arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.24904cs.AIcs.CR
收藏

人工智能可信性的基础

On the Foundations of Trustworthy Artificial Intelligence

ARC

TJ Dunham

AI总结:

本文提出确定性推理是可信AI的必要充分条件,并通过信任熵量化非确定性的成本,证明验证失败概率精确为1-2^{-H_T}。通过构建纯整数推断引擎,验证了跨架构模型的确定性,展示了AI信任的本质是算术问题。

摘要:

我们证明平台确定性推理是可信AI的必要且充分条件。我们将这一观点形式化为确定性学说,并引入信任熵来量化非确定性的成本,证明验证失败概率恰好等于1-2^{-H_T}。我们证明了确定性-验证崩溃:在确定性下验证需要O(1)次哈希比较;没有确定性,验证者将面临不可处理的成员问题。IEEE 754浮点运算本质上违反了确定性要求。我们通过构建一个纯整数推断引擎来解决这一问题,该引擎在ARM和x86架构上实现了位级相同的输出。在82个跨架构测试中,模型参数最多达6.7B,观察到零哈希不匹配。四个地理分布的节点产生相同输出,通过356个链上证明交易得到验证。每种主要的AI系统信任属性(公平性、鲁棒性、隐私、安全性和对齐)都预设了平台确定性。我们的系统,99,000行Rust代码部署在三个大洲,证明AI信任本质上是算术问题。

英文摘要:

We prove that platform-deterministic inference is necessary and sufficient for trustworthy AI. We formalize this as the Determinism Thesis and introduce trust entropy to quantify the cost of non-determinism, proving that verification failure probability equals 1 - 2^{-H_T} exactly. We prove a Determinism-Verification Collapse: verification under determinism requires O(1) hash comparison; without it, the verifier faces an intractable membership problem. IEEE 754 floating-point arithmetic fundamentally violates the determinism requirement. We resolve this by constructing a pure integer inference engine that achieves bitwise identical output across ARM and x86. In 82 cross-architecture tests on models up to 6.7B parameters, we observe zero hash mismatches. Four geographically distributed nodes produce identical outputs, verified by 356 on-chain attestation transactions. Every major trust property of AI systems (fairness, robustness, privacy, safety, alignment) presupposes platform determinism. Our system, 99,000 lines of Rust deployed across three continents, establishes that AI trust is a question of arithmetic.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑