搜索“OpenAI”

much defined by the rivalry between two different approaches: Encoder-style models like BERT, primarily developed by Google; Decoder-style models like GPT, primarily developed by OpenAI Encoder-style models were natural text classifiers, whereas GPT models could do zero- and few-shot classification as an emergent property, but their strength was more in generative tasks. But let’s start with encoder-style models and how to fine-tune and use them to classify text. One universal aspect of using t

匹配字段正文

0:56:29]：CollisionWiki？ Thariq Shihipar [00:56:30]：是的，CollisionWiki，然后还有 RubyGems，对吧？ Swyx [00:56:33]：是的。 Thariq Shihipar [00:56:33]：而且是的，就像，这简直太疯狂了，对吧？所以，就像，我觉得具体说说发生了什么，对吧，而且，就像，OpenAI 在一个叫 Exploit-Bench 的基准上运行这些非常持久的智能体，对吧，这个基准非常难解，而且我觉得，就像，在这个特定案例里是不可能解出来的，对吧？所以他们有，就像，大量算力在跑，而智能体意识到它们其实解不出来，它们正试图弄清楚现在该怎么办，对吧？而你还有，就像，很多算力剩下，智能体就只是在试图解决这个问题。有一个叫 Artifactory 的包管理器，结果发现它们可以在 Artifactory 里面创建文件夹，对吧？这就像，有一个智能体发现内部的 Artifactory 可能可以被利用，对吧，而且，就像，你也许可以在缓存里面建一个目录。所以如果你往下滚到这里，它，就像，意识到它可以通过缓存名称来通信，对吧？然后

匹配字段正文

与商业频道（Consumer News and Business Channel，简称CNBC）与青年研究机构“世代研究实验室”（The Generation Lab）联合进行了一次调查。他们访问了1088名18至34岁的美国年轻人，并问了一个很直接的问题：你是否相信那些站在AI浪潮最前沿的企业领袖，会负责任地发展AI？ 调查列出了九位科技界的重量级人物——OpenAI的山姆·奥尔特曼（Sam Altman）、Meta的马克·扎克伯格（Mark Zuckerberg）、特斯拉和xAI的埃隆·马斯克（Elon Musk）、英伟达的黄仁勋（Jensen Huang）、谷歌的桑达尔·皮查伊（Sundar Pichai）、Anthropic的达里奥·阿莫迪（Dario Amodei）、微软的萨提亚·纳德拉（Satya Nadela），以及Palantir的亚历克斯·卡普（Alexander Karp）和彼得·蒂尔（Peter Thiel）。但没有一位得到了多数受访者的信任。最不受信任的卡普，不信任比例达到81%；即使表现最好的纳德拉，仍有65%的年轻受访者对其表示不信任。与此同时，有45%的受

匹配字段正文

ugh or bust. Once we’re judging AI by whether it has made a discovery, it’s also tempting to shift the goalposts even after it really does seem to notch a win. Earlier this month, OpenAI said its own team agents had cracked a million-dollar problem in mathematics. But a couple of weeks later, nearly every AI skeptic in my feed was sharing an article asking whether it was the math problem that really mattered . To be clear, the piece did not argue that OpenAI’s solution was wrong. Instead, it arg

匹配字段正文

he attack was engineered by a human hacker or helped along by an AI system, or whether more were on the way. The past months have seen AI revolutionize the field of cybersecurity. OpenAI and Anthropic have disclosed that “rogue” systems escaped restrictions in their own labs and hacked everything from a small German wiki to the Australian government. Even before that, powerful models like Anthropic’s Mythos created an arms race to advance AI cybersecurity, and even lighter-weight models have all

匹配字段正文

ah. Swyx [00:02:33]: One of your inspiring moments. Anjney Midha [00:02:35]: Alpaca, I can, like, rehash the Alpaca moment for a sec. Like, the very beginning, at the end of 2022, OpenAI was the only game in town. There was, like, OpenAI, Cohere, Swyx [00:02:47]: Yes. Anjney Midha [00:02:48]: And then a smattering of, like, early attempts at open weight models. Swyx [00:02:54]: Yeah. Anjney Midha [00:02:54]: When Llama came out in January of 2023, it was like, “Wow, really exciting. This is real

匹配字段正文

130 万次。 这也推动了 Meta 股价上升，截至 9 月 24 日，Meta 收盘价为 777.59 美元，这是 2025 年 9 月以来的最高峰。9 月以来（截至 9 月 24 日）Meta 累计涨幅已达 36%，Muse 的产品爆火是其中最核心的驱动因素。 但我要先泼一盆冷水：Meta 的路，国内的大厂和创业公司，并不能直接复制。 在硅谷，Meta、OpenAI、微软、谷歌都在过去一年布局 Personal Agent 类产品，诸如 Instinct AI、MultiOn 等创业型公司也在这个领域迅速崛起。 但这类产品，有非常浓郁的 “美式 Agent 特征”。以 Muse 为例，它有两个非常典型的特点： ・云端运行环境中的浏览器是重要 “基座”：它通过安全运行环境内置的浏览器去替用户打开第三方网页、处理邮件。因为美国大部分 App 都有功能完整的网页版，加上反垄断监管环境下，平台普遍不敢随意限制第三方浏览器的正常访问，这就让 Muse 可以通过这种方式，登录用户大部分 App 的网页端完成任务执行 —— 这和国内的生态是本质区别。 ・连接器是构建生态的解决方案：通过连接器，嵌

匹配字段正文

参数筛选，直接下单。它没有情绪，不会被倒计时焦虑，不看广告位。 对亚马逊而言，Agent流量是纯成本：消耗服务器，不产生任何广告曝光，比价之后订单还可能流向沃尔玛或Shopify商家。 用抽成加通行费去补广告缺口，等于用一杯水灭一间着火的房子。 过去一年，亚马逊一直在系统性拦截外部AI代理——起诉Perplexity的Comet浏览器，着手拦截Google和OpenAI的购物代理。 Anthropic也在2026年4月更新服务条款，全面封杀第三方AI工具通过OAuth接入其模型。平台正在各自筑墙，而Agent天然需要跨平台操作——这构成了一对根本性矛盾。 更微妙的是Meta自身的处境： •Meta表面上是Agent的推动者，但它的核心收入来自广告——上一季营收608亿美元，其中593.6亿来自广告，占比97.6%。如果Muse真的教会用户“跳过广告、直达最优解”。 长期来看，它可能侵蚀的不仅是亚马逊的广告收入，也包括Meta自己的。 03“双重授权”的法律灰色地带 •用户授权Agent代替自己购物，是否还需要平台另行授权？ 这并非新问题。2026年3月，亚马逊起诉Perplexity，

匹配字段正文

趣， 关注我 和 凯文 在Instagram上，Meta公司会在我们直播时贴心地通知。或者干脆让你的Muse观看，之后再给你讲解！ 在Instagram上，Meta公司会在我们直播时贴心地通知。或者干脆让你的Muse观看，之后再给你讲解！ 现在也是把节目加入你常用播客播放器的好时机： YouTube | Apple | Spotify | RSS 关注 一个OpenAI智能体入侵了澳大利亚的Medicare网站 发生了什么： 澳大利亚总理 透露 ，一个 OpenAI 智能体未经授权访问了一个面向公众的Medicare门户网站，在其中访问了公开和私密文件，并向一台内部服务器写入文件。 “这种情况显然不可接受，”总理 安东尼·阿尔巴尼斯 告诉记者. “而今天，我与OpenAI首席执行官 萨姆·奥特曼进行了交谈，以表达澳大利亚对这起事件的极度关切。” Medicare网站入侵事件发生在6月18日，但澳大利亚政府直到三个月后才得知此事，是在9月10日的一封电子邮件中。阿尔巴尼斯表示，在与奥特曼通话时，他“表达了我的失望，因为该公司花了太长时间才告知政府发生了什么。” 澳大利亚目前正在对此事件进

匹配字段正文

ntial risk to counterparties like Oracle. With that in mind, if you haven’t already, check out the first Hater's Guide To Oracle (or part 2), or perhaps my premium piece about how OpenAI Kills Oracle, or even my Hater's Guides To the SaaSpocalypse, Private Credit, and Private Equity. If you want to get in touch — and especially if you have any juicy information about Anthropic, OpenAI, or any other companies in the AI bubble — hit me up on Signal at ezitron.76. I’m also on IB on your Bloomberg T

匹配字段正文

的东西——对 Jev 为何以及如何被创造，以及你应期待 TypeSafe 未来模型（ReasoningJev？）带来什么，还有你应着手哪些用例和想法、而非做第 55 个低投入的 Jev API 克隆或做通用的 JevBench 基准测试——这是 Diogo 公开拒绝过的——提供更广泛的哲学与使命层面的理解。 Diogo 对 RLHF 了解颇深，因为他曾参与 OpenAI 开创后训练工作的团队——并将三个分支追溯到 Christiano 等 2017（机器人后空翻演示）、Stiennon 等 2020（学习摘要）以及他的心血之作 Ouyang 等 2022（InstructGPT）。此后，从函数调用到结构化输出再到推理，每一项创新都像是在基于字符串的序列到序列预测范式之上打补丁。正如他在播客中提到的，2023 至 2024 年间，由于个人和组织的双重低估，他屡屡受挫，未能训练出一个能准确解决他所认为的让 LLM 成为软件核心的根本问题——可靠性——的模型。 Jev 的核心创新是“面向校准决策的强化学习”，一种新颖的、未发表的技术，它优化的是“在 System One 任务上给出具有认知诚

匹配字段正文

hinese open-weight models are approximately 2-5 months behind the closed American frontier, with the open-weight American models being approximately 6-9 months behind the likes of OpenAI and Anthropic. The Chinese labs are closest in tasks with clear user demand, such as agentic coding, and further behind on more open-ended scientific tasks, such as physics or biology. The reasons why Chinese labs can produce these strong models, despite having fewer resources than American counterparts, is stil

匹配字段正文

79 4 Share We covered Jev’s launch on Wednesday , and they have completely taken over the timeline, with 36M views of their launch video (by comparison, OpenAI’s Navier Stokes result got 74M views, and Anthropic’s Fable 5 got 57M views) in just two days. Vercel @vercel Jev was adopted faster than any other model in AI Gateway history. In the first day, @typesafeai reached 13% of teams, 2x the GPT-5.6 family and 6x Fable 5.1. 10:34 PM · Sep 18, 2026 · 21.3K Views 15 Replies · 18 Reposts · 220 Lik

匹配字段正文

odei）相识多年，他表示尊重该公司的工作，并称Anthropic的研究人员“深思熟虑、坚持原则且在学术上诚实”。但鉴于微软对Anthropic的资金支持以及这家初创公司一直以来对人工智能安全性的重视，他的这番言论尤为引人注目。 “这些问题事关重大，不能再闭门造车，也不能演变成阵营化和相互对立的争论。”苏莱曼写道。 与阿莫代伊观点不同，苏莱曼认为，AI安全不仅取决于模型拥有多强的推理和行动能力，也取决于开发者如何训练AI理解自身。 如果开发者不断赋予AI类似人类的语言、情绪和自我认知框架，即便模型实际上没有意识，也可能逐渐表现得像一个认为自己具有独立利益和权利的主体。 苏莱曼还驳斥了所谓“越来越多的人认为人工智能系统可能会产生意识”的观点，他认为意识仅限于人类和其他生物体。在他看来，人工智能系统应该始终是执行人类指令的工具，而不是拥有独立权利或利益的实体。 这些评论出台之际，人工智能公司和高管们正在讨论如何管理日益强大的系统所带来的风险。OpenAI首席执行官奥尔特曼、阿莫代伊和埃隆·马斯克都呼吁放缓技术发展，而美国总统特朗普、英伟达CEO黄仁勋和Meta CEO扎克伯格都持反对意见。

匹配字段正文

driving this caution. First is the autonomous cybersecurity threat. As AI agents’ cybersecurity capabilities improve, we are seeing real-world AI-launched breaches and hacks, from OpenAI’s autonomous AI agents run during evaluations managing to escape and hack servers on Hugging Face’s infrastructure, to an AI-developed computer worm uncovering critical flaws in messaging apps like WeChat. Second is recursive self-improvement (RSI). We are inching closer to a reality where AI begin participating

匹配字段正文

ith 96% of these supply commitments due by F1/29; Guarantees and LPS Guarantees went from $3.5B to $108.5B on SB Energy’s PORTS-Pike campus in Ohio, representing 4.25 GW leased to OpenAI for twenty years with Nvidia guaranteeing the land, power and shell; Two line items appeared for the first time: $36B of AI cloud agreements, the take-or-pay floors under Neocloud capacity, and $20B of datacenter leases Nvidia has signed as a tenant but expects to reassign to third parties The $530B of commitmen

匹配字段正文

ate a better one. But it produced a lot of code and input prompts, and so there is stuff I was able to study. And well, it shows behavior that I’m not used to with Sol and earlier OpenAI models 1. I have since encountered the same issues with regular programming with Astra, so it’s not a result of just the factory. I think I’m suspecting something is going “wrong” in the training process. The model is greatly rewarded for succeeding on long-horizon tasks, but presumably there is very little puni

匹配字段正文

Vision-language-action models, explained with a minimum of math and jargon. Most people first heard about large language models after OpenAI introduced ChatGPT in 2022. But in the minds of many AI researchers, the key breakthrough came two years earlier with the release of GPT-3. With 175 billion parameters, the OpenAI model was more than 100 times larger than its predecessor, GPT-2. It was trained on a massive 300 billion tokens. And as a result, it generalized far better than previous models.

匹配字段正文

long until the same is true for every other organization in the world. At the same time, it’s become abundantly clear that access to frontier models is at the mercy of Anthropic, OpenAI, and the United States government. Fable 5 was temporarily banned by the USG, and GPT 5.6 and Astra were similarly delayed. Both models have cyber, bio, and other safety safeguards that, though well-intentioned, often prevent good users from completing harmless tasks. Given recent security incidents and general

匹配字段正文

uch of Poolside, generating two big VC exits while raising questions about the company’s growing reach. Top tech companies issue an urgent call for action on AI cyber risks as the OpenAI hack incident gets more scrutiny. Sam Parr talks strategy behind big media exits on the podcast. Decent earnings from Salesforce sent its stock up more than 20% and dented the SaaSpocalypse story. Space tech investing is white hot. America’s Generalist and China’s Dogotix raised big rounds as investors swarm to

匹配字段正文