Poll of the Day > Computing power may increase 100x in the near future (shitty source)

(+) Yeah! (+)

https://www.independent.co.uk/life-style/gadgets-and-tech/news/ai-machine-learning-light-speed-artificial-intelligence-a9629976.html

---

Those things have been around forever, how do you think fiber optic cables work?

The biggest concern is getting it inside of processors, which, yeah getting glass wires laid out nicely is a lot harder than having some copper solution thrown on boards.

If the paper that they're referencing is linked easily, I could probably read more into it, but the basic idea isn't new.

---

shadowsword87 posted...

Those things have been around forever, how do you think fiber optic cables work?

"those things"? this is describing photonic TPUs, which definitely haven't "been around forever".

https://aip.scitation.org/doi/10.1063/5.0001942

the idea of photonic CPUs isn't new, but an actual strategy for implementing them definitely is, and TPUs that are 100x-1000x times faster/more efficient than what we already have is pretty scary.

---

Sahuagin posted...

"those things"? this is describing photonic TPUs, which definitely haven't "been around forever".

https://aip.scitation.org/doi/10.1063/5.0001942

From the same author, in 2007
https://arxiv.org/ftp/arxiv/papers/2007/2007.05380.pdf

They've gone around the block a few times with different names.

---

shadowsword87 posted...

They've gone around the block a few times with different names.

well, publishing multiple papers in the same domain is not necessarily a duplication of the same concepts. I would have to read both papers to know that they're talking about the same thing.

shadowsword87 posted...

From the same author, in 2007
https://arxiv.org/ftp/arxiv/papers/2007/2007.05380.pdf

this is discussing a method of implementing a very specific (photonic) processor that solves very specific problems (partial differential equations).

the other paper is discussing a method of implementing photonic TPUs.

I don't know what you're expecting. if these are the guys at the front of optical processor research, why shouldn't this be the kind of paper they'd publish?

---

Sahuagin posted...

the other paper is discussing a method of implementing photonic TPUs.

after reading a chunk of the paper, it is specifically a method of implementing parts of TPUs to allow fast matrix multiplication (using existing optical methods).

apparently matrix multiplication is an important part of neural-network machine learning, but has a computational time complexity of O(n^3) (the time it takes to compute increases with the cube of the width of the square matrix).

(there is also a problem I keep reading about recently regarding von Neumann architecture and the CPU-memory bottleneck, which is not present in GPU and TPU architectures.)

using electronic computers and some kind of optimization technique you can get matrix multiplication complexity down to O(n^2.373), but the paper is describing in great detail an optical hardware implementation that can compute matrix multiplications (with one varying input matrix, and one relatively non-varying matrix) in O(1) time complexity (specifically less than 20 picoseconds).

probably not worth going into much more detail than that. they use a germanium/antimony/selenium glass for the 'wires', with tungsten metal walls, and they have to use "heterogeneously integrated optimized photonic memory" for reasons, and you can see diagrams of the multiplexers and so on...

I don't know what there is to doubt about this. it's great to see advances in both realms of hardware neural-network processors and optical processors.

as for the thread title "Computing power may increase 100x in the near future", that's kind of misleading. the paper says they can increase TPU power efficiency by hundreds or thousands of times. this is not something that will directly apply to your home computer, but eventually this kind of work should lead to better home computers.


---