JEV CEO:RLHF是走弯路 | RLHF | OpenAI | 大语言模型 | 人工智能 | Diogo Almeida | JEV | 机器学习 | AI泡沫 | AI经济革命
前OpenAI核心成员、GPT-4与ChatGPT共同作者Diogo Almeida在两场技术分享中深入剖析了当前AI领域的核心矛盾:模型在辅助任务上表现惊艳,在自动化任务上却几乎不可用。他指出RLHF(基于人类反馈的强化学习)通过优化人类偏好,导致模型天然倾向于"过度承诺",产出的答案总是看起来貌似正确但实际可能有误。他重新解读了萨顿的"苦涩教训",提出"做对任务比数据重要,数据比算力重要",并以此解释Anthropic的崛起。他宣布自己的公司TypeSafe将放弃辅助能力,全力投入自动化方向,认为RLHF是AI发展史上一条本可避免的弯路,未来的AI应该是为自动化而生的,而不仅仅是更便宜地写代码。 https://www.youtube.com/watch?v=cJ0EOzey--o https://www.youtube.com/watch?v=TWeXnSoGIpg
Read Video · Transcripción e Insights
Genera transcripción e insights IA — Prueba gratis
Cuenta gratuita · sin tarjeta · 150 créditos al registrarte, suficiente para este episodio
- 📄 Transcripción completa con marcas de tiempo
- ✨ Resumen IA, palabras clave y mapa mental
- 💡 Puntos clave y citas destacadas
Episodios y videos listos para leer
Episodios de podcast

OpenAI Pauses Frontier Training, Elon's 100X Prediction Lands, Robot Beats Usain Bolt with Emad Mostaque | EP#282
Moonshots with Peter Diamandis
29 ago 20262:20:54EN
The Mystery of Sea Creatures (3/5): Could an orca give a TED Talk? | Karen Bakker
TED Talks Daily
18 jul 202615:27EN
Episode 68: How Timothy Baxter Built Baxter Research Into the Gold Standard of Criminal Research
Behind the Screens: Conversations with Background Screening Pros hosted by Les Rosen
20 jul 202659:28EN
Conflicto de roles: familia, trabajo y ocio | Experiencias compartidas | 419
Entiende Tu Mente
18 feb 202623:35ES
Paula Wehmeyer - VP Marketing und Talent bei General Intuition & Ex General Catalyst Partner (#40)
Mannheim Investment Club (MIC) Podcast
18 sept 202655:19DE
EP82《五次重置》:不想努力了怎么办?
纵横四海
23 may 20264:08:44ZH
Videos

21 Life-Changing Books Summarized in 18 Minutes
Daniel Pink
11 jun 202518:28EN
QWEN just CRASHED the industry
Wes Roth
4 ago 202615:21EN
Managed Agents - Don't Get Locked In
Sam Witteveen
13 sept 202610:49EN
Тайная мобилизация уже идет: облавы на улицах и нехватка солдат
Михаил Ходорковский
5 ago 20267:02RU
#39 Die Pest
99 mal Geschichte
8 ene 20261:02:15DE