I sometimes wonder if the future implications of neuromorphic computing are under-reported and discussed. Neuromorphic chips have the potential to be true human/computer interfaces, in the way traditional Von Neumann architecture silicon chips just can't be. AI trained on neuromorphic computers may be more human-like, and very different from AI trained on traditional silicon chips. If merging AI with a human brain was possible, it seems far more likely with these types of chips.
Finally, there's their fuel efficiency. That seems really futuristic compared to today's talk, from some AI leaders, of coal-fired AI data centers the size of Manhattan.
The world's largest neurocomputer simulates a monkey's brain
Neuromorphic computing just marked a major milestone, with a brain-like computer the size of a monkey's brain being achieved. What are the future implications of this technology?
byu/lughnasadh inFuturology
3 Comments
>What are the future implications of this technology?
Well, based on the way AI-marketed stuff is going, is it’ll put me out of a job faster than I anticipated and I’ll never see any benefit from it, just the consequences.
This order of magnitude neuromorphic systems exist for about 2 years now. Chinese Darwin 3 have the same number of synapses but more neurons per chip as Intel’s Laihi 2. It’s more an evutionary than qualitative change. Those SSNs are not yet as flexible as software defined ones, based on LLMs. However China made one significant step – they are able to make 64 chips on one wafer, already connected in the fab. That significantly reduces overall footprint of the system and allows for much faster interchip communication for 128m neurons. You just need 640 those wafer chips to reach neuron number of human brain. However at this time we don’t know yet how to effectively train them.
It’s an interesting area, but as another post implicitly mentioned it’s just kind of a slow one. I think the most intriguing thing in it right now is probably, what, still Numenta? Which to be fair, their Thousand Brains thing is a decent focus, and could potentially inform some breakthroughs in the field, but they tend to be quiet most of the time.