1. Thrive创始人Joshua Kushner批评硅谷VC的AI狂热 Thrive Capital创始人Joshua Kushner在其首封投资者信中，批评硅谷风险投资公司对AI的狂热情绪，认为行业过度关注技术细节而忽视长期方向。他主张Thrive采用集中投资策略，而非硅谷流行的'撒网式'投资。Thrive约90%资金投入每只基金的前15大投资，其2022年5.16亿美元早期基金（投资OpenAI、Anduril、SpaceX）目前价值超37亿美元。
EN: Joshua Kushner criticized Silicon Valley VCs for AI euphoria in Thrive Capital's first investor letter, warning that 'letting excitement weaken investment discipline would be a serious mistake.' Thrive concentrates ~90% of capital in top 15 investments per fund. Its 2022 $516M early-stage fund (invested in OpenAI, Anduril, SpaceX) is now worth over $3.7B.
🔗 https://techcrunch.com/2026/08/14/thrives-joshua-kushner-chides-silicon-valley-vcs-over-ai-euphoria/
1. 自动驾驶卡车正式获准在加州高速公路上测试 加州DMV于4月28日批准新规，解除对超过1万磅无人驾驶车辆在公共道路测试的禁令，Aurora Innovation和Kodiak AI获得测试许可。Kodiak已在Mountain View附近启动测试，但需配备安全驾驶员，且禁止在限速25英里/小时以下道路行驶。Teamsters California上周起诉DMV，指控其未研究经济影响。两家公司此前因禁令主要在德州测试，Aurora于2025年5月在德州推出商业服务，Kodiak于2025年1月在德州Permian Basin开始商业运营。
EN: California DMV approved new rules on April 28, lifting the ban on driverless vehicles over 10,000 pounds. Aurora and Kodiak received permits, requiring safety operators and prohibiting roads with speed limits under 25 mph. Kodiak started testing near Mountain View. Teamsters California sued the DMV last week over economic impact concerns. Both companies previously focused on Texas due to the ban.
🔗 https://techcrunch.com/2026/08/14/self-driving-trucks-are-officially-testing-on-california-highways/
2. Uber与Pony.ai计划将2000辆Robotaxi引入欧洲 Uber与自动驾驶公司Pony.ai宣布扩大合作，计划将2000辆Robotaxi引入欧洲市场。该合作将从此前克罗地亚萨格勒布的首个市场扩展至另外四个欧洲城市，标志着自动驾驶出行服务在欧洲的规模化落地。
EN: Uber and Pony.ai plan to deploy 2,000 robotaxis in Europe, expanding from Zagreb, Croatia to four additional European cities. This marks a significant scale-up of autonomous ride-hailing in Europe.
🔗 https://techcrunch.com/2026/08/14/uber-and-pony-ai-plan-to-bring-2000-robotaxis-to-europe/
3. 谷歌将允许用户移除AI生成内容的可见水印 谷歌于2026年8月14日宣布，将允许用户从其AI生成内容中移除可见水印。该公司表示，关闭此设置不会影响用于识别AI生成文件的不可见基准标记，引发对AI内容溯源和治理的讨论。
EN: Google announced on August 14, 2026 that users can now remove visible watermarks from AI-generated content. The company stated that disabling this setting does not affect invisible baseline markers used to identify AI-generated files.
1. 法国AI初创Kog披露GPU推理优化新进展：聚焦大模型与软件工程用例 法国初创公司Kog旨在通过软件优化从现有GPU中榨取更多推理性能，其CEO Gaël Delalleau透露，基于早期反馈，公司将软件工程作为首要用例，并放弃微调小模型路线，转而加速大模型开发。公司已开源其2B参数演示模型Laneformer 2B，实现单请求3000 tokens/秒，目标为'30倍更快LLM推理'。Kog获得Varsity VC等投资的种子轮。
EN: French startup Kog focuses on software optimization to extract more inference performance from existing GPUs. CEO Gaël Delalleau revealed software engineering as primary use case, abandoning fine-tuning small models for accelerating large model development. Open-sourced Laneformer 2B demo model achieving 3000 tokens/sec per request. Target: '30x faster LLM inference.' Seed round led by Varsity VC.
🔗 https://techcrunch.com/2026/08/14/kog-is-going-deeper-to-squeeze-more-inference-out-of-gpus/
本期总结：本期AI日报涵盖自动驾驶、AI投资、内容治理和推理优化四大领域。自动驾驶方面，加州开放自动驾驶卡车测试和Uber与Pony.ai的欧洲扩张标志着AI在交通领域的商业化加速，但监管争议（如Teamsters诉讼）也凸显了社会接受度问题。投资领域，Joshua Kushner对AI狂热的警告反映了行业对泡沫风险的反思。谷歌水印政策调整和Kog的推理优化则分别触及AI内容溯源和算力效率两大关键议题。整体来看，AI产业正从技术突破转向规模化落地与治理平衡阶段。