TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Duke University and IonQ Develop New Quantum Computing Gate

19 点作者 donutloop超过 3 年前

2 条评论

reikonomusha超过 3 年前
HN perhaps isn&#x27;t the right forum for this kind of commentary, but here it is anyway...<p>This is a schematic for a new physical implementation of a particular mathematical operation, on a specific modality of qubit. The same mathematical operation can be computed on any other quantum computer in production today; very easily so if you use a compiler like QUILC [0]. Some architectures even have ways of doing native Toffolis.<p>Unfortunately, I don&#x27;t personally believe the headline summary and following text are written in good faith.<p>&gt; Researchers discover a new method to perform the N-qubit Toffoli gate, a more efficient quantum operation helpful in scaling quantum algorithms.<p>They did discover a new way to perform this mathematical operation (the N-qubit Toffoli), but it&#x27;s highly dubious that it will &quot;scale&quot; <i>any</i> existing quantum algorithm up in any useful manner. Toffoli is like a quantum computing &quot;NAND&quot;-gate of sorts, it&#x27;s useful in building up larger computations (like adders and multipliers). The problem is, it&#x27;s not error corrected. So imagine building a full-adder, but every logic gate is inaccurate by about 1%. In series, those 1%&#x27;s rack up quickly.<p>&gt; New family of N-qubit gates can only be run on IonQ quantum computer architecture.<p>This is hugely misleading. You can do an N-qubit Toffoli on any computer. That&#x27;s what a compiler is for.<p>That&#x27;s like saying &quot;You can only do AES encryption on Intel processors because they have the instruction for it.&quot; It&#x27;s not true. You can do AES on any computing architecture, especially if you have an implementation in a portable programming language.<p>&gt; Once implemented, gate is expected to speed up several fundamental algorithms for quantum computing.<p>Like what? Again, it&#x27;s unlikely. A non-error-corrected Toffoli has limited utility, unless there&#x27;s deep research in NISQ algorithms using slightly unfaithful Toffoli operations profitably that I&#x27;m unaware of.<p>Later in the article they mention:<p>&gt; IonQ and Duke’s discovery may lead to significant efficiency gains in solving fundamental quantum algorithms, such as Grover’s search algorithm, variational quantum eigensolvers (VQEs), and arithmetic operations like addition and multiplication.<p>Grover requires a fault-tolerant oracle for non-trivial problems, so that&#x27;s out. Arithmetic the same (see above). Nobody to my knowledge has developed any reasonable Ansätze for VQE that use Toffoli gates that outperform existing ones—and VQE doesn&#x27;t exactly have a lustrous record anyway for doing anything useful at scale.<p>I think IonQ has made an interesting scientific achievement, and might even make for a few cool proof-of-concept demos, but it&#x27;s hard to believe it&#x27;ll lead to much of anything useful. Their speculation in the article, in my opinion, is squarely out-of-bounds.<p>[0] <a href="https:&#x2F;&#x2F;github.com&#x2F;quil-lang&#x2F;quilc" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;quil-lang&#x2F;quilc</a>
评论 #30357125 未加载
Strilanc超过 3 年前
A couple nits:<p>1. This is a theory paper. They haven&#x27;t actually performed the gate on hardware yet (or at least there was no data on that in the preprint). They need to experimentally show that directly doing the gate has lower error than decomposing it into the existing two qubit gates.<p>2. The article reads like this is the first time anyone has tried to do Toffolis natively. Neutral atom people also do native Toffolis [1].<p>1: <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2010.16036" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2010.16036</a>