Tolerable for human readers, these costs become critical when AI agents must understand, reproduce, and extend published work.
大多数人认为人类可读的论文同样适合AI理解,但作者认为传统论文对人类读者是可容忍的,但对AI理解研究过程却造成了'工程税',这反映了当前学术出版系统在AI时代的不适应性。
Tolerable for human readers, these costs become critical when AI agents must understand, reproduce, and extend published work.
大多数人认为人类可读的论文同样适合AI理解,但作者认为传统论文对人类读者是可容忍的,但对AI理解研究过程却造成了'工程税',这反映了当前学术出版系统在AI时代的不适应性。
Reporters may use AI tools vetted and approved for our workflow to assist with research
需要了解哪些AI工具被批准用于研究,以及这些工具如何辅助记者进行研究。
Ask ten different programmers how they use AI, and you can get ten different answers.
文章使用'十个程序员'的例子来说明AI使用方式的多样性,这是一个具体的样本数量。这个数字虽然小,但有效地说明了开发社区对AI工具的态度差异。这种表述方式简洁有力,但缺乏更大规模的调研数据来支持这一观察。
GPT‑5.5 found a proof of a longstanding asymptotic fact about off-diagonal Ramsey numbers, later verified in Lean. The result is a concrete example of GPT‑5.5 contributing not just code or explanation, but a surprising and useful mathematical argument in a core research area.
大多数人认为AI在数学研究领域仅能辅助计算或提供解释,无法独立进行创造性数学推理。但作者展示GPT-5.5能够发现并证明数学定理,这一突破挑战了数学研究作为纯粹人类活动的传统观念,暗示AI可能成为真正的'研究伙伴'而非仅是工具。
GPT‑5.5 found a proof of a longstanding asymptotic fact about off-diagonal Ramsey numbers, later verified in Lean. The result is a concrete example of GPT‑5.5 contributing not just code or explanation, but a surprising and useful mathematical argument in a core research area.
大多数人认为AI在数学研究中的作用主要是辅助计算和验证,但作者认为GPT-5.5能够独立发现数学证明,这在数学研究领域是革命性的。这一观点挑战了人们对AI在创造性思维和抽象推理领域能力的传统认知,暗示AI可能正在从工具转变为研究伙伴。
The application of LLMs in science is already underway... We believe that AI will ultimately bring a fundamental big change to scientific research across disciplines.
大多数人认为AI在科学研究中只是辅助工具,而作者认为AI将从根本上改变科学研究的结构和方式。这一观点与主流认知相悖,因为它暗示AI不仅是提高效率的工具,而是会重塑科学发现、合作和发表的本质。
Huge advances have been made in developing and building more capable models, driven by record investments—forecast by Gartner to grow to around $2.5 trillion in 2026 alone. In contrast, the investment in understanding how the technology works has been minuscule.
这一数据对比揭示了AI领域的一个令人惊讶的不平衡:巨额资金投入到构建更强大的AI系统,而用于理解这些系统如何工作的投资却微不足道。这种不平衡发展可能导致我们拥有强大但不透明的AI武器系统,而对其运作机制知之甚少。
And it’s not just office work. Multi-agent tools like Google DeepMind’s Co-Scientist let researchers use teams of AI agents to coordinate literature searches, generate and test hypotheses, design experiments, and more.
大多数人可能认为人工智能在办公室工作中的应用仅限于数据处理,但作者提出,多智能体工具甚至可以用于研究工作,如文献搜索和实验设计。
These skills act as an orchestration layer that helps scientists work through broad, ambiguous, and multi-step questions more effectively.
将AI描述为'编排层'而非简单工具,体现了AI在科学研究中角色的根本转变。这暗示未来科学家可能更像AI系统的指挥者,而非直接执行者,重塑科研工作流程。
Coding agents working from code alone generate shallow hypotheses. Adding a research phase — arxiv papers, competing forks, other backends — produced 5 kernel fusions that made llama.cpp CPU inference 15% faster.
这一声明揭示了AI代理在代码优化中的关键局限:仅基于代码的优化会产生浅显的假设。通过引入研究阶段,包括阅读学术论文、研究竞争项目和后端实现,代理能够发现更深层次的优化机会,实现了显著的性能提升。这表明AI代理需要更广泛的上下文信息才能做出有意义的创新。
Meta says its rebuilt pretraining stack can reach equivalent capability with >10× less compute than Llama 4 Maverick
令人惊讶的是,Meta声称他们重建的预训练栈只需要Llama 4 Maverick十分之一的计算量就能达到同等能力。这一效率提升是惊人的,表明AI模型训练可能正在经历一个范式转变,从单纯增加计算资源转向优化算法和架构。这可能会对整个AI行业的成本结构和竞争格局产生深远影响。
AMI Labs is not building a product for immediate deployment. This is a fundamental research effort, likely measured in years before commercial applications emerge.
在当今AI创业公司追求快速变现的环境中,作者认为AMI Labs正在进行的是基础研究,而非产品开发。这与大多数AI初创公司的商业模式背道而驰,暗示真正的AI突破需要长期投入而非短期商业考量。
AMI Labs is not building a product for immediate deployment. This is a fundamental research effort, likely measured in years before commercial applications emerge.
在当今追求快速商业化的AI环境中,大多数人认为AI研究应该迅速转化为产品。但作者指出AMI Labs正在进行基础研究,而非直接开发产品,这一观点挑战了科技行业对即时商业化的普遍期待,强调了基础研究的重要性。
Existing research often studies these demands separately: robotics emphasizes control, retrieval systems emphasize memory, and alignment or assurance work emphasizes checking and oversight.
大多数AI研究倾向于将控制、记忆和验证视为独立的问题领域,分别进行研究。但作者认为这种分离研究方法是有缺陷的,因为它们在自然系统中(如松鼠)是紧密耦合的。这一观点挑战了当前AI研究的分割方法,暗示未来的AI系统需要更综合的方法来同时处理这些相互关联的需求。
複雑なリサーチは、単一のクエリに対する回答の集積ではなく、アイデアの生成から、裏付けとなる証拠の探索、矛盾の解消、そして最終的なレポートとしての構造化まで、一連のプロセスを完遂する必要があります。
大多数人认为AI研究助手应该专注于提供快速、直接的答案,但作者强调复杂研究需要完整的'从想法到结构化报告'的完整过程。这与当前AI助手追求即时回答的主流认知相悖,暗示了质量比速度更重要,这是一个非共识的AI应用观点。
推論時により長く、深く思考させることでよりよいアウトプットを引き出せる。これが推論スケーリングの本質です。
大多数人认为AI应该追求更快的响应速度和更高的效率,但作者认为AI应该'长时间深度思考'才能产生更好的输出。这与当前AI行业追求即时响应的主流认知相悖,提出了一个反直觉的观点:计算效率的提升反而应该用于增加思考深度而非速度。
We are especially interested in work that is empirically grounded, technically strong, and relevant to the broader research community.
大多数人认为AI安全研究应该是高度理论化和抽象的,但作者强调需要实证基础和技术强度,这表明OpenAI正在将AI安全研究从纯理论领域转向更注重实际应用和可验证成果的方向,这与传统AI安全研究的精英主义倾向形成对比。
AI is already augmenting important parts of the AI research process itself, and that will only accelerate
for - quote - AI - AI is accelerating AI research itself
at any given time, the CCP may have a better idea of what OpenAI’s frontier advances look like than the U.S. government does.
for - AI - Chinese know more than US government about latest US frontier AI research
Interesting thought. This guy relates the upcome of AI (non-fiction) writing to the lack of willingness people have to find out what is true and what is false.
Similar to Nas & Damian Marley's line in the Patience song -- "The average man can't prove of most of the things that he chooses to speak of. And still won't research and find the root of the truth that you seek of."
If you want to form an opinion about something, do this educated, not based on a single source--fact-check, do thorough research.
Charlie Munger's principle. "I never allow myself to have [express] an opinion about anything that I don't know the opponent side's argument better than they do."
It all boils down to a critical self-thinking society.
26:30 Brings up progress traps of this new technology
26:48
question How do we shift our (human being's) relationship with the rest of nature
27:00
metaphor - interspecies communications - AI can be compared to a new scientific instrument that extends our ability to see - We may discover that humanity is not the center of the universe
32:54
Question - Dr Doolittle question - Will we be able to talk to the animals? - Wittgenstein said no - Human Umwelt is different from others - but it may very well happen
34:54
species have culture - Marine mammals enact behavior similar to humans
36:29
citizen science bioacoustic projects - audio moth - sound invisible to humans - ultrasonic sound - intrasonic sound - example - Amazonian river turtles have been found to have hundreds of unique vocalizations to call their baby turtles to safety out in the ocean
41:56
ocean habitat for whales - they can communicate across the entire ocean of the earth - They tell of a story of a whale in Bermuda can communicate with a whale in Ireland
43:00
progress trap - AI for interspecies communications - examples - examples - poachers or eco tourism can misuse
44:08
progress trap - AI for interspecies communications - policy
45:16
whale protection technology - Kim Davies - University of New Brunswick - aquatic drones - drones triangulate whales - ships must not get near 1,000 km of whales to avoid collision - Canadian government fines are up to 250,000 dollars for violating
50:35
environmental regulation - overhaul for the next century - instead of - treatment, we now have the data tools for - prevention
56:40 - ecological relationship - pollinators and plants have co-evolved
1:00:26
AI for interspecies communication - example - human cultural evolution controlling evolution of life on earth
this is a serious problem because all they need to do is automate AI research 00:41:53 build super intelligence and any lead that the US had would vanish the power dynamics would shift immediately
for - AI - security risk - once automated AI research is known, bad actors can easily build superintelligence
AI - security risk - once automated AI research is known, bad actors can easily build superintelligence - Any lead that the US had would immediately vanish.
nobody's really pricing this in
for - progress trap - debate - nobody is discussing the dangers of such a project!
progress trap - debate - nobody is discussing the dangers of such a project! - Civlization's journey has to create more and more powerful tools for human beings to use - but this tool is different because it can act autonomously - It can solve problems that will dwarf our individual or even group ability to solve - Philosophically, the problem / solution paradigm becomes a central question because, - As presented in Deep Humanity praxis, - humans have never stopped producing progress traps as shadow sides of technology because - the reductionist problem solving approach always reaches conclusions based on finite amount of knowledge of the relationships of any one particular area of focus - in contrast to the infinite, fractal relationships found at every scale of nature - Supercomputing can never bridge the gap between finite and infinite - A superintelligent artifact with that autonomy of pattern recognition may recognize a pattern in which humans are not efficient and in fact, greater efficiency gains can be had by eliminating us
Searching as exploration. White and Roth [71 ,p.38] define exploratory search as a “sense making activity focusedon the gathering and use of information to foster intellectual de-velopment.” Users who conduct exploratory searches are generallyunfamiliar with the domain of their goals, and unsure about howto achieve them [ 71]. Many scholars have investigated the mainfactors relating to this type of dynamic task, such as uncertainty,creativity, innovation, knowledge discovery, serendipity, conver-gence of ideas, learning, and investigation [2, 46, 71].These factors are not always expressed or evident in queriesor questions posed by a searcher to a search system.
Sometimes, search is not rooted in discovery of a correct answer to a question. It's about exploration. Serendipity through search. Think Michael Lewis, Malcolm Gladwell, and Latif Nasser from Radiolab. The randomizer on wikipedia. A risk factor of where things trend with advanced AI in search is an abandonment of meaning making through exploration in favor of a knowledge-level pursuit that lacks comparable depth to more exploratory experiences.
https://www.meetup.com/edtechsocal/events/296723328/
Generative AI: Super Learning Skills with Data Discovery and more!
LLMs are merely engines for generating stylistically plausible output that fits the patterns of their inputs, rather than for producing accurate information. Publishers worry that a rise in their use might lead to greater numbers of poor-quality or error-strewn manuscripts — and possibly a flood of AI-assisted fakes.
for: progress trap, progress trap - AI, progress trap - AI - writing research papers
comment
AI turns semantic relationships into geometric relationships
the shape which is say Spanish can't possibly be the same shape as English right if you talk to anthropologists they would say different cultures different cosmologies 00:14:45 different ways of viewing the world different ways of gendering verbs obviously going to be different shapes but you know the AI researchers were like whatever let's just try and they took the shape which is Spanish 00:14:59 and the shape which is English and they literally rotated them on top of each other and the point which his dog ended up in the same spot in both
ReconfigBehSci on Twitter: ‘Now #scibeh2020: Pat Healey from QMU, Univ. Of London speaking about (online) interaction and miscommunication in our session on “Managing Online Research Discourse” https://t.co/Gsr66BRGcJ’ / Twitter. (n.d.). Retrieved 6 March 2021, from https://twitter.com/SciBeh/status/1326155809437446144
AI is not analogous to the big science projects of the previous century that brought us the atom bomb and the moon landing. AI is a science that can be conducted by many different groups with a variety of different resources, making it closer to computer design than the space race or nuclear competition. It doesn’t take a massive government-funded lab for AI research, nor the secrecy of the Manhattan Project. The research conducted in the open science literature will trump research done in secret because of the benefits of collaboration and the free exchange of ideas.
AI research is not analogous to space research or an arms race.
It can be conducted by different groups with a variety of different resources. Research conducted in the open is likely to do better because of the benefits of collaboration.
Centre for Effective Altruism. (2020, June 13 & 14). EAGxVirtual 2020 Virtual Conference. https://www.youtube.com/playlist?list=PLwp9xeoX5p8NfF4UmWcwV0fQlSU_zpHqc
Young, V. A. (2020, May 20). Nearly Half Of The Twitter Accounts Discussing ‘Reopening America’ May Be Bots. Carnegie Mellon School of Computer Science. https://www.scs.cmu.edu/news/nearly-half-twitter-accounts-discussing-%E2%80%98reopening-america%E2%80%99-may-be-bots
Yang Yang: The Replicability of Scientific Findings Using Human and Machine Intelligence (Video). Metascience 2019 Symposium. https://www.metascience2019.org/presentations/yang-yang/
Hope, T., Borchardt, J., Portenoy, J., Vasan, K., & West, J. (2020, May 6). Exploring the COVID-19 network of scientific research with SciSight. Medium. https://medium.com/ai2-blog/exploring-the-covid-19-network-of-scientific-research-with-scisight-f75373320a8c