@MIAM_Schwan@BrainChip_inc Such a great point by Kevin regarding TOPS. Especially with heterogenous computation and the fragmentation of computation paradigms coming into play nowadays. How we compare performance needs to be rethought from this perspective
@michellelsun Researching tactile sensing through laser speckle patterns using event-based and neuromorphic hardware. Hoping to create a dense optical representation of touch for robotics, inspired by our somatosensory system
@_joe_harris_ Not to mention power management, and self-recovery on top of all things. You’re not going to have a continuous connection to the network either
But for me specifically - I think there's real merit to neuromorphic front ends(sensory processors) not just for robotics, but for most machinery. I can definitely see some real merit in machinery accident reduction as one idea
Rather than talking up neuromorphic systems - I want to focus on what I think could be a future set of systems that could be utilised in the field of robotics
@computeontology I'd just love for there to be more curiosity and intrigue with pursuing generalised intelligence through more than just the one path dependent architecture. Neuromorphic morphology and architectures should be explored a lot more than it is currently
Also if you want an epic conversation on neuromorphic computing with a better deep dive into the concepts I really recommend https://t.co/m2aOXRNgQ3. Was such a great listen
An AI for a centralised AI similar to our own systems. Reading this paper: https://t.co/oWz9PN9K9v got me thinking about the perspective of tactile sensing and having peripheral AI systems along with generalised systems
Here we've got a couple of layers. One is to represent a piezoelectric layer, and a second is a representation of a silicon skin making contact with an object. The point of contact that you see there is the speckle information
Testing synthetic event streams for tactile sensing before trials with a Prophesee EVS camera. An early step towards neuromorphic systems that could help robots react to touch. Below is an example of the synthetic run
@justforfunrun@BrainChip_inc@computeontology Thanks! Yeah good call. I’d actually love to once I move into the physical experimentation stage. I’m just about to submit our synthetic results
Working on my thesis for utilising light speckles to mimic tactile sensing that can act like our own touch receptors, I came across the @BrainChip_inc's Akida demo and I do wonder whether relying on a conventional sensor might concentrate attention on their product. Really cool
@justforfunrun@MIAM_Schwan@starkbier68@BrainChip_inc Jump on HotCopper and see how toxic it really is when it comes to them. It’s really bad how set groups are about technology. Ironically the thing that should be continuously changing
It's kind of weird to think about how our sense of touch is mostly about pressure. Building this speckle experiment really makes you think about how we operate vs how robots will utilise their senses