來源New York Times Technology•較早收集於 8h
研究人員啟動 40 億美元計畫研發自我改進 AI
前科技巨頭研究人員投入 40 億美元賭 AI 能自我進化。了解他們如何實現研發週期的自動化。
30 秒速覽
有什麼變化
Recursive Superintelligence 由前 Google、Meta 和 OpenAI 的研究人員創立。
為什麼重要
此舉標誌著 AI 領域向「遞迴自我改進」邁出重要一步，可能加速通往 AGI 的進程。這凸顯了業界致力於減少模型訓練生命週期中對人類依賴的趨勢。
下一步行動
密切關注 Recursive Superintelligence 發布的白皮書，以了解他們在自動化模型架構搜尋方面的技術路徑。
誰應關注：Researchers & Academics
關鍵要點
- •Recursive Superintelligence 由前 Google、Meta 和 OpenAI 的研究人員創立。
- •獲得 40 億美元資金，專注於開發自我改進的 AI 系統。
- •核心目標是實現人工智慧創建與擴展過程的自動化。
關鍵數字$500 million$650 million$4 billion40 億
深度解析
背景與延伸：來自公開資料，非原文內容。引用 13 個來源。
增強重點摘要
- •Recursive Superintelligence (RSI) was co-founded by Richard Socher, former chief scientist at Salesforce, and Tim Rocktäschel, an AI professor at University College London and former Google DeepMind scientist.
- •The company, incorporated in London approximately four months prior to April 2026, has secured between $500 million and $650 million in seed funding, not $4 billion, which is its reported valuation.
- •The funding round was led by Google Ventures (GV) and Greycroft, with additional investment from chipmakers Nvidia and AMD Ventures.
- •RSI's core mission is to develop AI systems that can autonomously improve their own architecture, training methods, evaluation processes, and research direction without continuous human oversight.
- •The company, currently operating with a team of less than 30 researchers and engineers across offices in London and San Francisco, remains in the research phase and has not yet launched commercial products or generated revenue.
技術深入
- RSI's approach centers on recursive self-improvement through open-ended algorithms to drive continuous innovation.
- The foundational architecture is based on the transformer model, optimized for context-aware language processing, multi-modal data integration, and recursive learning.
- The company aims to automate the entire AI development pipeline, including improving its own architecture, training methods, evaluation processes, and research direction.
- Key areas of expertise within the team include open-ended algorithms, quality diversity algorithms, AI-generating algorithms, self-improving coding agents, automated red teaming, capability discovery, prompt engineering automation, foundational world models, deep learning in NLP, vision transformers, and retrieval-augmented generation.
- RSI plans to run its first "Level 1" autonomous training system, indicating a phased approach to achieving full self-improvement.
前景展望基於引用來源的 AI 分析
Recursive Superintelligence will significantly accelerate the pace of AI development beyond current human-driven methods.
By automating the entire AI development pipeline, including research and architecture design, RSI aims to create a positive feedback loop that compounds improvements faster than human-led efforts.
The success of Recursive Superintelligence could lead to an 'intelligence explosion' as theorized by early AI pioneers.
The concept of recursive self-improvement inherently suggests a rapid, exponential increase in AI capabilities, potentially leading to superintelligence that far surpasses human understanding and control.
The development of self-improving AI systems like those pursued by Recursive Superintelligence will necessitate new paradigms for AI safety and governance.
The potential for unforeseen evolution and surpassing human control in self-improving systems raises significant ethical and safety concerns, requiring robust safeguards and regulatory frameworks.
時間線
2025-12
Recursive Superintelligence Inc. incorporated in London.
2026-04-18
Reports begin to surface about Recursive Superintelligence raising at least $500 million at a $4 billion valuation.
2026-05-13
Recursive Superintelligence officially emerges from stealth, announcing it has raised over $650 million at a $4.65 billion valuation.
- 2025-12Recursive Superintelligence Inc. incorporated in London.
- 2026-04-18Reports begin to surface about Recursive Superintelligence raising at least $500 million at a $4 billion valuation.
- 2026-05-13Recursive Superintelligence officially emerges from stealth, announcing it has raised over $650 million at a $4.65 billion valuation.
來源 (13)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
1Google Search Sourcevertexaisearch.cloud.google.com2Google Search Sourcevertexaisearch.cloud.google.com3Google Search Sourcevertexaisearch.cloud.google.com4Google Search Sourcevertexaisearch.cloud.google.com5Google Search Sourcevertexaisearch.cloud.google.com6Google Search Sourcevertexaisearch.cloud.google.com7Google Search Sourcevertexaisearch.cloud.google.com8Google Search Sourcevertexaisearch.cloud.google.com9Google Search Sourcevertexaisearch.cloud.google.com10Google Search Sourcevertexaisearch.cloud.google.com11Google Search Sourcevertexaisearch.cloud.google.com12Google Search Sourcevertexaisearch.cloud.google.com13Google Search Sourcevertexaisearch.cloud.google.com
AI 週報
閱讀本週精選 AI 大事摘要 →
AI 策展新聞聚合。所有內容版權歸原始發布者所有。
原始來源： New York Times Technology ↗
每週電子報
每週一封，可隨時退訂。