888 X 動態摘要|07/26
生成時間: 2026-07-26 05:03:06
總結
AI產業巨頭積極倡議開源模型以推進創新與安全,同時相關模型及開發工具在市場上展現顯著的應用普及與效率提升。
今日重點
1. 開放模型戰略
- 人物: Demis Hassabis (Google DeepMind CEO)
- 時間: Jul 24
- 熱度: 👀 53,231,232
- 觀察: Google DeepMind CEO強調開放模型對AI產業安全、網路安全、創新、普及及主權的重要性,並主張開放與閉源模型應並存。
- 意義: 指明未來AI生態系統將是開放與閉源模型並存的戰略方向,影響產業競爭格局、監管走向及技術投資分配。
2. Gemma模型普及
- 人物: Demis Hassabis (Google DeepMind CEO)
- 時間: 4h
- 熱度: 👀 133,501
- 觀察: Google的Gemma模型家族下載量已突破9億次,達成重要的市場普及里程碑。
- 意義: 反映開源模型在開發者社群中取得巨大成功,預示其在AI應用開發中的主導地位和潛在的商業化生態價值。
3. Meta力挺開源模型
- 人物: Yann LeCun (Meta AI掌門)
- 時間: 12h
- 熱度: 👀 30,109
- 觀察: Meta AI掌門Yann LeCun支持開放模型願景,並稱公司正持續投入並推動Gemma(原文如此)以利開發者,看好其生態系統。
- 意義: 產業巨頭持續表態支持開源模型,強化了開源AI在全球技術發展中的戰略地位與投資前景。
4. 模型訓練效率
- 人物: Thomas Wolf (Hugging Face 聯創)
- 時間: 22h
- 熱度: 👀 5,470
- 觀察: Hugging Face聯創觀察到Opus 5模型在FrontierCode基準測試上呈現有趣的非單調成功-努力曲線。
- 意義: 揭示了AI模型訓練與優化可能存在複雜性,對未來高效能模型開發策略及算力投入方式具參考價值。
5. Codex提升開發效率
- 人物: Peter Steinberger (OpenClaw→OpenAI)
- 時間: 22h
- 熱度: 👀 561,562
- 觀察: OpenAI前員工展示Codex模型如何以極簡方式,自動化處理繪圖並生成代碼的開發工作流程。
- 意義: 突顯AI工具在簡化複雜開發流程上的巨大潛力,預示開發效率的大幅提升和自動化趨勢,對軟體產業具顛覆性意義。
原始動態
重點 | Demis Hassabis
- 時間: Jul 24
- 熱度: 👀 53,231,232
- 原文: For my first post, I’m sharing a letter signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models.
其他 | Elon Musk
- 時間: 6h
- 熱度: 👀 720,386
- 原文: I mean c’mon. This is a real photo 😭
其他 | Elon Musk
- 時間: 6m
- 熱度: 👀 390,588
- 原文: The most powerful moving object ever created by humans
重點 | Demis Hassabis
- 時間: 4h
- 熱度: 👀 133,501
- 原文: Gemma model family crossed the 900M downloads. What a milestone!
重點 | Yann LeCun
- 時間: 12h
- 熱度: 👀 30,109
- 原文: Very much agree with this letter. 3 years ago we started Gemma aligned with this vision, that building a thriving open weight model ecosystem was critical and likely a huge component of our future industry. Today we continue to invest and push Gemma to be maximally useful for developers, and I’m personally very bullish on the open ecosystem it enables. Thanks and also great to see you here!
其他 | Ray Dalio
- 時間: 2h
- 熱度: 👀 35,366
- 原文: All these principles, rules, and checks and balances won't be worth much if you don't have capable people in positions of power who instinctually want to operate for the good of the community based on the agreed-upon principles. A company's leaders must have wisdom, competence, and the ability to have close, cooperative, and effective working relationships characterized by both thoughtful disagreement and commitment to following through with whatever the ideameritocratic process decides.
重點 | Thomas Wolf
- 時間: 22h
- 熱度: 👀 5,470
- 原文: Interesting non-monotonic success-effort curve for Opus 5 on FrontierCode
其他 | Yann LeCun
- 時間: Jul 24
- 熱度: 👀 53,271,191
- 原文: For my first post, I’m sharing a letter signed on why open models matter. AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models.
重點 | Peter Steinberger
- 時間: 22h
- 熱度: 👀 561,562
- 原文: "How to graph-max with Codex and 5.6 Sol: 1. Draw a graph (literally in any tool, even on paper) 2. Send it to Codex and say \"write a code mode script that implements this workflow, run it with\" There's no step 3, it just works. Seriously it's that simple."
其他 | Greg Brockman
- 時間: 17h
- 熱度: 👀 289,447
- 原文: stories like this should be told more
其他 | Huanusa
- 時間: 16h
- 熱度: 👀 192,394
- 原文: 轉:西安交大王树国校长发表所谓以下“不当言论”后, 中央组织部副部长彭金辉同志到会宣布中央决定: 王树国同志不再担任西安交通大学校长、党组副书....
其他 | Greg Brockman
- 時間: 12h
- 熱度: 👀 177,268
- 原文: "ChatGPT Work for using websites which require login:"
其他 | Ray Dalio
- 時間: Jul 24
- 熱度: 👀 642,515
- 原文: "It All Comes Down to the Strait of Hormuz: the Final Battle. When the war started in March, I shared this article explaining what was clear to me and what would likely happen. Now it is clear that it is happening and we will soon have the final battle and its result."
其他 | Demis Hassabis
- 時間: Jul 23
- 熱度: 👀 201,745
- 原文: Congratulations on your new role as UK's AI Minister in the Cabinet, great news for the UK AI ecosystem!
其他 | Elon Musk
- 時間: 1h
- 熱度: 👀 868,021
- 原文: beautiful engineering
其他 | Nassim Taleb
- 時間: 6h
- 熱度: 👀 93,362
- 原文: "This retard is claiming that once one changes religion, past history vanishes. So pagan Greece/Rome does not belong to Christian Greeks/Italians, etc. Note 1: Arabs did not populate the Levant, rather the opposite. Note 2: Palestinians speak Levantine, a dialect of Aramaic w/Canaanite & Arabic influence (see my article)."
其他 | Nassim Taleb
- 時間: 6h
- 熱度: 👀 70,624
- 原文: The left has been turning against Zionists because of a sense of justice & empathy with the Palestinians. The right has been turning against Zionists because of their outrage noticing the insidious control of politics, education & mainstream media.
其他 | Huanusa
- 時間: 19h
- 熱度: 👀 44,084
- 原文: 还有非常多的智者,都在不同环境中发挥着自己的力量 这是民族最后的希望👍
其他 | Huanusa
- 時間: 21h
- 熱度: 👀 35,069
- 原文: 巴菲特最经典的一次演讲 据说段永平自称当年看了不下十次
其他 | Ray Dalio
- 時間: Jul 23
- 熱度: 👀 77,320
- 原文: In order to assess well, one has to gain a threshold level of understanding and that takes time. Some people have the ability and the courage to hold people accountable, while most don't; having such ability and courage is essential. And the person doing the assessing must not have conflicts of interest--such as being in a subordinate position to the person they are intended to check on--that stand in the way of holding them accountable, including recommending that they be fired.
其他 | Huanusa
- 時間: 6h
- 熱度: 👀 1,048
- 原文: 网友的提醒,希望可以让更多人有所警觉 保护好自己的孩子
其他 | Peter Steinberger
- 時間: 14h
- 熱度: 👀 78,063
- 原文: Hit a new record with our autoreview skill. 66 rounds on a gnarly refactor. github.com
其他 | Balaji Srinivasan
- 時間: Jul 21
- 熱度: 👀 72,876
- 原文: Here's the official announcement from the Republic of Kazakhstan.
其他 | Greg Brockman
- 時間: 16h
- 熱度: 👀 60,398
- 原文: Pets make building fun
其他 | Huanusa
- 時間: 6h
- 熱度: 👀 2,275
- 原文: Article
其他 | Sean Kelly
- 時間: Jul 24
- 熱度: 👀 20,250
- 原文: on the bright side, if you have regrets about an email that you sent, you can now claim your AI agent is to blame
其他 | Peter Steinberger
- 時間: Jul 24
- 熱度: 👀 91,383
- 原文: I'm happy to see the unclonable encryption problem solved; Prabhanjan Ananth and Amit Sahai report their findings, with the help of the UCLA harness (link below). This was a challenge for the quantum crypto community for the last 6 years.
其他 | Sean Kelly
- 時間: 3h
- 熱度: 👀 5,074
- 原文: "Paper here:" "More here:"
其他 | Sean Kelly
- 時間: Jul 23
- 熱度: 👀 1,129,109
- 原文: "I am so tired of seeing these posts. I am so tired of seeing my beloved field of study be the testing lab for a technology that is disrupting everyday life at such a high pace. Mathematics were always about the beauty of seeing the ideas of a person unfolded in an extremely precise language so we can all look into a mathematician's mind and see their purest form of ideas, their understanding of a problem, their style, their way of communicating to other fellow researchers through mathematical results... this often transcends the common language humans use to communicate with each other and it is incredibly exciting to bump into shared ideas another person had thousands of kilometers away several years ago without even knowing them. I am extremely unexcited about AI-breakthroughs in Mathematics in the same way I am extremely unexcited about a picture created by AI or an AI-generated video, not to speak about AI-generated text. Mathematics requires soul and intention. Mathematics need not be useful by definition, Mathematics are pure art that need hard work and effort to understand. The high you get when a year of work gets condensed in a one-page result is incomparable to anything else and, certainly, a machine is not apt to relate to humans in such a humane way. I am not interested in the truth of the results. I am interested in how was the mind behind a result able to connect those dots. I am interested in the diversity human thoughts bring to Mathematics. I am not interested in a bunch of servers running probabilities to return a Theorem, even if true: that lacks the objective of Mathematics as I see them.
其他 | Sean Kelly
- 時間: Jul 20
- 熱度: 👀 17,076
- 原文: "Can humans be biased against AI? Does an LLM like Claude or a robot \"clanker\" matter less just because it is artificial: made of chips and wires instead of flesh and blood? With 5 preregistered studies, our new paper rigorously validates a psychometric scale for \"substratism\" 📝"