3 ms·

AI will need lower power consumption. Digital architecture will (ultimately) not be suitable.

gopalv 4y ago
The company (Mythic AI) in the article was featured on a Veritasium video[1] about analog computing. The interesting part for analog is obviously going to be wear and repeatability, because the digital layer auto-corrects the errors until it becomes damaged and unusable clearly, while analog will slide off somewhat gracefully but you need a way to detect it happening multiple layers later. AI is somewhat tolerant of errors since, the cumulative error across a lot of biases can approach zero if the errors are random. [1] - https://www.youtube.com/watch?v=GVsUOuSjvcg#t=15m44s https://www.youtube.com/watch?v=GVsUOuSjvcg#t=15m44s
- peepeepoopoo3 4y ago
  Mythic shut down recently. Well, technically they still exist on paper, but they're a shell of the former selves. It's a shame, really. Their technology is (was) amazing.
  - smsx 4y ago
    They're back actually - They recently raised a new round of funding and are working to release their next-gen chip. Source: https://mythic.ai/whats-new/mythic-raises-13-million-to-bring-its-next-generation-analog-computing-solution-to-market/ https://mythic.ai/whats-new/mythic-raises-13-million-to-brin... EDIT: I see this was actually mentioned in the article. Well, the new CEO happens to be the former CTO and is quite talented.
emmelaich 4y ago
IMNSHO, AI will the catalyst for going to nuclear power.
charcircuit 4y ago
Low power inference chips already exist and digital computing uses exponentially less power every year.
- mycall 4y ago
  Intel GNA has been in their last three gen CPUs, it is in many places already. I'm not sure if it is low power though.
artemonster 4y ago
Can you please provide any sort of justification for your (quite ridiculous) statement. We have quite a long history of signal processing migrating to digital precisely because it is better in every regard, INCLUDING POWER. Maaaaaybe, since AI neuromorphic-like circuits are quite fuzzy, they wouldnt require such strong noise resilience and precision, this could be a potential use case, but there is also a point of reliability and testability, which on these required scales for AI will be completely infeasible for analog circuits.
anthk 4y ago
No. Today you can get a netbook which can be powered by a phone charger.
- flangola7 4y ago
  Because you can push 200W over USB now

The company (Mythic AI) in the article was featured on a Veritasium video[1] about analog computing. The interesting part for analog is obviously going to be wear and repeatability, because the digital layer auto-corrects the errors until it becomes damaged and unusable clearly, while analog will slide off somewhat gracefully but you need a way to detect it happening multiple layers later. AI is somewhat tolerant of errors since, the cumulative error across a lot of biases can approach zero if the errors are random. [1] - https://www.youtube.com/watch?v=GVsUOuSjvcg#t=15m44s https://www.youtube.com/watch?v=GVsUOuSjvcg#t=15m44s

peepeepoopoo3 4y ago
Mythic shut down recently. Well, technically they still exist on paper, but they're a shell of the former selves. It's a shame, really. Their technology is (was) amazing.
- smsx 4y ago
  They're back actually - They recently raised a new round of funding and are working to release their next-gen chip. Source: https://mythic.ai/whats-new/mythic-raises-13-million-to-bring-its-next-generation-analog-computing-solution-to-market/ https://mythic.ai/whats-new/mythic-raises-13-million-to-brin... EDIT: I see this was actually mentioned in the article. Well, the new CEO happens to be the former CTO and is quite talented.

emmelaich 4y ago

IMNSHO, AI will the catalyst for going to nuclear power.

charcircuit 4y ago

Low power inference chips already exist and digital computing uses exponentially less power every year.

mycall 4y ago
Intel GNA has been in their last three gen CPUs, it is in many places already. I'm not sure if it is low power though.

artemonster 4y ago

Can you please provide any sort of justification for your (quite ridiculous) statement. We have quite a long history of signal processing migrating to digital precisely because it is better in every regard, INCLUDING POWER. Maaaaaybe, since AI neuromorphic-like circuits are quite fuzzy, they wouldnt require such strong noise resilience and precision, this could be a potential use case, but there is also a point of reliability and testability, which on these required scales for AI will be completely infeasible for analog circuits.

anthk 4y ago

No. Today you can get a netbook which can be powered by a phone charger.

flangola7 4y ago
Because you can push 200W over USB now