbiased, which is the main thing.
Cyber I'm not sure how many workshops the company is planning, but I do believe they include an in-depth look at akida for inference and autonomous learning, a session on running examples (CNN to SNN conversion and on-chip native learning), and a chance for people to perform their own experiements under the guinance of Brainchip staff.
BTW, on the weekend I listened intently to a Spiking Neural Network special on the This Week In Machine Learning Podcast, with the head of AI software platforms at Qualcom - there was no mention of Brainchip specifically when hardware was discussed (alas), it was a very interesting listen, though. Some interesting points were:
""ANN have shown us that you don't need as anything as complicated as *actually* modelling the biological neurons to achieve some pretty impressive resuts, in the same way that airplanes don't flap their wings. They're still very complicated systems, but they don't work exactly like the brain works. So with most of the work in spiking, recently we've seen hardware architectures that are event driven - we've seen quite a number of efforts in that area. And using those architectures to map ANN's onto substrates that are more energy efficient, more computationally efficient in some way by applying an attention mechanism, by using analogue circuitry in a lot of cases, different strategies where the angle is more about low power - i.e. activating only the parts of the network that need to be activated at any given point in time" (so only the neurons that are being used draw power.. but I'm guessing this might be similar to parrallel processing in CPU's - which make computers faster in theory, but as a developer you've gotta actually optimise your code to take advantage of the parrallelism - maybe neutral network programmers face the same issue...)
There was also talk about the difficulty of porting networks created on proprietary AI systems to other systems.... which I'm wondering is going to be a blocker for BRN at all (mainly becayse of something I read in an article about whetstone that mentioned, "in most cases, the final trained network is portable across different neuromorphic platfors.... (whereas) All the current work in this area - the Spinnaker project, True North, Akida - are for the most part using proprietary software technologies". Maybe it won't be, since we're supporting popular frameworks like Tensorflow - but will ask Sean Driver for his opinion.
Here's the podcast link in case anybody can decipher more than I was able to -
https://twimlai.com/twiml-talk-280-spiking-neural-nets-and-ml-as-a-systems-challenge-with-jeff-gehlhaar/
(20min delay)