AI should be illegal until we figure out how to deal with its consequences 在弄清楚如何应对AI的后果之前，AI应当被禁止

TL;DR

• The author argues AI development should be legally restricted until society can adequately address its social, economic, and psychological consequences
• AI's rapid advancement outpaces societal adaptation, creating profound career uncertainty as people invest years learning skills for potentially obsolete professions
• A key insight: any job AI can create, it can also automate, making technical literacy alone insufficient for job security
• The ultra-rich may shift from valuing currency to controlling physical resources (energy, land, raw materials) needed to operate AI infrastructure, concentrating power beyond current economic frameworks
• Legal prohibition could establish social norms around human-made work and spur development of verification systems (cryptographic proof, hardware attestations) even if full compliance is unachievable

Why It Matters

This piece challenges the dominant narrative of unrestricted AI acceleration by framing the issue as one of societal pacing rather than technological capability. For AI practitioners and policymakers, it raises critical questions about whether the industry's current trajectory accounts for second-order effects on employment, wealth distribution, and human culture. The argument for targeted restrictions with medical exceptions offers a concrete policy framework worth engaging with seriously.

Technical Details

• Proposes a legal prohibition framework with exceptions for medically beneficial applications (e.g., cancer research), suggesting a tiered regulatory approach rather than a blanket ban
• Envisions cryptographic proof of creative processes, trusted certification systems, and hardware-based attestations as mechanisms to verify human authorship in a post-AI creative economy
• Identifies a structural economic shift: wealth concentration may move from financial capital to physical resource control (energy, land, raw materials, infrastructure) as AI automates productive labor
• Argues that AI's defining advantage—learning and improving faster than humans—creates a recursive automation problem where even AI-adjacent jobs become automatable over time
• Suggests that social norms and legal restrictions can create value even without full compliance, drawing parallels to other illegal endeavors that persist despite prohibition

Industry Insight

• AI companies should proactively engage with societal impact frameworks rather than treating regulation as purely adversarial; preemptive responsibility could shape more sustainable policy outcomes
• The resource-control thesis suggests that long-term AI economic power may reside with those controlling energy and physical infrastructure, not just model developers—investment and partnership strategies should account for this shift
• Opportunities exist in developing verification and provenance technologies (human authorship certification, cryptographic process proof) that could become valuable infrastructure in a regulated AI ecosystem

摘要

作者主张，在社會能夠充分應對人工智能的社會、經濟和心理影響之前，應通過法律手段限制人工智能的發展
人工智能的快速發展超出了社會適應的速度，導致人們投入多年學習的技能可能很快過時，從而產生深遠的就業不確定性
一個關鍵見解：人工智能能夠創造的任何工作，它也能自動化，因此僅具備技術素養不足以保障就業安全
超級富豪可能從重視貨幣轉向控制運行人工智能基礎設施所需的實體資源（能源、土地、原材料），從而將權力集中到現有經濟框架之外
法律禁令可能確立關於人工創造工作的社會規範，並推動驗證系統（密碼學證明、硬件認證）的發展，即使完全合規無法實現

深度分析

簡要概括

• 作者主張，在社會能夠充分應對人工智能的社會、經濟和心理影響之前，應通過法律手段限制人工智能的發展
• 人工智能的快速發展超出了社會適應的速度，導致人們投入多年學習的技能可能很快過時，從而產生深遠的就業不確定性
• 一個關鍵見解：人工智能能夠創造的任何工作，它也能自動化，因此僅具備技術素養不足以保障就業安全
• 超級富豪可能從重視貨幣轉向控制運行人工智能基礎設施所需的實體資源（能源、土地、原材料），從而將權力集中到現有經濟框架之外
• 法律禁令可能確立關於人工創造工作的社會規範，並推動驗證系統（密碼學證明、硬件認證）的發展，即使完全合規無法實現

重要性

本文通過將問題框架為社會節奏而非技術能力，挑戰了不受限制的人工智能加速的主流敘事。對於人工智能從業者和政策制定者而言，它提出了關鍵問題：當前行業軌跡是否考慮了對就業、財富分配和人類文化的二階影響。針對性限制（醫療例外）的論證提供了一個值得認真對待的具體政策框架。

技術細節

• 提出帶有醫療有益應用例外（如癌症研究）的法律禁令框架，建議採用分層監管方法而非全面禁令
• 設想創意工作的密碼學證明