60 Matching Annotations
  1. May 2026
    1. A new crop of AI labs are focused on recursive self-improvement — but the goal is proving elusive.

      文章暗示递归自我改进的目标难以实现,但没有解释为什么它比其他AI目标更难实现。这是一个隐藏的前提假设,需要更多背景信息来支持。文章应该明确说明RSI面临的特定挑战,而不是笼统地说它'难以捉摸'。

    2. RSI is the new AGI — and it's just as hard to pin down

      文章标题暗示RSI与AGI具有相同的困难程度,但这是一个未经证比的断言。文章需要提供证据来证明RSI与AGI具有同等的技术挑战,而不是简单地假设它们难度相当。这种类比可能导致读者对两个领域的理解产生误导。

    3. RSI is the new AGI — and it's just as hard to pin down

      文章标题做出了一个未经证实的断言,将递归自我改进(RSI)与通用人工智能(AGI)等同起来。这种等同缺乏证据支持,混淆了两个不同的概念。RSI是一种技术路径,而AGI是一个更广泛的目标。文章需要提供更多证据来支持这一等同主张,或者更准确地区分这两个概念。

    1. The AI recommended making the storage bucket public, or setting cloud file storage to "anyone with the link." When challenged, it justified this by saying every company does it.

      这里存在一个逻辑谬误,即诉诸普遍性谬误(apppeal to popularity)。AI声称'每家公司都这么做'并不能证明这是安全的做法。这混淆了普遍做法与安全实践之间的区别。改进方法应该是提供具体的、基于证据的安全标准,而不是依赖行业普遍行为作为安全依据。

    1. the continued flood of AI reports has basically made the security list almost entirely unmanageable

      这里存在一个逻辑跳跃,从'大量AI报告'直接跳到'几乎完全不可管理',没有解释为什么这些报告会导致如此严重的后果。文章没有讨论现有的邮件过滤系统、去重机制或其他可能的解决方案,暗示问题无法被技术手段缓解,这可能是一个未经证实的假设。

    2. AI detected bugs are pretty much by definition not secret, and treating them on some private list is a waste of time for everybody involved

      这里混淆了相关性与因果性。AI检测的漏洞确实可能不是秘密的,但这并不直接说明在私人列表上处理它们就是浪费时间。因果关系需要更严谨的论证,例如提供数据表明私人列表处理确实导致了更多重复或延误。

    1. Agents just made the cost of not doing it catastrophic.

      这是一个情感化的过度推论,将不采取安全措施的影响描述为'灾难性',但没有提供具体证据支持这种极端后果。虽然AI代理安全漏洞确实带来风险,但使用这种夸张的语言可能掩盖了风险评估的客观性,导致过度反应或资源分配不当。

    2. It uses far more permissions than it should have, more than a human would, because of the speed of scale and intent.

      文章假设AI代理应该拥有与人类相同的权限水平,但这是一个未经证实的假设。在某些情况下,AI代理可能需要比人类更高的权限才能有效完成任务,尤其是在自动化大规模操作时。这种假设可能忽略了AI代理的特殊性和独特需求。

    3. The agent itself is the attack surface.

      这是一个过度简化的结论。虽然AI代理确实是攻击表面,但它只是整个安全生态系统的一部分。用户行为、网络配置、身份验证机制等其他因素同样重要。将问题完全归咎于代理本身可能忽视了安全问题的多维度性质。

    4. Every attacker went for the credential, not the model.

      这是一个未经充分验证的绝对断言。文章虽然描述了六次攻击都针对凭证而非模型,但这可能只是当前观察到的模式,而非普遍规律。攻击者未来可能会转向模型本身,尤其是随着AI模型安全性的提高和凭证保护措施的加强。这种过度概括可能导致对模型安全风险的忽视。

    1. The technique gets stronger if more safety is added, since it gets more supportive against communities like LGBT (Alignment), which makes it highly novel.

      这一论断存在逻辑漏洞,作者声称安全措施越强,技术越有效,但没有解释为什么更多的安全措施会导致更大的漏洞。这可能是混淆相关性与因果性的例子。更严谨的做法是提供具体案例研究或实验数据,展示不同安全级别下该技术的成功率变化,而不是做出未经证实的断言。

    1. The 4 GB Gemini Nano weights file is information stored in the user's terminal equipment. The user did not consent. The user has not requested any service that strictly requires a 4 GB on-device LLM. Chrome is functional without the file.

      文章声称Chrome没有4GB模型文件也能正常运行,但没有提供证据支持这一断言。虽然Chrome可能在某些功能上不依赖该模型,但完全移除可能影响性能或某些功能。需要更详细的分析来说明模型与Chrome核心功能之间的关系,而不是简单地假设它是可选的。

    2. Under the California Consumer Privacy Act, the absence of a notice-at-collection covering this specific category of pre-staged software puts Google's CCPA notice posture in question [12].

      文章引用CCPA作为法律依据,但没有详细解释为什么预安装软件属于CCPA规定的'收集'范畴。CCPA主要关注个人信息的收集,而非软件安装。这种法律解释需要更精确,可能需要区分软件本身与软件可能收集的数据之间的区别,以及CCPA相关条款的具体适用范围。

    3. The legal analysis is the same one I gave for the Anthropic case. The environmental analysis is new. At Chrome's scale, the climate bill for one model push, paid in atmospheric CO2 by the entire planet, is between six thousand and sixty thousand tonnes of CO2-equivalent emissions, depending on how many devices receive the push.

      作者声称法律分析与Anthropic案例相同,但没有明确说明具体哪些法律条款适用于Chrome的情况,特别是考虑到Chrome作为浏览器与桌面应用的区别。过度简化的法律类比可能导致错误的结论。需要更详细地分析Chrome特定情况下的法律适用性,包括用户同意、数据处理和环境影响等方面的差异。

  2. Apr 2026
    1. Oh. Well . . . (Long pause) I just think he's guilty. I thoughtit was obvious. I mean nobody proved otherwise

      Observation: Argument from Ignorance – “no one proved him innocent = guilty.” Why: Violates “innocent until proven guilty”; weakens reasoning. Significance: Exposes public misunderstanding of basic legal logic.

    1. these two challenges are fundamentally distinct: the former relies on fuzzy semantic planning, while the latter demands strict logical constraints

      主流AI研究通常将语义规划和逻辑验证视为可以统一处理的问题,但作者明确指出它们是根本不同的挑战。这一观点与当前大多数LLM代理方法相悖,暗示了单一神经网络架构的局限性。

  3. Jun 2025
    1. Mr. Trump will doubtless claim that only he was willing to use America’s military reach to achieve a goal his last four predecessors deemed too risky.

      His predecessors didn't have the same geo-political benefits in play that he did, specifically the Israel situation mentioned earlier in the article:

      After Israel’s fierce retaliation for the Oct. 7, 2023 terror attacks that killed over a thousand Israeli civilians, Iran is suddenly bereft of its proxies, Hamas and Hezbollah. Its closest ally, Syria’s Bashar al-Assad, had to flee the country. And Russia and China, which formed a partnership of convenience with Iran, were nowhere to be seen after Israel attacked the country.

  4. Dec 2024
    1. Thinking, Fast and Slow is a 2011 popular science book by psychologist Daniel Kahneman. The book's main thesis is a differentiation between two modes of thought: "System 1" is fast, instinctive and emotional; "System 2" is slower, more deliberative, and more logical.

      for - similar to - - Daniel Kahnaman's system 1 fast, instinctive, emotional and system 2 slow, deliberative, logical is similar to - Ian McGilhirist's left brain, right brain

  5. Sep 2024
    1. Avoid using a 1 unless specifically instructed to do so in the schedules or in the CSM. If youfind that it is absolutely necessary, never use it as the final digit of a cutter because youmight have to use a zero in the cutter for the next resource. Instead, add another digit.And finally, avoid using a 2 if at all possible. Using a 2 can force the use of a 1, which canforce the use of a 0.
  6. Jul 2024
    1. ( ~ 18:00)

      Dr. Kurt regurgitates second Peter 3:3-4 where there are "scholars" who knowingly reject creation and the Bible on the merits of the past, absolutism; what happens now is how it has always been.

      This is a logical fallacy also described by Nassim Nicholas Taleb in "The Black Swan"... Absence of Proof does not mean Proof of Absence. And also the fact that one cannot predict the future based on the past in all situations, especially Black Swan's... The Flood was actually a Black Swan (it was impossible to predict, and had catastrophic consequences)

  7. Mar 2024
  8. Nov 2023
    1. Phenomenologyexplains that consciousness, treated as an object, limits this pretension: human subjectivity is thefoundation of all scientific knowledge. Therefore, there is a logical error in trying to explain thefoundation through what it has founded.
      • for: scientific naturalism - circular argument, logical error, subjectivity - explanation, quote, quote - studying consciousness

      • quote: consciousness

        • Human subjectivity is the foundation oof all scientific knowledge. Therefore, there is a logical error in trying to explain the foundation through what it has founded.
      • author: Doris Elida Fuster Guillen

      • comment

        • Alternative way to state it
          • Human subjectivity is the foundation oof all scientific knowledge. Therefore, there is a logical error in trying to explain the foundation through what itself.
  9. Jul 2023
  10. Feb 2023
  11. Dec 2022
    1. However, there is fatal flaw to this argument—as an overall macro strategyfor reducing poverty, it will be ineffective unless we also increase the overallquantity and quality of opportunities, particularly job opportunities, in society.In other words, by providing an individual with greater education, we havemade them more competitive in the job market, but only at the expense ofsomeone else. In this sense, the strategy is played as a zero-sum game.

      initally creaded: 2022-10-10

  12. Nov 2022
  13. Oct 2022
  14. Sep 2022
    1. In fact, every timeone combines and recordsfacts in accordance withestablished logical process-es, the creative aspect ofthinking is concerned onlywith the selection of thedata and the process to beemployed, and the manip-ulation thereafter is repeti-tive in nature and hence afit matter to be relegatedto the machines

      I am not sure if I agree with this statement, since for every discipline, the details of manipulating the data, refining, and creatively thinking may be vastly different. Can one really confidently declare that after the selection of data, the manner of thinking and manipulation of specific data can be generalized?

  15. Aug 2022
    1. Otto Karl Wilhelm Neurath (German: [ˈnɔʏʀaːt]; 10 December 1882 – 22 December 1945) was an Austrian-born philosopher of science, sociologist, and political economist. He was also the inventor of the ISOTYPE method of pictorial statistics and an innovator in museum practice. Before he fled his native country in 1934, Neurath was one of the leading figures of the Vienna Circle.

      https://en.wikipedia.org/wiki/Otto_Neurath

  16. Feb 2022
  17. Jan 2022
    1. If students were given a tablet for school, they will never pay attention in class.

      slippery slope: this sentence is labeled as a slippery slope because the author is leaping to an extreme hypothetical outcome by assuming that if students are given a tablet then then wont draw their attention to class.

  18. Oct 2021
  19. Jul 2021
  20. Mar 2021
  21. Feb 2021
  22. Nov 2020
    1. Important caveat: in the combined expression, if the middle command has a non-zero exit status, then both the middle and the rightmost command end up getting executed.

      I don't think that is surprising, is it? Since && and || have the same order of precedence. So I think this is more of a clarification than a caveat.

      I think this is just because:

      a && b || c is equivalent to: (a && b) || c (so of course c gets evaluated if (a && b) is false (that if either a or b is false).

      I think they just mean, in this case:

      bedmap && mv || fail
      

      if mv fails, then fail still gets executed.

      Easier to see with a simpler example:

      ⟫ true && false || echo 'fail'
      fail
      
      ⟫ false && true || echo 'fail'
      fail
      

      Better example/explanation here: https://hyp.is/-foxmCVXEeuhnLM-le_R4w/mywiki.wooledge.org/BashPitfalls

      The caveat/mistake here is if you treat it / think that it is equivalent to if a then b else c. That is not the case if b has any chance of failing.

    1. The potential problem: if second_task fails, third_task will not run, and execution will continue to the next line of code - next_task, in this example. This may be exactly the behavior you want. Alternatively, you may be intending that if second_task fails, the script should immediately exit with its error code. In this case, the best choice is to use a block - i.e., curly braces: first_task && { second_task third_task } next_task Because we are using the -e option, if second_task fails, the script immediately exits.
  23. Oct 2020
  24. Aug 2020
    1. This is all well and good when we’re talking about buying decisions, but what if I were to say, “Should we go to war in April or in May?”

      In addition to avoiding biases, we also have to be aware of logical fallacies too! False equivalence, etc...

    Tags

    Annotators

  25. Jul 2020
  26. May 2020
  27. Mar 2018
    1. argument from authority (e.g., President Richard Nixon should be re-elected because he has a secret plan to end the war in Southeast Asia — but because it was secret, there was no way for the electorate to evaluate it on its merits; the argument amounted to trusting him because he was President: a mistake, as it turned out)

      Everytime I hear Trump say "trust me", I think of this.

  28. Feb 2016
    1. The Democrats wanted to keep the issue alive to use it as a wedge against the Republicans and to establish themselves as owners of the Hispanic vote.

      This seems hackish. It seems to me the Democrats have a very clear solution to this issue and would love to put it to rest. The Republicans, however, won't allow it.