GPT-4
PulseAugur coverage of GPT-4 — every cluster mentioning GPT-4 across labs, papers, and developer communities, ranked by signal.
16 天有情绪数据
-
Meta 的 Llama 2 超越开源 LLM 排行榜，支持商业用途
Meta 发布了 Llama 2，一个开源的大型语言模型，它迅速成为同级别模型中的最先进水平，性能超越了其他开源模型。该模型在 2 万亿个 token 上进行了预训练，具有更长的上下文长度，并在人类反馈强化学习方面进行了大量投入。Llama 2 可用于商业用途，满足了不能将敏感数据发送给外部 API 提供商的组织的关键需求，并为 GPT-4 等专有模型提供了替代方案。
-
NVIDIA AI研究员、Replit CEO探讨多模态AI的未来
Replit 举办了一场网络研讨会，邀请 NVIDIA AI 研究员 Jim Fan 和 Replit CEO Amjad Masad 共同探讨生成式AI的进展。讨论强调了AI多模态日益增长的重要性，通过整合图像、视频和3D数据，实现更丰富的系统交互。他们还谈到了大型语言模型的演变、ChatGPT界面等用户体验的改进，以及模型能力超越参数数量的趋势。讨论最后预测了AI对编码和各行业的未来影响，并重点介绍了Replit自身的AI编码助手…
-
LLM-powered Biographies
Eugene Yan experimented with several large language models, including GPT-4, Claude-v1.2, and Cohere-xlarge, by asking them to generate his biography. He observed that while the models captured the general essence of hi…
-
OpenAI 发布 GPT-4，多模态模型在基准测试中表现出人类水平的性能
OpenAI 发布了 GPT-4，一个大型多模态模型，能够处理文本和图像输入以生成文本输出。该新模型在各种专业和学术基准测试中表现出人类水平的性能，显著优于其前身 GPT-3.5。OpenAI 通过广泛的对齐工作，提高了 GPT-4 的事实性、可控性和对安全护栏的遵守程度。该模型的文本功能可通过 ChatGPT 和 API 使用，图像输入功能正在准备更广泛地发布。
-
2023 Year In Review
METR, an AI safety research organization, detailed its 2023 accomplishments, including developing methodologies for evaluating AI agents on autonomous tasks and contributing to OpenAI's GPT-4 system card. The organizati…
-
Enterprises prioritize data foundations for scalable AI adoption and business value
Organizations are increasingly adopting AI not just for isolated use cases but as a strategic portfolio of value models to drive business reinvention. This involves building workforce fluency, enabling AI-native custome…
-
OpenAI secures $40B funding, advances AI safety research and regulation
OpenAI has announced significant funding rounds, with one raising $6.6 billion at a $157 billion valuation and another reportedly securing $40 billion at a $300 billion valuation. The company is also focusing on AI safe…
-
OpenAI、Google DeepMind 加强人工智能网络安全防御，应对不断演变的威胁
OpenAI 正在积极应对人工智能在网络安全领域的双重用途特性，详细介绍了其加强防御和减轻滥用的努力。该公司正在增强其模型在代码审计和漏洞修复等防御任务中的能力，旨在使防御者能够应对日益复杂的威胁。OpenAI 还报告称，已挫败了五个与国家有关联的威胁行为者的活动，并指出当前的人工智能模型在恶意网络行动方面提供的能力有限，仅比现有工具略有提升。
-
OpenAI trains AI with human preference feedback; Chip Huyen proposes predictive model routing
OpenAI and DeepMind have developed a new algorithm that learns desired behaviors from human feedback, reducing the need for explicit goal functions. This method uses a three-step cycle where humans compare two agent beh…
-
Ilya Sutskever 离开 OpenAI；Jakub Pachocki 被任命为新首席科学家
Ilya Sutskever 将离开 OpenAI，Sam Altman 宣布 Jakub Pachocki 将接任首席科学家一职。Pachocki 此前曾负责 GPT-4 和 OpenAI Five 的研究，现在将领导公司迈向 AGI 的进程。OpenAI 还概述了几个关键研究领域，包括检测隐蔽 AI 系统、构建用于编程竞赛的代理、网络安全防御以及创建复杂的代理模拟。