【方法】这种设计展示了数据处理的谨慎方法，通过分离素材生成和历史更新，防止了数据不一致问题。这种分阶段处理方法值得其他类似工具借鉴，特别是在处理增量数据时。

【局限】这一关于数据处理的声明提到了零数据保留和隐私安全处理的测试，但未详细说明这些措施的具体实施方式和限制。需要了解这些技术如何平衡安全监控与隐私保护，以及它们在多大程度上能够有效防止数据泄露和滥用。

这个数据点展示了税务处理的规模：数万份报税表和数百万份文件。这解释了为什么自动化如此重要—人工处理如此大规模的数据不仅耗时而且容易出错。'tens of thousands'和'millions'之间的比例关系也显示了每份报税表通常涉及数十份支持文档的复杂性。

OpenRouter声称拥有800万全球用户，每月处理100万亿个token（约每周25万亿）。这是一个相当大的用户规模和处理量，但需要验证这些数据的计算方式和来源。在AI基础设施领域，这类用户指标是评估平台价值的重要指标。

在法律文档处理中达到90.9%的准确率，特别是在处理模糊文档编辑任务时的智能提升，展示了AI在专业领域的深度应用能力，这种进步将极大扩展AI在法律和合规领域的应用价值。

tokenizer的更新虽然增加了token使用量，但提高了文本处理效率，这反映了AI模型在基础架构上的持续优化，这种优化虽然带来短期成本增加，但长期将提升AI的处理能力和准确性。

这些惊人的使用指标展示了AI技术在实际应用中的大规模采用。特别是每分钟处理150亿个token的能力，反映了企业对AI处理能力的巨大需求，以及AI已经从实验阶段进入实际工作流程的临界点。

引入基于截图的LLM评估器和并行化功能是一个令人惊讶的创新。通过截图评估AI模型的性能，可以更直观地理解自动化过程中的视觉理解能力，而并行化功能则大大提高了基准测试的效率，这代表了AI系统评估方法的重要进步。

令人惊讶的是：Claude code的并发处理能力如此强大，能够同时运行12个子代理，但同时也暴露了其API使用限制的脆弱性，几分钟内就达到20倍的限额，这表明即使是高级AI模型也存在明显的使用边界，可能影响大规模应用场景。

令人惊讶的是：Artemis安全公司在短短几个月内就处理了每小时超过10亿个安全事件，这种数据处理规模反映了现代企业面临的网络安全威胁的惊人频率和复杂性。

SIAC 滤波器的「超收敛」性质令人印象深刻：对多项式次数为 q 的 DG 解进行 SIAC 后处理后，收敛阶从 q+1 跃升至 2q+1——精度几乎翻倍，却几乎不增加计算代价。这是数值分析中罕见的「免费午餐」：滤波本身是线性操作，计算量微乎其微，却能将误差的收敛速率提升一个整量级。

「在特定 token 位置追踪当前生效的情绪概念」——这句话揭示了一个深刻洞见：情绪不是持续状态，而是逐词涌现的动态标注。这与人类神经科学中「情绪是对当前感知的实时评估」高度吻合，暗示 LLM 在没有神经元的情况下，重演了大脑皮层处理情绪的某种计算逻辑。

大多数人认为实时视频处理需要极高的计算资源和帧率才能有效，但作者仅用两块80G加速器就实现了2 FPS的实时系统，并达到了最先进的性能。这一结果挑战了高性能视频处理需要大量计算资源的共识，暗示通过优化算法和架构可以显著降低实时视频处理的计算门槛。

大多数人认为现有的视频大模型可以通过简单的触发-响应管道或描述式叙述来处理实时视频流，但作者认为这种方法对于开放式问答和长时程交互效果有限。这是一个反直觉的观点，因为它挑战了当前视频处理领域的常规做法，暗示需要更集成的端到端方法来真正实现实时视频理解。

大多数人可能认为处理复杂文档的最佳方式是将其作为一个整体来处理，保持上下文完整性。但作者提出将复杂文档分解为多个子任务并由子代理分别处理的方法更有效，这一方法挑战了文档处理中'整体优于部分'的传统认知，暗示分解策略可能更适合处理超长文档。

虽然并行处理在计算领域常见，但将其应用于LLM代理系统中的信息搜索任务可能出人意料，因为大多数LLM应用仍采用顺序处理模式，作者的观点挑战了这一现状。

大多数人认为计算资源应该用于当前任务，但作者提出利用未充分利用的计算资源并行预测额外令牌的创新方法，这挑战了传统计算资源分配的常识，暗示了AI计算效率的全新可能性。

https://en.wikipedia.org/wiki/Emily_M._Bender

for - key insight - paraphrase - adjacency - memory consolidation - sleep - massive unconscious parallel processing - From a memory perspective, - sleep can be understood as critically important for normal memory function, - given the lasting ramifications of consolidation. - Consolidation is the establishment of new connections - anchoring recent memories within relevant knowledge networks - While consolidation happens, some conscious experience (the dream) may be synthesized as the memory processing unfolds - Dreams reflect a storyline generated to make sense of a subset of activated memory fragments. - Consolidation that wires new connections happens across the entire cerebral context, without the constraints that come with conscious experience. - Unconscious processing during sleep takes advantage of massive parallel processing to connect all these thoughts together. - Dreams reflect a small portion of overnight memory consolidation work.

for - adjacency - brain - identity - immune system - immune system involved with identity at a microscopic level - which molecule is part of "you"? - which moleculev is NOT part of you - immune system preceded neural system

? - maybe there is similarity between neural processing and immune system response?

for - adjacency - Indyweb dev - natural language - timebinding - parallel vs serial processing - comparison - spoken vs written language - what's also interesting is that spoken language is timebinding, sequential and our written language descended from that, - in spite of written language existing in 2D and 3D space, it inherited sequential flow, even though it does not have to - In this sense, legacy spoken language system constrains written language to be - serial - sequential and - timebound instead of - parallel - Read any written text and you will observe that the pattern is sequential - We constrain our syntax it to "flow" sequentially in 2D space, even though there is absolutely no other reason to constrain it to do so - This also reveals another implicit rule about language, that it assumes we can only focus our attention on one aspect of reality at a time

for - stats - internet - average connection speed - hasn't increased as much as storage and processing power

The infamous Apple typewriter memo is 40 years old ... by [[Richard Polt]]

Chapeau! Een Belgische website kaart dit aan in de context gezond leven.

How to get round the idea of processing which is a stumbling block for fun?

processed notes are akin to the amount of nutritive value in processed food...

I often find that people who discuss "workflows" and the idea of "processing" their notes are the ones who are falling trap to the anxiety-producing side of the work.

BD should have found more exciting words for "processing" which he uses two more times in the next paragraph.

This relates to Luhmann's quote about only doing what is easy/fun/flow:<br /> - https://hypothes.is/a/TQyC1q1HEe2J9fOtlKPXmA<br /> - https://hypothes.is/a/EyKrfK1WEe2RpEuwUuFA7A

Compare: - being trapped in the box: https://hypothes.is/a/AY7ABO0qEeympasqOZHoMQ - idea of drudgery in the phrase "word processing"

Typewriter Video Series - Episode 147: Font Sizes and the Writing Process by [[Joe Van Cleave]]

typewriters for note making

double or 1 1/2 spacing with smaller typefaces may be more efficient for drafting documents, especially first drafts

editing on actual paper can be more useful for some

Drafting on a full sheet folded in half provides a book-like reading experience for reading/editing and provides an automatic backing sheet

typewritten (or printed) sheets may be easier to see and revise than digital formats which may hide text the way ancient scrolls did for those who read them.

Jack Kerouac used rolls of paper to provide continuous writing experience. Doesn't waste the margins of paper at the top/bottom. This may be very useful for first drafts.

JVC likes to thread rolls of paper into typewriters opposite to the original curl so as to flatten the paper out in the end.

For true deep processing and learning, intellectualism, one must think beyond the single source they are consuming and think about everything they know. Although keep in mind selective attention for true learning and thinking.

This process is habitualized by means of Zettelkasten and further aided in tool like hypothes.is

Very interesting method, and way of looking at the signal. "What makes a document exceptional because something is common within itself and uncommon without".

Indeed. But testing in general, as it is done currently, in modern formal education, discourages deep learning as opposed to shallow learning.

Why? Because tests with marks implore students to start learning at max 3 days before the test, thus getting knowledge into short-term memory and not long term memory. Rendering the process of learning virtually useless even though they "pass" the curriculum.

I know this because I was such a student, and saw it all around me with virtually every other student I met, and I was in HAVO, a level not considered "low".

It does not help that teachers, or the system, expect students to know how to learn (efficiently) without it ever being taught to them.

My message to the system: start teaching students how to learn the moment they enter high school

Reply to Nidhi Sachdeva: Nidhi Sachdeva, PhD Just went through the micro-lesson itself. In the context of teachers using to generate instruction examples, I do not argue against that. The teacher does not have to learn the content, or so I hope.

However, I would argue that the learners themselves should try to come up with examples or analogies, etc. But this depends on the learner's learning skills, which should be taught in schools in the first place.

Reply to Nidhi Sachdeva: I agree mostly, but I would advice against using AI for this. If your brain is not doing the work (the AI is coming up with the story/analogy) it is much less effective. Dr. Sönke Ahrens already said: "He who does the effort, does the learning."

I would bet that Cognitive Load Theory also would show that there is much less optimized intrinsic cognitive load (load stemming from the building or automation of cognitive schemas) when another person, or the AI, is thinking of the analogies.

https://www.linkedin.com/feed/update/urn:li:activity:7199396764536221698/

Are Travel Loyalty Programs at Risk? by [[Elaine Glusac]] 2024-03-28

Visa and Mastercard Agree to Cap Swipe Fees in Settlement by [[Tara Siegel Bernard]]

First, ask yourself if the item is actionable. Then, series of stuff you might do: throw away, reference, someday/maybe, incubate (calendar/tickler)

Hartlib Papers 30/4/47A, Ephemerides 1640, Part 2.

In extolling the Ark of Studies created by Thomas Harrison, Samuel Hartlib indicated that the combinations of information and the potential links between them created the "argumentative part" of the system. In some sense this seems to be analogous to the the processing power of an information system if not specifically creating its consciousness.

1:35:00 The gap effect or spacing effect as time interleaved wherein information is processed. Embracing boredom and taking none stimulative breaks aids in this.

Reliance on ETL seen as risky

I’m not convinced about this. Real-time API connectivity between systems is a great goal…very responsive to changes filtering through disparate systems. But a lot of “modern” processing is still done by ETL batches (sometimes daily, sometimes hourly, sometimes every minute).

for - semantic folding - semantic fingerprint - natural language processing - NLP - cortical.io - Numenta

Event logs allow for gradual processing of events.

(1:20.00-1:40.00) What he describes is the following: Most of his notes originate from the digital using hypothes.is, where he reads material online and can annotate, highlight, and tag to help future him find the material by tag or bulk digital search. He calls his hypothes.is a commonplace book that is somewhat pre-organized.

Aldrich continues by explaining that in his commonplace hypothes.is his notes are not interlinked in a Luhmannian Zettelkasten sense, but he "sucks the data" right into Obsidian where he plays around with the content and does some of that interlinking and massage it.

Then, the best of the best material, or that which he is most interested in working with, writing about, etc., converted into a more Luhmannesque type Zettelkasten where it is much more densely interlinked. He emphasizes that his Luhmann zettelkasten is mostly consisting of his own thoughts and is very well-developed, to the point where he can "take a string of 20 cards and ostensibly it's its own essay and then publish it as a blog post or article."

Pages is so much more impressive than you'd expect in so many ways, but damn...

The way styles are handled still perplexes the shit out of me... even after consuming this document.

The essence for this video is correct; active learning, progressive summarization, deep processing, relational analytical thinking, even evaluative.

Yet, the implementation is severely lacking; marginalia, text writing, etc.

Better would be the use of mindmaps or GRINDEmaps. I personally would combine it with the Antinet of course.

I do like this guy's teaching style though 😂

Code for data processing and model training should be separated as different modules.

(13:36) Perhaps this is also why visual information is so useful for learning and cognition (see GRINDE)... Maybe the visual medium should be used more in instruction instead of primarily auditory lectures (do take into account redundancy and other medium effects from CLT though)

When it comes to thinking, the Zettelkasten solves an important issue which is the problem of scope, which is impossible at the current moment in mindmapping software such as Concepts.

Mainly, Zettelkasten allows you gain a birds-eye holistic view of a topic, branch, or line of thought, while allowing you to at the same time also gain a microscopic view of an "atomic" idea within that thought-stream, therefore creating virtually infinite zoom-in and zoom-out capability. This is very, very, beneficial to the process of deep thinking and intellectual work.

CIFAR-10 performance results are overestimates since some of the training data is essentially in the test set.

During the coronavirus pandemic, many vendors facing inflation began to pass along the 3% (or more) credit card processing fees to their customers. Previously many credit card companies would penalize vendors for doing this (and possibly cut them off). This fee was considered "the cost of doing business".

Some vendors prior to the pandemic would provide cash discounts on large orders because they could circumvent these fees.

Does this affect (harm) inflation? Is it a form of Gresham's law at play here? What effect does this have on credit card companies? Are they so integral to the system that it doesn't affect them, but instead the customers using their legal tender?

I've always disliked the idea of "work" involved in "making" notes and "processing" them. Framing zettelkasten and knowledge creation in terms of capitalism is a painful mistake.

the quote is from https://blay.se/2015-06-21-living-with-a-zettelkasten.html

Not sure that I agree with the framing of "algorithmic processing" here as it was done manually by a person pulling out facts. But it does bring out the idea of where collecting ends and synthesis of a broader thesis out of one's collection begins. Where does historical method end? What was the purpose of the collection? Teaching, writing, learning, all, none?

This is a key difference in knowledge mastery...

Rank (social or otherwise) can be a signal for predictability from the perspective of consensus views for how to behave towards others with respect to the abilities or values being measured.

Ranking people for some sort of technical ability may be a better objective measure rather than ranking people on social status which is far less objective from a humanist perspective. In employment situations, individuals are more likely to rely on social and cultural biases and racist tendencies rather than on objective measures with respect to the job at hand. How can we better objectify the actual underlying values over and above the more subjective ones.

rank as signal processing

Can I use some of these sorts of methods with respect to corpus linguistics over time to better identified calcified words or archaic phrases that stick with the language, but are heavily limited to narrower(ing) contexts?

https://genizalab.princeton.edu/events/2022/princeton-machine-learning-and-future-philology-symposium

Was this recorded?

The Friedberg Judeo-Arabic Project contains a large corpus of Judeo-Arabic text which can be manually searched to help improve translations of texts, but it might also be profitably mined using information theoretic and corpus linguistic methods to provide larger group textual translations and suggestions at a grander scale.

Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜. In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT '21). Association for Computing Machinery, New York, NY, USA, 610–623. https://doi.org/10.1145/3442188.3445922

https://www.nlpdemystified.org/course

I have thought long about the same issue and beyond. The triple (wiki, Hypothesis, donations) could be a working way to search for OER, form a social group processing them, and optionally support the creators.

I imagine that as follows: a person wants to learn about X. They can head to the wiki site about X and look into its Hypothesis annotations, where relevant OER with their preferred donation method can be linked. Also, study groups interested in the respective resource or topic can list virtual or live meetups there. The date of the meetups could be listed in a format that Hypothesis could search and display on a calendar.

Wiki is integral as it categorizes knowledge, is comprehensive, and strives to address biases. Hypothesis stitches websites together for the benefit of the site owners and the collective wisdom that emerges from the discussions. Donations support the creators so they can dedicate their time to creating high-quality resources.

Main inspirations:

Deschooling Society - Learning Webs

Building the Global Knowledge Graph

Schoolhouse calendar

https://www.researchgate.net/profile/Robert-Amsler

Apparently follow my blog. :)

Makes me wonder how we might better process and semantically parse peoples' personal notes, particularly when they're atomic and cross-linked?

By waiting some amount of time (days/weeks/a few months) between originally reading something and processing one's notes on it allows them to slowly distill into one's consciousness. It also allows one to operate on their diffuse thinking which may also help to link ideas to others in their memory.

https://dictionary.apa.org/elaboration

https://www.explainpaper.com/

Another in a growing line of research tools for processing and making sense of research literature including Research Rabbit, Connected Papers, Semantic Scholar, etc.

Functionality includes the ability to highlight sections of research papers with natural language processing to explain what those sections mean. There's also a "chat" that allows you to ask questions about the paper which will attempt to return reasonable answers, which is an artificial intelligence sort of means of having an artificial "conversation with the text".

cc: @dwhly @remikalir @jeremydean

The practice of repeating items as a means of attempting to place them into short-term memory is called maintenance rehearsal. Examples of this practice include repeating a new acquaintance's name or perhaps their phone number multiple times as a means of helping to remember it either for the short term or potentially the long term.

Research on the levels-of processing model of memory indicates that maintenance rehearsal is not as effective at promoting long term memory as methods like elaborative rehearsal.

Seaman, K. L., Christensen, A. P., Senn, K., Cooper, J., & Cassidy, B. S. (2022). Age Differences in the Social Associative Learning of Trust Information. PsyArXiv. https://doi.org/10.31234/osf.io/b38rd

https://maggieappleton.com/joining-ought

!- question : processing of differences * what does this mean?

https://www.openbookpublishers.com/

Considering how I've been publicly speaking and behaving (melodramatically, that is) - as someone who has returned to using my iPhone as my primary working device - this sort of oversight is precisely what I expected, actually, What I did not expect of Apple was to respond as early as the next numeric release to this omission.

Running this very first build of iOS 16, I can indeed that Apple has thought of at least one original context for Quick Note creation, but obviously, it's quite hard to say at this point.

Anywho/how, here's what it looks like at the moment.

Immediate knowledge capture by highlighting, annotating, or other means when taking notes can help to decrease cognitive load. This is similar to other productivity methods like quick logging within a bullet journal system, writing morning pages, or Getting Things Done (GTD). By putting everything down in one place, you can free your mind of the constant need to remember dozens of things. This frees up your working memory to decrease stress as you know you've captured the basic idea for future filtering, sorting, and work at a later date.

Lehnen, N., Glasauer, S., Schröder, L., Regnath, F., Biersack, K., Bergh, O. V. den, & Werder, D. von. (2022). Post-COVID symptoms in the absence of organic deficit—Lessons from diseases we know. PsyArXiv. https://doi.org/10.31234/osf.io/yqar2

Fernandez-Castaneda, A., Lu, P., Geraghty, A. C., Song, E., Lee, M.-H., Wood, J., Yalcin, B., Taylor, K. R., Dutton, S., Acosta-Alvarez, L., Ni, L., Contreras-Esquivel, D., Gehlhausen, J. R., Klein, J., Lucas, C., Mao, T., Silva, J., Pena-Hernandez, M., Tabachnikova, A., … Monje, M. (2022). Mild respiratory SARS-CoV-2 infection can cause multi-lineage cellular dysregulation and myelin loss in the brain (p. 2022.01.07.475453). https://doi.org/10.1101/2022.01.07.475453

How to use Relational Databases to process logs

https://legacy.imagemagick.org/discourse-server/viewtopic.php?t=28963

A domain-specific language for encoding legal interpretations.

https://www.kapwing.com/settings/account

Set up using my gmail account.

The mechanical work of moving an index card from one position to another (and potentially changing or modifying links to it in the process) allows us to call card catalogues paper machines. This property is not shared by information stored in codices or scrolls and thus we do not call books paper machines.

Sloman, S. A. (2021). How Do We Believe? Topics in Cognitive Science, 0(2021), 1–14. https://doi.org/10.1111/tops.12580

Did I ever really find this man intelligent??? Things sincerely do make a lot more sense now. Such a specific lack of aspiration.

Marco, M. A., Meg. (n.d.). How Your Brain Tricks You Into Taking Risks During the Pandemic. ProPublica. Retrieved September 2, 2021, from https://www.propublica.org/article/how-your-brain-tricks-you-into-taking-risks-during-the-pandemic?token=l24Nh-wDyBgy53bhcy5jGvQh1IDRcxzE

Baek, E., Hyon, R., López, K., Porter, M. A., & Parkinson, C. (2021). Lonely individuals process the world in idiosyncratic ways. PsyArXiv. https://doi.org/10.31234/osf.io/yt872

Facet.ai, an AI-based image editing software

Jung, Y., Lee, Y. K., & Hahn, S. (2021). Web-scraping the Expression of Loneliness during COVID-19. PsyArXiv. https://doi.org/10.31234/osf.io/59gwk

Antico, L., & Corradi-Dell’Acqua, C. (2021). Far from the eyes, far from the heart. COVID-19 confinement dampened sensitivity to painful facial features. [Preprint]. PsyArXiv. https://doi.org/10.31234/osf.io/ewvp7

Possibly useful summaries of a collection of papers and resources, including some by Andy Clark.

A brief overview of predictive processing.