Cool little #neuralnetwork project recreating #lena using minimal #neurons https://incoherency.co.uk/blog/stories/l.. Some other people's results is really quite cool: https://imgur.com/a/BrNhwkF Currently set as my #wallpaper !
Using #embedded #magnets in #muscle #tissue could for sure be the way forwards for #bionic #limbs https://hackaday.com/2021/08/21/magnets-.. Using #neurons and #electrodes is flakey as hell at the best of times. Someone has pointed out how this may affect a #patient 's ability to go through an #mri though... Personally I would be looking towards a #neuralnetwork to predict #movement #intention ...
Random #brainfart : The current state of #neuralnetworks is completely #inaccurate for how #neurons actually operate. We should rename them to 'neuron-inspired networks' (NIN). One of the largest issues for me is that each neuron's affect on the processing of information is *overweighted* - i.e. there is no local #spatial #redundancy built into the #network structure. I'm not devaluing the field (it clearly produces results) and I'm not suggesting there is a better approach, just that it's not there yet.
#aardvark raises a thought provoking point when roughly comparing #neurons and #transistors https://aardvark.co.nz/daily/2021/0707.s.. As I've discussed with people before, there's a good chance we have enough computation now, we just need the right algorithms. I think we need to re-think AI structures, etc, but I think we could get close to something human-like with what we have - or at least have a very good idea of what we are missing.