AI 撬动量子计算 Anthropic 称 Claude 设计的量子电路超越顶尖现有方案AI Unlocks Quantum Computing: Anthropic Says Claude-Designed Quantum Circuits Beat Top Existing Approaches


AI free token/AI免费积分

IT之家 10 月 17 日消息,Anthropic 于 10 月 16 日发布研究称,其 AI 模型 Claude 在几乎无需人工干预的情况下,针对 8 类量子计算任务,外观出比已发表最优电路更高质的思路。

IT之家援引博文介绍,量子计算机依靠量子比特运行,但当前设备通常仅有数百到数千个量子比特,且对热量、振动等环境噪声极为敏感。量子比特和运行时间都是稀缺资源,因而让量子计算走向实用的关键,是减少电路所需的量子比特数和量子门数量。

量子比特(qubit)是量子计算的“信息单元”:经典计算机用比特表示 0 或 1,而量子比特(qubit)则可以处于 0、1 的叠加态,即同时带有两种状态的“概率幅”。不过量子比特格外脆弱。热量、振动、电磁噪声等都会造成退相干和计算错误,因而当前量子设备通常只有数百到数千个量子比特,且需要极低温、隔离和纠错机制。

而量子门是对一个或多个量子比特施加的幺正变换,相当于经典逻辑门(如 AND、NOT)的量子版本,但作用对象是叠加态和纠缠态。量子电路就是按时间顺序排列的一组量子门,外加初始化、测量等步骤。

衡量量子电路的“成本”通常看两个指标:需要多少量子比特,以及需要多少量子门。量子比特越少、门数越少,硬件压力、运行时间和噪声累积通常就越低。

Anthropic 研究团队使用通俗语言,向 Claude 阐述了来自化学、优化、物理模拟、数论和数值计算等领域的 8 项任务,并设定了自动评估程序。

该程序会检验 Claude 提出的电路是否正确,并按“量子比特数 × 量子门数”打分;Claude 随后依据反馈迭代改进,形成闭环外观流程。

结果表明,Claude 发现的新外观不仅减少了量子比特和量子门数量,而且并非小幅微调既有思路,更是提出了新的数学与结构思路。

AI 撬动量子计算 Anthropic 称 Claude 设计的量子电路超越顶尖现有方案

例如,在模拟硫化铁分子时,Claude 利用分子的对称性和重复结构,将部分结果计算一次后重复使用;在量子优化算法的解码环节,它将最昂贵步骤的成本降低至 12 分之一。

其中一项代表性成果涉及椭圆曲线点加法。Claude 经过约 53 小时的自主探索,提出了无需存储完整分支历史即可反向运行的可逆算法,并形成一系列在量子比特数量与运行时间之间取得不同平衡的新外观。

AI 撬动量子计算 Anthropic 称 Claude 设计的量子电路超越顶尖现有方案

椭圆曲线点加法的资源。粗体行是 Claude 的外观;其余为先前发表的最佳外观。

研究人员认为,这类进展有望降低实用量子应用对硬件规模的要求,让新药外观、高质电池、新型催化剂和更清洁的肥料生产等难题的量子计算应对思路更早落地。不过,研究也指出,量子电路外观只是构建容错量子计算机的一环,纠错、硬件控制等难题仍需协同突破。

AI 撬动量子计算 Anthropic 称 Claude 设计的量子电路超越顶尖现有方案

本文来源:https://www.ithome.com/1/014/317.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

IT House news on October 17th, Anthropic released a study on October 16th, saying that its AI model Claude, with almost no human intervention, has a higher quality idea than the published optimal circuit for 8 types of quantum computing tasks.

IT House quoted the blog post as saying that quantum computers run on qubits, but current devices usually have only hundreds to thousands of qubits and are extremely sensitive to ambient noise such as heat and vibration. Both qubits and runtime are scarce resources, so the key to making quantum computing practical is to reduce the number of qubits and quantum gates required by the circuit.

A qubit is a "unit of information" for quantum computing: a classical computer uses bits to represent 0 or 1, while a qubit can be in a superposition of 0 and 1, that is, a "probability amplitude" with two states at the same time. But qubits are exceptionally fragile. Heat, vibration, electromagnetic noise, etc. can cause decoherence and calculation errors, so current quantum devices usually have only hundreds to thousands of qubits, and require extremely low temperature, isolation, and error correction mechanisms.

The quantum gate is a quantum version of the classical logic gate (such as and, not), but the object of action is the superposition state and the entangled state. A quantum circuit is a set of quantum gates arranged in chronological order, plus initialization, measurement, and other steps.

Measuring the “cost” of a quantum circuit typically looks at two metrics: how many qubits are needed, and how many quantum gates are needed. The fewer qubits and gates, the lower the hardware pressure, runtime, and noise buildup.

Using popular language, the Anthropic research team explained eight tasks to Claude from the fields of chemistry, optimization, physical simulation, number theory, and numerical computation, and set up an automatic evaluation program.

The program verifies that the circuit proposed by Claude is correct and scores it according to "number of qubits × number of quantum gates"; Claude then improves iteratively based on feedback to form a closed-loop appearance process.

The results show that the new appearance discovered by Claude not only reduces the number of qubits and quantum gates, but also not only fine-tunes the existing ideas, but also puts forward new mathematical and structural ideas.

For example, when simulating iron sulfide molecules, Claude uses the symmetry and repetitive structure of the molecules to compute partial results once and reuse them; at the decoding stage of the quantum optimization algorithm, it reduces the cost of the most expensive step to one twelfth.

One of the representative results involves elliptic curve point addition. After about 53 hours of independent exploration, Claude proposes a reversible algorithm that can run in reverse without storing a full branch history, and forms a series of new appearances that strike a different balance between the number of qubits and the running time.

Researchers believe that such progress is expected to reduce the hardware scale requirements for practical quantum applications, so that quantum computing ideas such as new drug appearance, high-quality batteries, new catalysts and cleaner fertilizer production can be implemented earlier. However, the study also pointed out that the appearance of quantum circuits is only part of the construction of fault-tolerant quantum computers, and problems such as error correction and hardware control still need to be collaboratively broken.

This article is from: https://www.ithome.com/1/014/317.htm For copyright issues, please write to: 13951431913#139.com or submit a review on the right.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry