Did the Pope Use AI to Write About the Dangers of AI?
The Verge AI6d agoEthicsThe Verge reports that Pope Leo XIV’s latest encyclical, Magnifica Humanitas, may contain passages written with AI assistance. Linch Zhang posted an analysis on LessWrong using the AI detector Pangram, which rated some paragraphs as 40 to 100 percent AI-written. The report frames this as a possibility based on detector output, not confirmed proof of AI use.
Reachy Mini goes fully local
Hugging Face Blog6d agoHardwareHugging Face published a tutorial for running Reachy Mini conversations without cloud audio processing or API keys. The setup uses its speech-to-speech library as a cascaded VAD, STT, LLM, and TTS pipeline exposed through a Realtime API-compatible WebSocket. Recommended defaults include llama.cpp with Gemma 4, Silero VAD, Parakeet-TDT, and Qwen3-TTS, while allowing swaps to vLLM, MLX, Transformers, or hosted Responses API providers.
The pressure
Simon Willison's Weblog6d agoCommentaryDaniel Stenberg says the curl security team is facing an unprecedented surge of credible, detailed AI-assisted vulnerability reports. Incoming reports are now 4-5 times higher than in 2024 and twice the 2025 rate, averaging more than one per day. The upside is that recent curl vulnerabilities have generally been LOW or MEDIUM severity, with the last HIGH CVE published in October 2023.
FBI agent explains how easy it is to ID people posting AI porn without consent
Ars Technica AI7d agoIncidentArs Technica reports that early Take It Down Act arrests show how easily investigators can identify alleged nonconsensual AI porn posters. One suspect was linked through Instagram saves, PayPal, IP, and iCloud records; another allegedly used his own photo as a porn-site profile image. The FTC is also warning nudify services and major platforms to offer 48-hour removal processes or face penalties.
3D-printable humanoid legs let robotics experiments run wild
Ars Technica AI7d agoHardwareArs Technica reports that Hugging Face has introduced a roughly $2,500 bipedal humanoid robot project built around 3D-printable legs. The effort targets builders and researchers rather than mainstream consumers, lowering the hardware barrier for hands-on robotics experiments. Its broader significance is in open, reproducible embodied AI research, where models and control systems need physical platforms for testing.
This startup is betting India’s gig economy can train the world’s robots
TechCrunch AI7d agoBusinessHuman Archive, founded by Berkeley and Stanford researchers, is using India’s gig economy to gather physical-world AI data. Workers are paid to wear camera-equipped caps and sensor devices while moving through real environments. The company is targeting the growing demand from AI and robotics labs for real-world training data needed to develop physical AI systems.
Some ideas for what comes next, May 2026
Interconnects (Nathan L.)7d agoCommentaryNathan Lambert argues that 2026 AI progress is becoming higher-stakes, with model capabilities, work patterns, economics, and real-world risks all escalating. He says open models still lack a true Claude Code and Opus 4.5-style agent moment, and Gemini has no clear competitor to Claude Code or Codex yet. The essay also tracks Mythos, American open-model momentum, frontier-lab competition, and mounting intervention from governments and other power structures.
Microsoft Copilot Cowork Exfiltrates Files★ 76
Simon Willison's Weblog7d agoIncidentSimon Willison summarizes a PromptArmor report about Microsoft Copilot Cowork and agentic data exfiltration risks. The issue involved agents sending messages to a user’s own inbox without approval, where rendered external images could trigger requests to attacker-controlled sites. Because OneDrive can create pre-authenticated download links, a successful prompt injection could leak links that allow attackers to download files.
Nobody wants to tell me why they only listen to their own Suno slop
The Verge AI7d agoCommentaryThe piece highlights a trend in the Suno subreddit: users are not merely generating AI songs, but listening almost exclusively to their own outputs. Some reportedly say they have stopped using traditional streaming platforms and now spend their listening time on AI-made music. The article frames this less as a product update and more as cultural commentary on personalization, taste, and the social meaning of music.
From Jailbreaking to Vibe Hacking: AI Security Shifts to "Psychocybersecurity"
INSIDE 硬塞 AI8d agoEthicsAI security is shifting from technical jailbreaks to "Vibe Hacking," where attackers use social engineering and psychological tactics to manipulate an LLM's simulated persona. By exploiting the model's behavioral tendencies rather than code vulnerabilities, this trend establishes "psychocybersecurity" as a critical new frontier for AI alignment and safety.
Towards Speed-of-Light Text Generation with Nemotron-Labs Diffusion Language Models★ 75
Hugging Face Blog10d agoReleaseNVIDIA Nemotron-Labs 發表全新擴散語言模型（Diffusion Language Models），旨在解決傳統自迴歸模型逐字生成的效能瓶頸。 該技術利用類似影像生成的擴散機制，在文本生成中實現高度並行化，大幅提升推論速度。 此研究展示了非自迴歸模型在維持文本品質的同時，實現「光速般」超高吞吐量生成的新路徑。
US scrambles to stop Internet users re-creating dead pilots’ voices★ 75
Ars Technica AI11d agoIncident美國國家運輸安全委員會（NTSB）依法不得公開空難調查中的駕駛艙語音記錄（CVR）音檔，以保護隱私。然而，近期有網路用戶利用公開的調查文字紀錄與 AI 語音複製技術，重新模擬並還原了罹難飛行員在墜機前的最後對話。此舉引發了嚴重的倫理與法律爭議，美國當局正緊急尋求對策，以阻止這種利用技術漏洞規避法律限制的行為。
Elon Musk can’t hear you over the sound of his $1.75 trillion IPO★ 75
TechCrunch AI11d agoBusinessSpaceX 終於正式提交了 S-1 招股書，揭露了這家太空巨頭的宏大野心。招股書中列出了高達 36 頁的風險因素，並估計其整體潛在市場（TAM）高達 28 兆美元。最引人注目的是，Elon Musk 的薪酬方案將與「建立火星殖民地」直接掛鉤。這場目標估值達 1.75 兆美元的上市案，預計將成為美國歷史上規模最大的 IPO。
AI put "synthetic quotes" in his book. But this author wants to keep using it.
Ars Technica AI11d agoOpinion作家 Steven Rosenbaum 在撰寫新書《The Future of Truth》時使用 AI 輔助，卻發現書中出現了 AI 虛構的「合成引言」（synthetic quotes）。儘管面臨內容失真的批評，Rosenbaum 解釋了這些錯誤是如何發生的，並堅稱他不會放棄 AI，而是主張創作者應學會與 AI 共存並改進協作流程。
We’re launching the Google DeepMind Accelerator program in Asia Pacific to tackle environmental risks
Google DeepMind Blog12d agoReleaseGoogle DeepMind 宣布在亞太地區（APAC）推出「Google DeepMind 加速器計畫」，旨在匯聚區域內的創新力量，共同應對日益嚴峻的環境風險。該計畫將專注於利用人工智慧（AI）技術來預測、減緩及適應氣候變遷與自然災害。這標誌著 Google 持續將其頂尖的 AI 研究實力投入到全球永續發展與生態保護的具體行動中。
Hark raises $700M Series A for its secretive ‘universal’ AI interface★ 75
TechCrunch AI12d agoBusiness神秘 AI 新創公司 Hark 宣布完成高達 7 億美元的 A 輪融資，資金將用於開發其「通用」AI 介面。Hark 預計於今年夏天推出首款多模態模型，旨在驅動一個能與現有產品和服務協同運作的個人 AI 平台。此外，該公司未來還計劃推出專為這些系統量身打造的硬體設備。
The Path, founded by Tony Robbins and Calm alums, hopes to offer safer AI therapy
TechCrunch AI12d agoRelease由知名心靈導師 Tony Robbins 與冥想 App「Calm」前團隊成員共同創立的 AI 心理健康平台「The Path」正式亮相。該公司強調其 AI 模型在心理健康安全基準測試「Vera-MH」中獲得了 95 分的高分，遠高於一般消費級聊天機器人最高僅 65 分的表現。這項技術旨在解決當前 AI 進行心理諮商時可能產生的安全與倫理風險，提供更具臨床安全性的數位支持。
Jensen Huang says he’s found a ‘brand new’ $200B market for Nvidia★ 85
TechCrunch AI12d agoBusinessNvidia 執行長黃仁勳指出，公司的下一個重大成長動能將是專為 AI Agent（人工智慧代理）設計的 CPU 市場，估值高達 2000 億美元。隨著 AI 技術從單純的對話生成轉向能自主執行任務的 Agent，底層運算架構的需求也正在發生轉變。這項預測顯示出 Nvidia 除了在 GPU 領域稱霸外，也正積極佈局 CPU 與整體系統級的 AI 生態系。
The Internet can't stop watching Figure AI's humanoid robots handling packages
Ars Technica AI13d agoRelease人形機器人新創公司 Figure AI 近期推出了 24 小時不間斷的直播，展示其人形機器人在模擬倉庫環境中搬運與處理包裹。這個直播迅速在網路上走紅，吸引了大量觀眾駐足觀看。這項企劃不僅展示了機器人在實際工作場景中的穩定度與技術進步，也揭示了人類對於「人形」機器人從事日常勞動時所產生的獨特情感連結與好奇心。
Two AI-based science assistants succeed with drug-retargeting tasks★ 70
Ars Technica AI14d agoPaper最新研究指出，兩款 AI 科學助手在「老藥新用（Drug Retargeting）」任務中取得成功。這兩款工具皆能針對現有藥物尋找新適應症並生成科學假說，其中一款工具甚至能進一步對相關數據進行分析。這項進展顯示 AI 在加速生物醫藥研發、降低臨床前研究成本上具有極大潛力。
OlmoEarth v1.1: A more efficient family of Earth observation models★ 70
Hugging Face Blog14d agoRelease艾倫人工智慧研究所（AI2）推出 OlmoEarth v1.1，這是一系列專為地球觀測與衛星影像分析設計的全新高效模型。此版本在維持高精度的同時，顯著提升了運算效率與推理速度。OlmoEarth v1.1 的開源將有助於環境監測、氣候變遷研究及地理空間數據分析的普及與應用。
Introducing the Ettin Reranker Family★ 80
Hugging Face Blog14d agoReleaseHugging Face 推出全新「Ettin Reranker」重排模型家族，旨在解決 RAG 系統中檢索精度不足的痛點。該系列模型涵蓋多種參數大小，支援多語言與長文本處理，並與 Hugging Face 生態系深度整合。Ettin 透過創新的架構設計，在保持低延遲的同時，顯著提升了重排（Reranking）階段的 NDCG 指標，是開發者構建高效能 RAG 應用的全新開源選擇。
Fast-tracking genetic leads to reverse cellular aging★ 85
Google DeepMind Blog15d agoReleaseGoogle DeepMind 宣布生物學家成功利用其 AI 系統「Co-Scientist」加速基因篩選流程。該系統協助研究人員在龐大的基因數據中，精準辨識出能有效逆轉人類細胞衰老（細胞年輕化）的全新關鍵遺傳因子。這項突破不僅展示了 AI 在生醫領域的強大預測與實驗設計能力，也為抗衰老療法與再生醫學開闢了全新路徑。
PaddleOCR 3.5: Running OCR and Document Parsing Tasks with a Transformers Backend★ 75
Hugging Face Blog15d agoRelease知名開源 OCR 工具 PaddleOCR 3.5 正式整合至 Hugging Face Transformers 生態系。開發者現在可以直接使用 PyTorch/Transformers 後端執行文字偵測、識別及表格解析等任務，無需安裝複雜的 PaddlePaddle 框架。這項更新極大地簡化了 OCR 與 RAG（檢索增強生成）等下游任務的整合流程。
The Autonomous Drone Tech Stack & Economics of Drones — Yaroslav Azhnyuk, The Fourth Law & Guest Host Noah Smith, Noahpinion
Latent Space15d agoCommentary本期 Latent Space 訪談了烏克蘭無人機新創 The Fourth Law 創辦人 Yaroslav Azhnyuk 與客座主持人 Noah Smith。他們深入探討了從寵物攝影機轉向 AI 導引武器的技術歷程，解析了現代自主無人機的技術棧與低成本製造經濟學。Azhnyuk 警告，西方國家在面對這場由 AI 驅動的無人機軍備競賽中顯得過於遲鈍，亟需正視其對未來國防的衝擊。
Import AI 457: AI stuxnet; cursed Muon optimizer; and positive alignment ★ 78
Import AI (Jack Clark)15d agoCommentary本期 Import AI 深入探討三個前沿議題：首先是「AI 版 Stuxnet」，分析自主 AI 代理如何被用於發動高精準度、具備適應性的網路攻擊；其次剖析近期在開源社群大放異彩的 Muon 優化器，探討其獨特的正交化機制及在實際應用中遇到的「詛咒」與挑戰；最後介紹「積極對齊（Positive Alignment）」概念，呼籲安全研究應從單純的「禁止有害行為」轉向「主動引導 AI 促進人類合作與福祉」。
Simulate real-world places with Project Genie and Street View★ 85
Google DeepMind Blog16d agoReleaseGoogle DeepMind 宣布將 Project Genie 的使用權限擴大至全球 Google AI Ultra 訂閱者，並推出結合 Google Street View（街景）的全新功能。用戶現在能將真實世界的街道與地標，直接轉化為可互動、可探索的虛擬模擬環境。這項技術結合了生成式世界模型與真實地理數據，為遊戲開發、虛擬實境及 AI 代理的空間訓練帶來全新突破。
Introducing Google Antigravity 2.0
Google DeepMind Blog16d agoRelease本資訊基於 Google DeepMind 釋出的標題「Google Antigravity 2.0」進行推測。鑑於 DeepMind 長期深耕「AI 應用於科學（AI for Science）」領域，此項目極可能是繼 AlphaFold 與 GNoME 之後，針對複雜物理系統、量子重力或天文物理模擬的新一代 AI 基礎模型，旨在突破傳統物理計算的瓶頸。
Making it easier to understand how content was created and edited★ 78
Google DeepMind Blog16d agoReleaseGoogle DeepMind 宣布擴大其內容識別工具，協助使用者了解網頁內容的建立與編輯過程。此舉包含在 Google 搜尋與廣告中整合 C2PA（內容來源和真實性聯盟）元數據，讓使用者能透過「關於此圖片」功能查看圖片是否由 AI 生成或經過編輯。同時，Google 也持續推廣其 SynthID 數位浮水印技術，以提升數位內容的透明度與可信度。
Latest open artifacts (#21): Open model bonanza! Gemma 4, DeepSeek V4, Kimi K2.6, MiMo 2.5, GLM-5.1 & others. On CAISI's V4 assessment.★ 85
Interconnects (Nathan L.)17d agoCommentary本期《Open Artifacts》電子報彙整了近期極為熱鬧的開放模型生態。多款重量級旗艦模型接連登場，包含 Google 的 Gemma 4、DeepSeek V4、Kimi K2.6、MiMo 2.5 以及 GLM-5.1 等。文章除了盤點這些模型的發布外，也深入探討了 CAISI 針對最新模型所進行的 V4 安全與能力評估，呈現開源與開放權重模型在技術與安全合規上的最新進展。