筛选结果
11 篇排序
35:19EN → 简体中文SallyAnn DeLucia:如何编写你的第一个 AI 评估SallyAnn DeLucia · Arize AI从智能体的真实追踪记录中提炼具体的通过与失败标准,再构建一个能测试、能与人工判断对齐,并能在生产环境运行的大模型裁判。
2:10:42EN → 简体中文没人审核的考卷主持 Nathan Labenz前沿实验室从一群小供应商那里购买强化学习(reinforcement learning)环境,却几乎没人核查这些测试是否贴合真实工作。
48:00ENA Two-Second Vote Is Training Models to Be Clickbait主持 Jacob EffronSurge AI CEO Edwin Chen tells Redpoint why LMArena rewards length and emojis, how one lab's models got worse for a year, and why frontier labs are no longer on the same path.
54:53ENA Free Voting Board Burned $2 Million. Then It Raised $100 Million.主持 Jacob EffronDatabricks co-founder Ion Stoica explains why LMArena exists, why China is winning open-source models, and why reliability—not another scaling law—is still the thing holding AI back.
44:21ENMercor's CEO: Knowledge Work's Next Job Is Writing Evals主持 Jacob EffronBrendan Foody tells Redpoint how Mercor turned AI hiring and expert data into a $100 million round — and why he thinks a huge slice of white-collar work will become creating tests for models.
57:11ENA Million Tokens Cannot Write a Good Joke主持 Jacob EffronPrinceton's Arvind Narayanan on why agent evals mislead, inference scaling hits a wall, and AI looks more like the internet than an overnight revolution.
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.
1:23:37ENNine 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.
1:54:17EN → 简体中文Metaculus 在 2020 年 1 月就关注到了新冠,如今它要让 AI 也加入预测主持 Nathan LabenzNathan Labenz 与 Metaculus 首席执行官 Deger Turan 谈认知安全、为期一年的机器人锦标赛,以及为什么一个有用的预测,比一个准确的预测更难做到。
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。
1:07:48EN → 简体中文大多数生产力指标都是谎言:Nicole Forsgren 谈 AI 时代如何衡量开发者体验主持 Lenny RachitskyDORA 与 SPACE 的提出者解释,为何代码行数和旧流水线指标在 AI 时代不够用,以及怎样从倾听、问卷开始改进开发者体验,并把结果接到领导真正关心的事上。









