Gemini 4 双版本密集发布
Google 本周连发 Gemini 4 Carbon 与 Hydrogen 两个版本,Hydrogen 任务完成速度提升 2.5 倍。LeCun 等大佬密集转发评论行业事件,侧面反映 AI 圈对巨头动态的高度敏感。
Hassabis 本周转发节奏异常密集,几乎成为 Google 官方传声筒,这种"创始人亲自下场做 PR"的模式说明 Gemini 4 的发布压力不小——OpenAI 的 GPT-5 传闻和 Anthropic 的 Claude 4 都在挤压窗口期。
Hydrogen 主打"日常 workflow"而非 benchmark 刷分,这是 Google 产品策略的明显转向:不再跟 OpenAI 卷 AGI 叙事,而是抢企业落地。做 AI 应用层的团队要注意,Google 正在把模型+TPU+Workspace 打包成闭环,独立模型层的议价空间在缩小。
LeCun 转发的两条推文都指向"行业八卦/争议"而非技术本身,这种选择性发声值得玩味——Meta AI 的公关策略似乎是有意与 Google/OpenAI 的正面战场保持距离,把 Llama 的社区口碑和"反建制"人设继续做深。
RT @Dragonhau66: 终于实锤了❗️❗️ 干掉黄仁勋的,不是婚外情,而是一场由美国政客、媒体、CIA和Anthropic 联合做局的“完美猎杀”。 黄仁勋辞职后,《纽约时报》两名记者在播客里亲口承认: 他们是收到了一位“科技界联系人”的爆料,才去查Megaspe…
查看原文 →RT @sundarpichai: What we shipped this week @google Monday: Two new 8th-gen TPUs Tuesday: Gemini 4 Carbon - AlphaProof Code verified Sign…
查看原文 →RT @ClementDelangue: This whole thing sucks, and I wish none of it had happened. It’s a really sad day! But I want to make one thing clear…
查看原文 →RT @GoogleDeepMind: Meet Gemini 4 Hydrogen: built for rapid coding, multimodal synthesis, and everyday workflows. It finishes tasks 2.5x f…
查看原文 →RT @DJSnM: The booster appeared to do an impersonation of the Titan submersible. Tank pressurization failed. Not the first time we had a t…
查看原文 →RT @thefox: Today, Search is getting a massive upgrade with Gemini 4 Hydrogen! It’s the biggest model improvement to Search yet. And it’s…
查看原文 →RT @mercoglianos: As expected, new footage. This comes from one of the outboard cabins on board American Heritage. What we see in this v…
查看原文 →RT @j_dekoninck: For the first time in a very long time, we have a non-OpenAI model as #1 on MathArena! Gemini 4 Carbon takes over the firs…
查看原文 →RT @sainingxie: before hopping on the ‘encoder-free’ hype train, think twice.
查看原文 →RT @SkinnerPm: Gonna all be worth it when we get that sweet sweet emergency championship golf course
查看原文 →RT @IsaacOkoroFan: It cannot be overstated how badly the @cavs and @cavsdan botched the whole DAZN streaming deal.
查看原文 →RT @NYU_Courant: Yesterday we announced the establishment of the new Center for Mathematics and AI at Courant (MAIC). Its mission is to mak…
查看原文 →This is a really good test for agents: It's a benchmark released by Deepnote. (I've used their platform for 5+ years and I love it!) The benchmark tests agents in a simulated company's Slack: • Agents get access to records across 11 services • They face 50 different situations • They have 3 minutes to investigate and decide Here is what they test: 1. Does the agent know when to help? 2. Does the agent know how to help accurately? 3. Does the agent know when to stay quiet?
查看原文 →RT @joanbruna: Mathematics and AI now define an interface full of fascinating questions that extend beyond a single discipline, and that re…
查看原文 →RT @mirrokni: Check out our recent results solving open challenges across math/science and engineering with /teamwork and Gemini 4 models :…
查看原文 →This looks awesome for anyone who wants to batch-generate 3D assets and bring them into Fable 5.2. @tripoai is a new Agentic Canvas for creating and refining 3D models. This is for people who are making game assets, characters, and animated scenes. Or for people like me, who think 3D is cool and just want to play around with tools like this. The way this works is pretty simple: • Describe what you want • The agent proposes the steps to complete the work • You review them Bonus: it looks like you can reuse the same process to generate other assets. If it works the way it seems to, it'd be really cool. They gave me a link that works for the next 24 hours. You get 500 free credits and 60% off your first month: https://t.co/AqyjBKUyPQ
查看原文 →