斯坦福区块链和AI会议海报

3月中旬，斯坦福大学举办了一个区块链和AI会议，将教授，初创企业首席执行官和风险投资家（VCS）汇集在一起​​。

In mid-March, Stanford University hosted a Blockchain and AI conference, bringing together professors, startup CEOs, and venture capitalists (VCs). The event aimed to highlight the convergence of two major technologies: blockchain and AI. However, the conference could have benefited from highlighting Bitcoin and AI further, given Bitcoin's market dominance and the emerging innovations on Bitcoin Layer 2 solutions.

3月中旬，斯坦福大学举办了一个区块链和AI会议，将教授，初创企业首席执行官和风险投资家（VCS）汇集在一起​​。该活动旨在强调两种主要技术的融合：区块链和AI。但是，鉴于比特币的市场优势以及比特币第2层解决方案的新兴创新，会议可能会从突出比特币和人工智能进一步中受益。

One of the main challenges with the conference was that blockchain and AI have largely evolved as separate disciplines—with different investors, entrepreneurs, academics, and communities. While the idea was to merge the two fields, many speakers remained focused on their own domain, failing to establish clear connections between them. Perhaps a more fitting title would have been the Blockchain OR AI Conference.

会议的主要挑战之一是区块链和AI在很大程度上发展为单独的学科，包括不同的投资者，企业家，学者和社区。虽然这个想法是将这两个领域合并，但许多发言人仍然专注于自己的领域，未能在它们之间建立明确的联系。也许更合适的头衔是区块链或AI会议。

For example, a venture investor presented an overview of the AI industry, showcasing impressive advancements in image, audio, and code generation. Meanwhile, a DeepMind researcher discussed adversarial machine learning, a phenomenon where slight manipulations to input data can drastically alter an AI's output. One striking example involved modifying just a few pixels in an image of a cat—causing the AI to misclassify it as guacamole.

例如，一个风险投资者概述了AI行业，展示了图像，音频和代码生成的令人印象深刻的进步。同时，一位深度研究人员讨论了对抗机器学习，这是一种现象，其中轻微的操作输入数据可以大大改变AI的输出。一个引人注目的示例涉及在猫的图像中仅修改几个像素，从而导致AI将其错误分类为鳄梨调味酱。

On the blockchain side, discussions revolved around various protocols, but much of the technology remains highly experimental—or, in some cases, non-existent yet. Blockchain-AI integrations are still in their infancy, with practical implementations yet to emerge.

在区块链方面，讨论围绕各种协议进行了讨论，但是许多技术仍然具有很高的实验性，或者在某些情况下尚不存在。区块链-AI集成仍处于起步阶段，实际实施尚未出现。

Proof of Computation

计算证明

One of the more insightful contributions came from Dan Boneh, an applied cryptographer at Stanford. He discussed SNARKs (succinct non-interactive arguments of knowledge) and zero-knowledge proofs, which address a fundamental cryptographic problem: proving knowledge of a computation in an efficient way.

最有见地的贡献之一来自斯坦福大学应用的密码师Dan Boneh。他讨论了Snarks（知识的简洁性非交互论点）和零知识证明，这解决了一个基本的加密问题：以有效的方式证明计算的知识。

This principle is well-established in both blockchain and cryptography. For example: It’s computationally expensive to factor a large number into its two prime components, but verifying via multiplication is computationally cheap. It’s expensive to find a block header whose hash meets a target threshold, but verifying that it does is inexpensive.

该原理在区块链和密码学中都具有良好的成就。例如：将大量数量计入其两个主要组件的计算昂贵，但是通过乘法进行验证在计算上是便宜的。找到一个标题符合目标阈值的块标头很昂贵，但是验证它确实很便宜。

This asymmetry between computation and verification is critical in blockchain systems, where nodes constantly validate the work of others. In Bitcoin, nodes verify signatures and miners' proof of work. SNARKs extend this concept, enabling cryptographic proofs that are verifiable without revealing sensitive data.

在区块链系统中，计算和验证之间的不对称性至关重要，在区块链系统中，节点不断验证他人的工作。在比特币中，节点验证签名和矿工的工作证明。 Snarks扩展了此概念，从而实现了可以证明的加密证明，而无需揭示敏感数据。

As AI agents become increasingly autonomous, a major challenge will be verifying computation while preserving privacy. Many are hesitant to upload sensitive data to OpenAI due to concerns over data security and prefer using their own models.

随着AI代理的越来越自主，一个重大挑战将是在保留隐私时验证计算。由于对数据安全性的疑虑，并且更喜欢使用自己的模型，因此许多人不愿将敏感数据上传到OpenAI。

This creates a market demand for privacy-preserving verification—a mechanism that allows users to prove an AI model executed a computation correctly without revealing the underlying data. Such a solution could unlock AI applications in domains like healthcare, defense, and finance, where data security is paramount. This will likely become a multi-billion-dollar industry in the next decade.

这创建了对隐私验证的市场需求，该机制允许用户证明正确执行计算的AI模型而不揭示基础数据。这样的解决方案可以在数据安全性至关重要的情况下在医疗保健，国防和金融等领域中解锁AI应用程序。在未来十年中，这可能会成为数十亿美元的行业。

Interestingly, this concept originates from blockchain via networks to implement such cryptographic techniques. As Boneh pointed out, the idea of one machine cheaply verifying the expensive computation done by another emerged out of Bitcoin. But it may have a second, large application in AI.

有趣的是，该概念源自通过网络区块链来实施此类加密技术。正如Boneh所指出的那样，一台机器的想法便宜地验证了另一个机器从比特币中出现的昂贵计算。但是它可能在AI中有第二个大型应用。

I hope to see future conferences place a greater emphasis on Bitcoin's contributions to these fields. BitVM, for example, leverages ideas from zero-knowledge proofs to create bridges between Bitcoin and new Layer 2 protocols—potentially enabling AI agents to interact with Bitcoin's ecosystem.

我希望看到未来的会议更加重视比特币对这些领域的贡献。例如，BITVM利用零知识证明的想法来创建比特币和新第2层协议之间的桥梁，这有可能使AI代理与比特币的生态系统进行交互。