AMD Patent Proposes Teleportation to Make Quantum Computing More Efficient | Hacker News Reader

AMD Patent Proposes Teleportation to Make Quantum Computing More Efficient

tomshardware.com

34 pointspunnerud6 comments

(Disclaimer: I work on google's quantum team and I'm not gonna read the patent for legal reasons.)

From the perspective of someone in the field, the title of this article sounds kinda funny. The notable thing would be if the design of some large scale quantum system somehow avoided teleportation. Imagine an article that said "AMD proposes transistors to make computing more efficient" and you'll get a sense of what I mean.

Teleportation is one of the first things you will learn about in an introductory quantum computing course. It is an incredibly useful building block, with even more useful generalizations (e.g. gate teleportation [1], delayed-choice routing [2]; I gave a tutorial on this topic last year [3], although it wasn't targeted at beginners so maybe HN is the wrong audience). It is not surprising to see teleportation in quantum architectures.

[1]: https://en.wikipedia.org/wiki/Quantum_gate_teleportation

[2]: https://arxiv.org/abs/1210.4626

[3]: https://www.youtube.com/watch?v=Gl4iW1kYLKo

The title is clickbait. It sent me down a long rabbit hole reading old research papers about experiments into quantum teleportation. I'm not going to pretend I understand any of this, but it seems people have been able to entangle particles to demonstrate a "teleportation" effect whereby doing something to one affects the other even as they move apart in space, but because you need the original quantum state to reconstruct any information and can only get that through a classical channel, you're still constrained by the speed of light, so it seems like you get your cake and eat it, too. There is spooky action at a distance somehow, but physics doesn't break and we don't get subspace radio or time travel or anything.

But apparently fiber lines aren't good enough to not break the entanglement, so the only place this really works well is in space, and China has put a satellite in orbit for exactly that purpose, to generate quantum key pairs for unbreakable encryption.

So how the heck does this work in a computer? Do quantum computers have to run in an artificial vacuum?

You can use fiber to distribute entanglement. You can use entanglement distillation to deal with noise. You can use entanglement swapping to deal with router hops. But it will be awhile before anyone can actually build a large scale quantum network. IMO it's not hard once you have quantum computers to use for the routers, but another way of saying that is that a good quantum network is harder to make than a good quantum computer.

The interesting aspect of teleportation, from a system design perspective, is that it gives you more flexibility in which parts have to be quantum and which can be done classically. Instead of needing a quantum channel now, you can instead use a classical channel now if you have previously used a quantum channel to share entanglement.

As someone who knows little of quantum computing, can anyone explain what the article means by 'Multi-SIMD' in a conventional computing context?

...Taking a shot in the dark here, because I opened the article expecting some writer mixed up MIMD and SIMD, but there might be some nuance here that makes the Multi-SIMD make sense.

Single Instruction Multi Data refers to implementing processing logic that performs the same operation on multiple data. The article seems to be using Multi-SIMD to describe a state of affairs where the processor is capable of determining data-dependence, and splitting out dependency coupled SIMD instructions from non. Those non-coupled instructions get scheduled to run in parallel, but then you're left with these dependent ones to go through serially.

Given Quantum teleportation being mentioned, the only thing that even starts to make sense to me is that you're implementing some sort of coupling via entanglement whereby state put in place in the teleporter gets rectified and instantiated in the receiver the moment the prior calculation is done, skipping wire propagation delays. That sounds way too handwavey and dubious for me to even pretend it sounds reasonable.

Be kinda cool if it was pulled off though. There's be a bunch of seriously wicked innovations possible if you could feasibly delete a good portion of intermediary interconnects and replace it with spooky action at a distance while still getting reliable results.

Upvoted and When the teleport devices are finally coming ? /weekend rant

Reply on news.ycombinator.com