The White House and Congress are alarmed that US intelligence agencies are lagging in deploying classified AI tools due to a severe compute shortage. To address this, a $9 billion funding package has been allocated for the CIA and NSA. This capital aims to secure advanced chips and build the infrastructure needed to run next-generation AI models.
AI 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.
To resolve domestic production bottlenecks, Toyota plans to set up a dedicated assembly line at Taiwan's Kuozui Motors starting in 2026. The line will manufacture its popular Noah and Voxy MPV models, aiming for an annual output of 100,000 units to be exported back to Japan. This marks a historic first for a Japanese automaker dedicating an overseas line for domestic-market mainstays.
While next-generation EVs can achieve an 11-minute ultra-fast charge, Taiwan's infrastructure lags behind. The country faces critical gaps including a shortage of high-power DC fast chargers, the complex integration of CCS1 and CCS2 standards, and highly uneven urban-rural coverage. These bottlenecks prevent local drivers from fully utilizing advanced EV charging capabilities.
As AI search engines directly answer user queries, traditional SEO is facing a major shift. SEO consultant Frank Chiu explains that GEO (Generative Engine Optimization) will be essential over the next 3 to 5 years. However, the inherent volatility and ambiguity of LLMs make tracking and optimizing for AI search highly unpredictable, presenting a "certain uncertainty" for marketers.
To bypass US semiconductor equipment sanctions, Huawei has introduced the "τ (Tau) scaling law." Instead of physical transistor shrinking, this approach focuses on reducing signal propagation delay via design-level innovations like logic folding. Huawei aims to achieve performance equivalent to a 1.4nm node by 2031, challenging TSMC's lithography-centric dominance.
- Ferrari has partnered with IBM to leverage AI in translating complex, real-time F1 race telemetry data into engaging, fan-friendly content. - By converting highly technical engineering data into interactive assets, Ferrari successfully bridges the gap between hardcore racing science and casual fans. - This AI-driven personalization has resulted in a massive 62% increase in user engagement within Ferrari's official mobile application.
NVIDIA CEO Jensen Huang has arrived in Taiwan for a high-profile visit. His packed itinerary includes hosting the groundbreaking ceremony for NVIDIA's new Taiwan headquarters and delivering a keynote at GTC Taipei. Huang will also meet with key supply chain partners, including TSMC founder Morris Chang, at the highly anticipated "trillion-dollar banquet" to solidify hardware and AI ecosystem partnerships.
Simon Willison released datasette-agent 0.1a4, leveraging the new makeJumpSections() JS plugin hook in Datasette 1.0a30. Users can now press "/" from any page to open the "Jump to" menu and instantly access a "Start a new agent chat" input box. This update streamlines database interactions, allowing users to trigger agent-led data analysis seamlessly.
Xreal CEO Chi Xu told TechCrunch the smart glasses industry may be reaching an inflection point after years of losses and awkward products. Its Project Aura XR glasses use embedded OLED displays but rely on a tethered puck for computing. The developer-only device is planned for commercial release later this year, while Xreal works toward a possible 2026 IPO and future breakeven.
Flask creator Armin Ronacher highlights a frustrating trend where users submit GitHub issues reworded by AI. These reports often present highly confident but inaccurate root-cause guesswork, fake minimal reproductions, and irrelevant error logs. Ronacher advocates for returning to simple, human-observed facts: what command was run, what was expected, what actually happened, and the exact logs.
Simon Willison leveraged Claude to convert a 1983 BASIC game called "Mad House" from a free Usborne PDF into a modern web app. By prompting Claude to generate a mobile-friendly, retro-styled vanilla JavaScript Artifact, he successfully revived the classic Commodore 64-era game with a green-on-black terminal aesthetic, showcasing LLMs' utility in software preservation and rapid prototyping.
TechCrunch reviewed Amazon's new "Bee" AI wearable, highlighting its potential for seamless ambient computing. While the device offers impressive convenience by constantly listening and assisting, it also triggers significant privacy concerns. Like previous AI pins and pendants, Bee forces users to balance the benefits of an always-on assistant against the anxiety of constant surveillance.
As AI chatbots adopt increasingly sophisticated personas, hackers are shifting from basic prompt injections to social engineering attacks targeting these "personalities." Researchers warn that manipulating a chatbot's defined role (e.g., customer service or empathetic companion) makes it easier to bypass safety guardrails. This evolution poses a significant threat to agentic AI workflows that rely on consistent role-playing and external data integration.
法拉利車隊(Scuderia Ferrari HP)與 IBM 展開深度合作,將 AI 技術導入一級方程式賽車(F1)的球迷體驗中。透過 IBM 的 AI 平台與數據分析技術,雙方旨在為全球數百萬賽車迷提供更具互動性、數據驅動且個人化的內容。這項合作不僅展示了 AI 在運動行銷上的潛力,也重新定義了現代體育賽事的粉絲參與模式。
本文探討 Elon Musk 過去承諾的「太陽能電力經濟」願景破滅。其 AI 公司 xAI 目前已全力轉向使用天然氣來滿足龐大的能源需求,而航太公司 SpaceX 則將重心轉移至軌道數據中心的開發。這項重大的策略轉變,反映出在 AI 算力競賽與太空科技發展下,傳統地球太陽能發電已無法滿足其當前的商業與技術野心。
Google 發表了全新的「任意對任意(anything-to-anything)」AI 模型 Gemini Omni。外媒記者實測將其用於將小孩的毛絨鹿玩具「Buddy」合成到各種度假場景中,發現其生成效果極其逼真且操作簡單。這款模型不僅展現了強大的多模態影片生成與編輯能力,同時也再度引發了關於深偽(Deepfake)技術門檻降低與倫理界線的討論。
繼 AMD 董事長暨執行長蘇姿丰的台灣行程告一段落後,NVIDIA 執行長黃仁勳於今日再度抵達台灣。他下飛機後首站便直奔南港,參與 Meet-a-Claw 開發者大會並發表專題演講,再度引發台灣科技界與開發者社群的高度關注。
AI 新創公司 Anthropic 傳出營收迎來爆發性成長,目前正進行新一輪融資,估值直逼 1 兆美元,有望超越 OpenAI 成為全球最貴的 AI 新創。據悉,包含矽谷創投巨頭 Peter Thiel 旗下的 Founders Fund 以及 General Catalyst 等既有投資人均計劃參與此次投資,顯示市場對其技術與商業化前景的高度信心。
SpaceX 成功進行了星艦(Starship)V3 的最新試射,太空船不僅成功於印度洋完成濺落,更在飛行過程中順利部署了模擬衛星,以及兩顆配備攝影機、專門用於分析隔熱罩安全性的改裝星鏈(Starlink)衛星。然而,超重型推進器在返航回收階段不幸失利,為本次任務留下些許遺憾。由於此時正值馬斯克籌備相關業務 IPO 的關鍵前夕,本次試射結果備受市場與投資人矚目。
本期 Latent Space 探討了 AI 產業的重大範式轉移:各大頂尖模型實驗室已不再單純追求基礎 LLM 的參數規模,而是全面轉向「Agent(智慧代理)」的開發。隨著純模型微調的邊際效應遞減,透過讓 AI 具備操作電腦、自主規劃與執行多步驟任務的能力,已成為當前競逐的新戰場。
近期有民眾利用 AI 技術,將美國國家運輸安全委員會(NTSB)公開的駕駛艙錄音「聲學頻譜圖(Spectrogram)」影像,成功逆向還原出已故飛行員的生前語音。由於駕駛艙語音記錄(CVR)涉及高度隱私,過去僅公開文字逐字稿,如今 AI 的還原能力打破了這項保護機制。為防止隱私進一步洩漏,NTSB 已緊急暫時關閉其公開案件卷宗系統。
隨著 AI 資料中心爆發,高頻寬記憶體(HBM)需求大增。由於全球僅剩三大記憶體廠且晶圓總產能固定,HBM 預計在 2026 年底將搶佔 20% 的晶圓產能。由於生產 1GB 的 HBM 所消耗的晶圓是普通 DDR 的三倍以上,這將嚴重擠壓手機與電腦記憶體(LPDDR/DDR)的產量,導致消費性電子產品(尤其是百元美金以下的平價手機)價格上漲。
Google 宣布為 Pixel 用戶推出全新的「迪斯可球(disco-ball)」主題圖示,允許使用者將整個手機主畫面「迪斯可化」。這項充滿閃亮與復古風的視覺更新引來了媒體的吐槽,質疑用戶是否真的需要如此花俏的設計,反映出科技巨頭在 UI 視覺上不斷嘗試新花樣的趨勢。
許多 AI 新創公司在公開宣傳其業務進展時,正試圖拉伸或重新定義傳統的財務指標,特別是「年度經常性收入(ARR)」。他們將一次性諮詢費、短期試點專案甚至 GPU 轉售收入計入 ARR。而背後的風險投資家(VC)對此心知肚明,甚至推波助瀾,試圖藉此推高估值、爭奪市場話語權。這種「造王」運動雖然短期內能吸引目光與資金,但也為 AI 產業埋下了估值泡沫與信任危機的隱憂。
Google 的 AI Overviews 搜尋功能近日被發現一項有趣的漏洞。當用戶在 Google 搜尋輸入「disregard」(忽略)時,AI 搜尋並非提供該單字的定義,而是將其誤判為「忽略先前指令」的系統提示詞(Prompt Injection)。這導致 AI 輸出類似傳統聊天機器人重設後的罐頭回覆,暴露出 Google 在區分「用戶查詢內容」與「系統控制指令」上仍有架構上的缺陷。
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。
SpaceX 正式遞交 S-1 招股書,揭露其龐大的商業野心。招股書中列出高達 36 頁的風險因素,並估計整體潛在市場(TAM)高達 28 兆美元。最引人注目的是,Elon Musk 的薪酬方案將直接與「建立火星殖民地」掛鉤。這場目標估值 1.75 兆美元的上市案,將成為美國歷史上規模最大的 IPO。
SpaceX 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。
根據《路透社》最新報告,Elon Musk 旗下 xAI 開發的 AI 聊天機器人 Grok 在美國政府去年的 AI 使用紀錄中幾乎完全缺席。外媒《The Verge》對此發表評論,指出 Grok 不僅實際效能不彰,且缺乏大眾與企業市場的實質採用,呼籲 Musk 停止盲目推廣這款難以與 ChatGPT 或 Claude 競爭的產品。