Claude Designs Quantum Algorithm in 53 Hours, Reduces Operations by 17x

Anthropic's artificial intelligence (AI) model Claude has autonomously designed eight quantum computing procedures, achieving improved efficiency in seven of them. In some cases, resource efficiency was enhanced by over ten times. The company announced the research findings on April 16, with a focus on how AI can optimize quantum computing operations and potentially impact cryptocurrency security systems.
Claude autonomously designed eight quantum computing procedures, improving resource consumption in seven of them. Reducing both the number of qubits required and computation steps is critical for practical quantum computers. The research has drawn attention because efficient quantum calculations could affect the computational resources needed for cryptographic decryption.
Bitcoin and Ethereum use elliptic curve cryptography for transaction signatures. While it's difficult to reverse-engineer private keys from public keys with conventional computers, a sufficiently large quantum computer could undermine this security premise. If private keys are exposed, attackers could move assets protected by those keys.
Anthropic focused on the 'point addition' operation used repeatedly in elliptic curve cryptography-based quantum calculations. This process calculates a new point using two points on an elliptic curve and is crucial for cryptographic decryption algorithms. By optimizing this operation, overall calculation efficiency can be improved.
The research team provided Claude with natural language specifications of computational procedures and built a program to verify accuracy while evaluating designs based on the product of qubit count and computation steps. Claude iteratively submitted and refined its designs using validation results. This approach allows AI to explore more efficient methods without requiring humans to design detailed calculation processes.
In one case, Claude spent approximately 53 hours with minimal human intervention searching for a new computational method targeting the resource-intensive point addition operation. The resulting design required 12% more qubits than existing minimum designs but reduced computation steps by about 17 times (to 1/17). This optimization path slightly increased qubit usage while significantly lowering overall computational burden.
However, this research does not mean that Bitcoin or Ethereum's cryptographic systems have been decrypted. Anthropic did not conduct direct attacks on cryptocurrencies in this study and did not specify the full scale of quantum computers needed for decryption. Improving calculation efficiency is distinct from building a functional quantum computer capable of decrypting cryptography.
Current quantum computers remain limited, operating at hundreds to thousands of qubits. Google previously estimated that fewer than 500,000 qubits might be required to decrypt Bitcoin signatures as of March. However, actual requirements may vary based on calculation methods and error correction, making simple comparisons challenging.
Anthropic also suggested future possibilities where AI could optimize not just individual quantum computing procedures but entire system designs tailored to specific problems. This could expand AI's role from optimizing software-level algorithms to influencing broader quantum hardware design.
The research demonstrates that AI can improve quantum calculation efficiency rather than posing an immediate threat to cryptocurrency security. However, continued optimization using AI may influence the speed of quantum computer practicality and long-term cryptographic security responses in the future.
Korean Source
This article is an English localization of a Korean-language crypto news report. Original headline: 클로드가 53시간만에 찾아낸 양자 계산법…연산 횟수 1/17 축소