Magnon-based computation could signal computing paradigm shift | Hacker News Reader

Magnon-based computation could signal computing paradigm shift

phys.org

84 pointsisaacfrond27 comments

Life is too short to read articles titled "x-based y could signal z paradigm shift" from phys.org.

Regrettable that academia didnt learn from industry how to deploy, only how to spin. When everything is framed as an "investment", researchers aren't incentivized to discover but to squeeze out "returns" even when there arent much

Perhaps, more precisely, it's a shame that governments turned research funding into a private sector game like start-up funding.

It seems no mystery why western governments are saying insane things like they need to be competitive in the quantum computing space: their frontal-lobes (the research industry) has been hijacked by the hype machines formerly constrained to the private sector.

This really it's the nail on the head and it has other consequences.

I was talking with a colleague about how it is to lead a research group and essentially it is very much like running a start up always in the phase of trying to attract funding without the prospect of things ever becoming self sustainable and there being a great monetary reward.

This is so bad it's not even clear what they are spinning. Is this a memory device? A compute element? What?

Physicists have been fooling around with yttrium iron garnet for decades.[1][2] It apparently has really weird interactions between photons and magnetic fields. That's promising, because semiconductors came from studying weird interactions between electric charge and current flow in unusual solid materials. Yttrium iron garnet been tried experimentally for microwave phase-shifters and beam-forming antennas, but doesn't seem to have appeared in products yet.

So, decades of papers, but no products yet. There's real physics there, but the hype is overblown. Probably because someone needs funding and grad students to work in this obscure area.

[1] https://www.sciencedirect.com/topics/physics-and-astronomy/y...

[2] https://www.nature.com/articles/s41535-017-0067-y.pdf

> Regrettable that academia didnt learn from industry how to deploy, only how to spin.

Hey, this was from phys.org and spin matters a lot to certain physicists!

More seriously: with funding being less prevalent and the cultural shift to the dominance of commerce-worshipping Gradgrinds, university press offices and even individual professors are under a lot of pressure to point out the practicality (which today means “marketability”) of everything, even an abstract mathematical proof!

I can’t count how many academic presentations I’ve heard that ended with 15-20 minutes with the author trying to justify the work. They are professors, not business people, and they almost always sound like they are struggling to jump through that hoop. The time would have better been spent talking further about the actual subject matter!

I’m old enough to remember when the memristor was going to signal a computing paradigm shift.

Memristors are inherently not practical, and there were plenty of people pointing this at the time.

Magnetronics are an open question, nobody really knows how useful it can be. (But they have more promise for energy efficiency than miniaturization and cost-cutting.)

> Memristors are inherently not practical

There were practical products based on it, shipped in volume!

What do you mean they were "inherently" not practical?

Diffusion based products have a minimum size that prevents they being used anywhere where they could be useful.

What practical products did you see?

HP still knocks the dust off that one every few years, touts their major achievements towards it each time, and implies it's only a couple of years until it'll hit the market.

It has been a few years, which means we're probably close to the next round of articles.

I heard a rumour that they sold of the IP rights piecemeal to so many different entities that it's nearly impossible to actually legally use the technology in a product anymore.

Shoehorned in a mention of AI…

Did I understand this right: they can do 0 to 1, but haven’t figured out 1 to 0 yet?

That’s a big “could”.

I don’t like the tag links the site inserts in to the text. If you follow one, you generate a nice extra hit on the site, but I struggle to see the value.

it’s an ad for their lab and research. that was clear to me before I clicked the link. I didn’t know anything about magnetic approaches to this tho so I appreciate the overview tbh

So if I am understanding correctly this YIG-nanomagnet device expands on the work of spin-wave computation in a hybrid CMOS setting and allows for read and write which would allow in memory computation? How does a YIG-nanomagnet work? Does it hold its state after a charge is applied or does it need a constant amplitude to keep it in either a 0 or 1 state? I find this very interesting and would love to read more about the work.

This is an alt name for spin wave conputing.

I'm still waiting for phononic computing to become a thing. (Note: phononic, not photonic).

https://aip.scitation.org/doi/10.1063/1.4919584

>"Nanomagnet reversal happened only when the spin wave hit a certain amplitude, and could then be used to write and read data."

Reminds me of Mr. Spock (being tested by an automated AI test administrator) in "Star Trek IV: The Voyage Home", when the AI asks the following question:

"Adjust the sine wave of this magnetic envelope so that anti-neutrons can pass through it but anti-gravitons can not..."

https://www.youtube.com/watch?v=u2ooUXjNPS8&t=51s

(A classic line, still to this day! <g>)

Isn't there a little law of thermodynamics here about doing work and dissipating energy?

To be clear, there is some dissipation with magnons, just a lot less than Joule heating. But exciting a magnon will heat up your sample, because magnons don't live forever and decay after some time. That energy has to go somewhere.

Reversible computers avoid this by producing arbitrarily little entropy

How do you deal with all the output "garbage" signals that reversible computing produces? Do they all get sent somewhere, and the place where they get sent to discards them but heats up?

Being able to choose where you want do dissipate heat would be a game changer in itself

From my recollection of the Feynman lectures on the subject, basically yes. Nevertheless the overall theoretical efficiency is still considerably higher than an irreversible machine.

It’s not an area where I’ve followed the subsequent research though. I’d be delighted if someone who knows more chimed in. There are also some obvious advantages for being able to choose precisely when you do a reset and generate your waste heat.

The garbage output signals get recycled indefinitely. Only the useful output ever needs to generate heat.

It was thought for some time that they needed be turned to heat, but that's actually untrue. Using quantum gates, it is possible to do an arbitrary classical computation and recycle the anciliary bits.

Neat! Any reference you could share for me to read more?

There is a short but sufficient Wikipedia article on it - it is called uncomputation : https://en.wikipedia.org/wiki/Uncomputation?wprov=sfla1

The textbook I used in my quantum computing class dealt with the subject far better, but unfortunately it's not public yet :/

Reply on news.ycombinator.com