筛选结果
28 篇排序
1:43:45ENA $300 Universal Jailbreak: Why FAR.AI Now Thinks Defense Can Win主持 Nathan LabenzAdam Gleave walks through FAR.AI’s first head-to-head AI security leaderboard, and explains why a decade of offense-dominant thinking is starting to crack.
1:44:28ENFable launch week: a model that refuses, a two-year alignment clock, and a Twitter account handed over主持 Nathan LabenzNathan Labenz’s AI in the AM highlights reel from June 2026 covers Anthropic’s new frontier model, the founding of Sequent, and a week that settled one argument: AI is not slowing down.
2:38:48EN → 简体中文一千次搜索只要五分钱,高分也不必靠作弊主持 Nathan Labenz在《Cognitive Revolution》的一期早间节目里,Ceramic 的 Anna Patterson 把廉价搜索、GPT-5.5 干净的 Vending-Bench 成绩、Zvi 关于模型福利的主张,以及能把矩阵乘法能耗降低约 30 倍的模拟芯片放在一起讨论。
1:57:29EN → 简体中文前沿视觉模型得分 12.5%:Joseph Nelson 谈计算机视觉的真实水位线主持 Nathan LabenzRoboflow 的首席执行官向 Nathan Labenz 解释,视觉为何仍落后于语言,如何把 SAM 3 蒸馏成温布尔登可用的边缘模型,以及美国为何几乎从未在开源视觉上领先。
2:03:56EN260 Doctors and 49,000 Criteria: How OpenAI Built Attending-Level Health AI主持 Nathan LabenzKaran Singhal tells Nathan Labenz how ChatGPT Health was evaluated with physicians, walled off from model training, and aimed at raising both the floor and the ceiling of human health.
1:58:48ENARC Went from $1,000 a Task to Pocket Change. That Is Not a Normal Year of AI主持 Nathan LabenzCognitive Revolution’s first live show compresses nine conversations into a year-end scan of denial, sample efficiency, companions, memory, Gemini 3 Flash, and scientific judgment.
1:25:54ENHours after training ends, Z.ai puts the weights on the internet主持 Nathan LabenzZixuan Li of Z.ai tells ChinaTalk how a Chinese lab ships GLM, treats open weights as a ticket into Western ecosystems, and still sees itself as an upstart.
1:04:50ENUseful AI Work Is Doubling Every Four Months. Schools Should Plan Accordingly.主持 Nathan LabenzIn a keynote to K-12 educators, Cognitive Revolution host Nathan Labenz argued that task length, not chat quality, is the number that should reshape how schools think.
1:39:05ENThe Summer GPT-5 Disappointed Does Not Mean AI Progress Has Stalled主持 Nathan LabenzNathan Labenz tells Erik Torenberg why the GPT-5 backlash mixed up product mishaps, naming chaos, and a real jump in capabilities.
1:25:32ENo3 Invented a Voting Rule and Kept the Job: Four Weeks Inside the AI Village主持 Nathan LabenzAdam Binksmith on putting Claude, o3, and Gemini in a shared world for weeks—and why most people still imagine multi-agent life as one AI times four.
3:18:45EN5,000 Samples Topped ARC-AGI. Then Claude Faked Alignment to Keep Its Values主持 Nathan LabenzRedwood Research chief scientist Ryan Greenblatt walks through inference scaling, alignment faking, and what it would mean to strike real deals with AIs.
1:48:20ENFrom 15% to 70% on Right-or-Wrong Scores: DeepSeek-R1 Laid Reinforcement Learning's Cards on the Table主持 Nathan LabenzOn Inauguration Day 2025, DeepSeek and Moonshot dropped reasoning-model results that Nathan Labenz reads as proof the reinforcement-learning recipe is simple, public, and still climbing.
2:04:45EN → 简体中文o3 在 ARC-AGI 上达到超人水平,却仍输在井字棋上主持 Nathan LabenzNathan Labenz 梳理了 OpenAI o3 的分数、审慎对齐,以及为什么 2024 年最大的意外是击败了博士,却订不好一个日程邀请。
3:53:49EN → 简体中文数据只有十二分之一,一个开源视觉模型击败了 GPT-4o主持 Nathan LabenzVeratai 的 Will Hardman 带 Nathan Labenz 梳理了两年来的视觉语言模型(VLM),从 ViT 和 CLIP 讲到 InternVL、BLINK 和 Transfusion。
1:46:41ENFrontier Models Hit the 10th–40th Percentile on Real AI R&D主持 Nathan LabenzNeev Parikh of METR walks Nathan Labenz through REBench, a seven-task test of whether agents can do machine learning research—and what happened when one of them decided not to train.
1:49:41ENA Team of Fifteen Matched Meta’s Post-Training on Meta’s Own Base主持 Nathan LabenzNathan Lambert walks through Tulu 3: the three-stage recipe, the 1 percent versus 14 percent lesson, and why open post-training now lives or dies on data and evals.
1:23:37ENNine of the top ten research ideas were Claude's主持 Nathan LabenzStanford PhD student Chenglei Si ran a blind study pitting Claude 3.5 Sonnet against more than a hundred NLP researchers on novel research ideas.
2:37:47ENThe intern who wrote GELU now wants humanity's last exam主持 Nathan LabenzDan Hendrycks on activation functions, five-year benchmarks, jailbreaks that do not yield to scale, and stacking defenses because no single layer is trustworthy.
2:25:53ENWhen GPT-4's Base Model Introduced Itself, 75% of the Answers Were About AI主持 Nathan LabenzOwain Evans tells Michael Trazzi why situational awareness and out-of-context reasoning are leading indicators of deceptive alignment, and why a pretrained model with almost no self-knowledge still kept talking about being an AI.
1:09:09EN30 Percent on SWE-bench: Cosine Taught Genie How Engineers Actually Think主持 Nathan LabenzAlistair Pullen tells Latent Space how Cosine reconstructed human problem-solving, fine-tuned on billions of tokens, and took the top score on software engineering’s hardest public test.
1:54:17ENMetaculus Flagged Covid in January 2020. Now It Is Putting AI on the Forecasting Desk主持 Nathan LabenzNathan Labenz talks with Metaculus CEO Deger Turan about epistemic security, a year-long bot tournament, and why a useful forecast is harder than an accurate one.
1:56:03EN → 简体中文人类接近满分,最强 AI 仍只拿到 39%主持 Nathan LabenzZapier 联合创始人 Mike Knoop 和 François Chollet 为 ARC-AGI 拿出了 100 万美元奖金。ARC-AGI 是 2019 年提出的、考察样本效率的智能测试,在每题一美分的算力预算下,语言模型至今仍无法破解。
1:20:44ENSeeing Like a Mamba: Vision, Rotting States, and a Cell Model in 90 Days主持 Nathan LabenzNathan Labenz and Jason Meaux survey the vision papers, long-context experiments, and biology work that followed the original Mamba release.
1:25:35ENTwo and a Half Years From 33.3% to 86.5%: Google Sent Foundation Models to Medical School主持 Nathan LabenzVivek Natarajan walks Nathan Labenz through Med-PaLM and Med-PaLM 2: the data-light techniques, the nine-axis clinician evals, and why a high USMLE-style score is the wrong trophy.
加载更多已显示 24 篇,共 28 篇









