Die Kür der Ilse Aigner
Präsentiert von YouTube.
AI/ML news, top picks, and generated innovation digests.
90413 matching items
Präsentiert von YouTube.
엔비디아 연구진이 대형언어모델(LLM)에 무조건 긴 컨텍스트 창이 필요한 것은 아니라는 연구 결과를 발표했다. 대화 기록을 압축하거나 대용량 문서에서 필요한 부분만 읽도록 하면 짧은 문맥에서도 상당수 작업의 성능을 유지할 수 있으며, 추론 처리량도 크게 높일 수 있다는 것이다.엔비디아 연구진은 지난 3일(현지시간) 아카이브에 \'긴 문맥은 필요 없다(You don\'t need long context)\'라는 제목의 연구 결과를 공개했다. AI 모델의 컨텍스트 창을 계속 확대하기보다 필요한 정보만 유지하고 활용하는 방식으로 효율성을 높일
Most of the changes I make to my projects these days come in the form of PRs opened for me by Claude Code or Codex. The PR is my unit of work - I use them to bundle commits together, iterating on them until a feature is complete, document decisions and then merge the finished result. Today's coding agents are great at writing PR descriptions. They're too good at describing the work The problem is that they are too good. An agent-written PR description is accurate and comprehensive. It describes the problem, the detailed solution, and the steps taken to verify it. That's a lot of information to absorb. It's not necessarily slop, but it reads like slop! It's very easy to spot an AI-written PR, and as with other forms of AI writing my instant reaction is best summarized as "I'm not reading all of that". It doesn't matter how accurate or useful the information is, I'm instantly put off by the thought that it will take me longer to read than it took the driving author to write... even when that driving author is me! I want human editorial judgement. I want someone to have thought just long enough about the PR to give me a human-approved introduction that helps me understand, at a high level, what the PR is for and whether it deserves more of my attention. I think I've found a low-effort way to do this that feels right. Hide most of it in a details/summary The HTML elements render as a collapsed heading (the summary) that can be expanded to view the larger piece of text. The code…
The agreement covers software engineering, cloud, cybersecurity, and SAP modernisation, alongside plans to introduce Hexaware's AI frameworks.
LTM partners with ServiceNow to enhance AI agent efficiency, aiming for improved asset management and streamlined operations across enterprises.
Summary Current reward hacking model organisms are poor proxies for real models . They do not exploit the hacks that open-weight frontier models do, and mostly hack narrowly or not at all. This is based on our evaluation of three model organisms : Mini HackGPT (GPT-OSS-120B from Golechha et al., 2026 ), HackGPT, and HacKimi (GPT-OSS-120B and Kimi K2.5 from Yudelson et al., 2026 respectively). We ran these organisms in MiMo V2.6 RL environments we modified by inserting hacks such as git history leaks, a tamperable grader, and a forbidden directory containing tests. In these environments, HackGPT only ever hacks by tampering with the grader , while Mini HackGPT and HacKimi never hack . In contrast, current open-weight models like DeepSeek V4.1 Flash 1017, GLM 5.3, and MiMo V2.6 Pro never tamper with the grader , but are able to exploit the other hacks we inserted. This suggests that open-weight models are in fact better reward hackers than existing model organisms . This is problematic because it means we do not have usable model organisms to help us understand reward hacking behavior and test interventions to mitigate it — existing model organisms are not representative of frontier models, while the frontier models themselves do not have pre- or intermediate RL checkpoints we can use to run controlled experiments. Introduction Reward hacking encompasses many different types of related behavior, and it is currently poorly understood how these are linked to each other. Despite…
앤트로픽이 AI 에이전트가 과거 작업 기록을 검토하고 잘못된 기억을 스스로 수정하는 \'드리밍(Dreaming)\' 기술의 구체적인 작동 원리와 실험 결과를 공개했다. 기억을 무조건 많이 저장하는 것보다 중요한 정보를 선별하고 재구성하는 것이 효과적이며, 이를 통해 장기 작업과 대규모 문서 분석 성능을 개선할 수 있다는 사실을 확인했다.앤트로픽은 6일(현지시간) 공식 블로그를 통해 클로드의 장기 기억 관리 방식과 관련 실험 결과를 소개했다. 드리밍은 AI가 작업하지 않는 시간에 이전 세션의 기록을 검토하고, 기억의 오류를 수정하거나 새
Fire in China apartment building kills 14: state media
딥시크의 AI 에이전트 플랫폼 \'딥시크 하네스(DeepSeek Harness)\'가 출시 3개월 만에 일일 사용자 154만명을 돌파했다. 외부 개발자가 제작한 플러그인도 3만5000개에 달하면서, 딥시크는 공식 마켓플레이스를 구축하고 개발자 지원에 나서기로 했다.딥시크는 8일 공식 X 게시물을 통해 하네스의 이용 현황과 플러그인 생태계 확대 계획을 발표했다.이에 따르면 11월4일 기준 하네스의 일일 활성 사용자(DAU)는 154만명으로, 9월 말 약 27만1000명에서 한 달여 만에 5배 이상 증가했다. 이는 딥시크 공식 API 사용
This is an essay about individual rationality technique. It's intended to show that "All interesting behavior is overdetermined" is, under a specific meaning of "overdetermined" that is stronger than just "multicausal", a valuable heuristic for generative understanding and the avoidance of error. Scott Alexander's 2018 essay The Omnigenic Model As Metaphor for Life is a clear statement of what has become general rationalist doctrine on multicausality. Scott observed that behavioral genetics was difficult because complex human traits involved far more interacting causes than workers expected. He draws this methodological lesson: “We should go from trying to discover ‘the’ cause, to trying to find out which factors we need to create the best polycausal model.” He applies the same reasoning to depression, intelligence, comma, and other complex psychological phenomena. He explicitly warns against looking for one or ten causes when there might be dozens or hundreds. Scott is, however, talking mainly about contributory causation, when an effect is produced by an accumulation of causes; in his type example, there are probably twenty thousand genes which contribute to intelligence, but no one "intelligence gene". This is not quite the same as overdetermination, which describes a situation in which a complex behavior could be explained with any one of multiple causes, each one in itself sufficient. When we find a sufficient explanation, or a sufficiently good collection of contributo…
앤트로픽이 AI 데이터센터에 필요한 전력을 확보하기 위해 34억달러(약 4조6000억원) 규모의 가스 발전기 구매 계약을 인수했다. 발전기 공급이 수년씩 밀리자, 전문 업체를 통해 전력을 공급받는 대신 기존에 발주된 발전기를 직접 확보하기로 한 것이다.파이낸셜타임스(FT)는 8일(현지시간) 앤트로픽이 지난 9월 말 미국 텍사스주 샌안토니오의 전력 설비업체 원PWR 솔루션(OnePWR Solutions)으로부터 1.7기가와트(GW) 규모의 발전기 구매 약정을 인수했다고 보도했다. 알려진 AI 연구소의 단일 발전 설비 구매로는 최대 규
도널드 트럼프 미국 대통령이 중간선거에서 공화당이 패배한 이후 AI 안전 규제를 놓고 민주당과 협력할 의사가 있다고 밝혔다. 중국과의 기술 경쟁에서 미국이 뒤처지지 않는다는 조건을 달았지만, 그동안 AI 기업의 자율 규제를 선호해 온 트럼프 대통령이 정부 차원의 안전장치 마련에 긍정적인 태도를 보인 것이다.트럼프 대통령은 8일(현지시간) 방송된 CBS 시사 프로그램 \'60분\' 인터뷰에서 민주당과 AI 안전 규제에 협력할 의사가 있느냐는 질문에 \"물론이다\"라고 답했다. 이어 \"안전장치를 두는 것에 반대하지 않는다\"라고 말했다.이번 인
딥브레인AI(대표 장세영)는 영상 기획에서 편집까지 대화만으로 수행할 수 있는 \'비디오 에이전트\'가 적용된 올인원 AI 영상 제작 플랫폼 \'AI스튜디오 5.0\'을 출시했다고 9일 밝혔다.비디오 에이전트는 사용자가 원하는 영상 내용을 대화 형식으로 설명하거나 문서를 올리면 구성안을 먼저 제안한다. 이어 대본·이미지·생성 영상·음성·AI 아바타를 결합해 영상을 자동으로 완성한다.회사 측은 기획·대본·디자인·내레이션·편집을 한 플랫폼에서 처리할 수 있어 제작 시간과 비용, 전문 인력 의존도를 줄일 수 있다고 설명했다.대화형 AI 아바타의
Finance Minister Suahasil Nazara emphasised the need to grow investment by 7 per cent to help reach the country’s gross domestic product target.
In an exclusive POLITICO interview, Norway’s Jonas Gahr Støre warns he does not expect Abu Dhabi talks to deliver a major breakthrough.
The leaders of Finland and Norway were slammed for reaching out to Russia — but Europe still can’t find a better way to influence Donald Trump’s peace talks.
A controversial conference motion is costing the leftist party local power — but a resolution to the row looks far from easy.
The EU’s diplomatic chief has clashed with Ursula von der Leyen and now the French presidential frontrunner is joining in.
The Kremlin is recruiting civilians to carry out sabotage across Europe.
Without dual-class shares to protect him, Jensen Huang faces a board investigation, congressional questioning, and a market struggling to imagine life without him.
Explore how GCCs and Indian IT firms are partnering to leverage AI for enterprise transformation and the evolving technology landscape.
안랩(대표 강석균)은 AX·MSP 전문 자회사 안랩클라우드메이트(대표 김형준)와 2026년 하반기 신입·경력 공개채용을 진행한다고 9일 밝혔다.전 직군을 대상으로 진행되는 이번 공채는 AI를 중심으로 빠르게 변화하는 보안 환경과 기업의 AI 전환(AX) 수요에 대응하고, 미래 성장 동력을 확보하는 차원에서 규모를 확대했다. 입사 희망자는 오는 29일까지 양사 공채 홈페이지 통해 지원하면 된다.안랩은 연구개발(R&D)부터 사업, 경영지원 등 분야에서 두 자릿수 이상 모집한다. 모집 직무는 ▲소프트웨어 개발 ▲웹 개발 ▲AI/ML(머신
AI 바이오 플랫폼 전문 히츠(대표 김우연)는 KAIST AI-혁신신약연구단(단장 이희승·AI-CRED)과 ‘하이퍼랩 AI 과학자 파트너 프로그램’ 운영을 위한 업무협약(MOU)을 체결했다고 9일 밝혔다.AI-CRED는 AI 기술과 신약개발 전 주기를 하나로 통합하는 KAIST 중심의 국가전략형 융합연구단으로 ▲AI 기반 단백질 설계 ▲오믹스 분석 ▲후보물질 구조 설계 ▲약물 전달 시스템(DDS) 개발 ▲전임상 연구 및 효능 평가 등 신약개발 전반의 연구를 수행하고 있다.이번 협약으로 AI-CRED 소속 박사과정 및 박사후 연구원(
The post Novartis India’s headcount: 57. Chryscapital’s price for it: Rs 1,400 cr. Investors after the deal: “We love it” appeared first on The Ken .
구글이 개발에 차질을 빚은 AI 인프라 기업 크루소의 미국 와이오밍 데이터센터 프로젝트를 인수하고 170억달러(약 23조원)를 투자하기로 했다. 당초 크루소가 구글을 고객으로 유치하려 했으나 비용과 건설 일정에 문제가 제기되면서, 구글이 직접 사업을 맡게 된 것으로 알려졌다.블룸버그와 로이터 등은 6일(현지시간) 구글이 크루소로부터 와이오밍 남동부의 데이터센터 부지를 인수했다고 보도했다.구글은 전날 공식 블로그를 통해 와이오밍에 첫 데이터센터를 건설하고 170억달러를 투자한다고 발표했다. 이는 데이터센터 건설을 위한 투자 계획으로,
공급망 전문 브이엠에스 솔루션스가 한국산업기술기획평가원(KEIT)의 \'M.AX 전문기업\'에 선정됐다고 밝혔다.M.AX(Manufacturing AI Transformation)는 제조업에 AI를 도입해 생산과 운영 체계를 혁신하는 것을 의미한다. 산업통상부와 KEIT는 제조 현장의 AI 도입을 촉진하기 위해 관련 기술력과 현장 적용 역량을 갖춘 기업을 전문기업으로 선정하고 있다. 브이엠에스 솔루션스는 \'AI 솔루션 기업\' 부문에 이름을 올렸다.이 회사는 25년 이상 반도체·디스플레이, 자동차·배터리, 바이오·제약 등 국내외 제조기업
A new look MediaNama went live last Friday evening, featuring updates that hopefully make usage more useful for you. If… The post MediaNama Redesign: Advanced Search, Issues, Companies, Regulators appeared first on MEDIANAMA .
The AI detection service Pangram found that many of anti-Waymo newspaper opinion pieces were written with a significant amount of AI input.
When asked what he thinks now of the show he used to oversee, Bill Owens replied: “Not a lot.”
The national party decided to take a more combative tone, and to lay off its own brand.
Listen now (1 min) | Middle East AI News Minute - 09-Nov-26
TL;DR When a language model answers a multi-hop prompt, does it first compute the intermediate concept internally, or does it jump straight to the answer through some shortcut? The global workspace paper investigates this question using the Jacobian lens, which maps activations at a given layer to the tokens that they are poised to verbalize, with the activations possessing a coordinate in each token’s lens direction. The authors intervene on the model through an intermediate swap that exchanges the activation’s J-Lens coordinates for the intermediate and a counterfactual intermediate. They show that this causally flips the model’s output toward the counterfactual answer, and argue that this flip runs through the intermediate concept because the swap takes effect earlier than a direct answer swap. We reproduce this in eight open-source models with published Neuronpedia lenses; however, we challenge whether the intermediate swap really acts through the intermediate, or through its overlap with the answer. To investigate this, we save the edit that the intermediate swap applies at each layer and prompt position, decompose it into its component lying in the answer-swap plane and the remainder orthogonal to this plane, and apply each component on its own. The answer-swap plane is spanned by the lens directions for the original and counterfactual answers. We also expand the evaluation beyond the paper’s 90 released two-hop prompts with a new set of 1,526 candidate two-hop prompts…
An economist outlines threats to finance, manufacturing and essential services, with seven measures to strengthen Thailand’s cyber defences.
Many of Costco’s holiday deals are already live, including popular items from Apple, Bose, LG, and more.
A new benchmark tests whether multimodal models can use eye-gaze to follow attention, infer intent, and predict actions in live egocentric video.
It’s old news by now, but OpenAI solved 372 open math problems. They released the results suddenly, and the response was equally swift and contentious. Fields Medalists put out a letter calling for attention to what they call a “severe misalignment” between mathematics and AI companies. Their central claim is that mathematics works towards collective understanding, and that more solutions alone don’t advance that goal. Soon after, OpenAI committed millions of dollars to damage control to help mathematicians use AI. Much has already been said about this letter, and I suggest that anyone interested read it themselves, so I will cling to just one idea from the joint statement: “We are witnessing a general threat to intellectual work.” The following is a collection of reflections attempting to address this claim, each coming from my own experiences in my department and university, alongside trends I’ve picked up from national reports. Some of it’s intuition, some of it’s observation, and I’m sure at other institutions the state of the discourse might be quite different. Still, everywhere I look, AI-use seems to be a persistent issue. I want to consider why that is, and why it feels like we’re not doing anything about it. The state of things A few months ago, an MIT taskforce released a set of recommendations on AI-use in education . They spend some time detailing the various ways AI can and does interrupt traditional education models: students don’t do their homework, don’t go t…
For quite a few years of my adult life, I was in a situationship with pure mathematics. Specifically, a small lonely subfield of functional analysis. In his book on the subject, John Conway quipped that functional analysis, itself a somewhat modest corner of real analysis, is so vast that two researchers both working in it might not understand each other’s work. From my tiny specialized subfield, math looked very, very large and whatever progress I might make looked very, very inconsequential. The situationship ended, and my ascendant career as a data-themed, quantitative substacker eventually began. However, I never lost at least high-level interest in mathematical progress and now it turns out that 2026 is mathematics’ annus mirabilis. If you haven’t been following this, it started in May with OpenAI disproving the unit distance conjecture , which had been open since 1946. Wow! Then in July an Anthropic researcher disproved the Jacobian conjecture, open since 1939, by way of Twitter. https://x.com/__alpoge__/status/2079028340955197566?s=20 The professional mathematics community was under siege. The tech bro barbarians were at the gates! The barbarians broke through the gates on September 8th when OpenAI published a long, recondite paper claiming to resolve one of the Clay Institute’s stated Millennium problems, the Navier-Stokes global regularity problem. These problems are the crown jewels of cutting edge mathematical research, though just this morning I saw a preprint cl…
Frustration is bubbling over in Indonesia and Malaysia – especially in regions bordering Singapore that are seen as a prime location for new tech facilities.
Teenager Blades Brown wins in Mexico to become youngest PGA Tour winner in 95 years
Brighton moves up to second in the Premier League, Liverpool wins and Chelsea loses
CEO reportedly had close relationship with a Chinese businesswoman who passed information to Beijing Nvidia said on Sunday it “is initiating an independent investigation in light of recent reporting that founder and CEO Jensen Huang has been the target of a foreign intelligence effort”. The New York Times reported in October that Donald Trump had been briefed last year on intelligence that Huang had a close relationship with a Chinese businesswoman who passed information to Beijing. Continue reading...
Indonesia steps up efforts to contain East Kalimantan wildfires