内容来源：https://www.wired.com/story/openai-atlas-browser-chrome-agents-web-browsing/

内容总结：

美国人工智能公司OpenAI于本周二正式发布旗下首款独立浏览器“Atlas”，该产品深度整合了ChatGPT智能对话功能，标志着硅谷科技企业围绕生成式AI展开的浏览器革新竞赛进入新阶段。

这款浏览器突破传统设计理念，将AI助手置于交互核心。用户可通过侧边栏实时向ChatGPT咨询网页内容，更可授权AI代理自动执行点击操作、完成在线任务。OpenAI首席执行官萨姆·奥尔特曼在直播发布会上表示：“AI技术带来了十年一遇的浏览器革新契机。自多标签页功能问世后，这个领域已沉寂太久。”

目前Atlas已向全球macOS系统的ChatGPT用户开放，Windows和移动版本正在开发中。基础功能免费使用，但AI代理等高级特性仅向ChatGPT Plus或Pro订阅用户开放。

与传统搜索引擎体验形成鲜明对比的是，Atlas将智能对话结果作为首要呈现内容。当用户搜索电影评论时，系统会优先展示聊天机器人风格的整合答案，传统网页链接则转为次级选项。项目主设计师瑞安·奥鲁克透露：“我们重构了搜索逻辑，用户可自由切换至链接、图片、视频等分类标签。”

值得关注的是，浏览器内置的“记忆存储”功能可记录用户搜索历史，通过分析行为模式智能推荐兴趣内容，甚至自动恢复曾访问的相关网页。这种持续学习能力建立在此前ChatGPT记忆功能基础之上。

面对行业变局，谷歌已为Chrome浏览器注入包括Gemini聊天机器人在内的多项AI功能，微软早年也在Edge浏览器中整合Bing AI侧边栏。新兴企业Perplexity今年推出的免费AI浏览器Comet同样引发市场关注。尽管OpenAI此前曾在ChatGPT中测试网页任务代理工具，但实际运行速度与准确性仍有提升空间，此次独立浏览器的推出将推动行业进入新一轮技术迭代周期。

中文翻译：

当地时间本周二，美国开放人工智能研究中心（OpenAI）宣布推出一款名为"Atlas"的新型浏览器，这款浏览器将与ChatGPT直接整合。Atlas内置了侧边栏窗口，用户可向ChatTSsT咨询与所访问网页相关的问题。该浏览器还配备了人工智能助手，能代替用户点击操作并完成各类任务。

OpenAI首席执行官萨姆·奥尔特曼在发布会直播中表示："我们认为人工智能代表着十年一遇的机遇，让我们重新思考浏览器的定位。标签页功能固然出色，但此后浏览器领域再未见重大创新。"Atlas的推出正值硅谷争相运用生成式人工智能重塑网络体验之际。谷歌也为其广受欢迎的Chrome浏览器推出了大量AI功能，包括可启动Gemini聊天机器人的"闪光"按钮。目前Chrome仍是全球使用最广泛的浏览器。

OpenAI表示Atlas浏览器即日起向全球macOS系统的ChatGPT用户开放，Windows系统和移动端版本正在开发中。该浏览器可免费使用，但其人工智能助手功能仅面向ChatGPT Plus或ChatGPT Pro订阅用户开放。

浏览器主设计师瑞安·奥鲁尔克在直播中介绍："我们通过Atlas对ChatGPT的搜索功能进行了重大升级。"当用户在Atlas搜索栏查询影评时，系统会优先呈现聊天机器人风格的答案，而非传统谷歌搜索常见的蓝色链接集合。

如今除了核心搜索结果，用户还能切换至其他标签页查看相关网站链接、图片、视频或新闻资讯。这与谷歌Chrome的体验形成对比：Atlas以AI聊天助手为核心，网站链接列表和图片结果反居次要地位。

直播中重点演示的另一项功能是"浏览器记忆"。这个可选功能基于ChatGPT现有记忆工具升级而来，能根据用户历史交互记录存储个人信息。浏览器可追溯用户过往搜索记录，据此推荐感兴趣的内容和操作建议，例如自动执行检测到的常规网络操作，或返回曾访问过且对当前项目有帮助的网页。

过去数年里，科技巨头与初创企业持续尝试将AI植入浏览器。微软是早期探索者之一，曾将名为"必应"的AI工具以侧边栏形式集成至Edge浏览器。此后Opera、Brave等浏览器厂商也持续优化各类AI整合方案。引人注目的还有Perplexity公司今年推出的免费AI浏览器Comet。

尽管这是OpenAI首次推出独立浏览器，但并非该公司首次尝试融合生成式AI与网络浏览。今年早些时候，OpenAI曾在ChatGPT内推出智能助手工具，可代用户执行比价选品等任务。《连线》杂志实测发现，当放任该机器人在网上点击操作、完成调研甚至下棋时，其反应迟缓且时有差错。

英文来源：

OpenAI announced on Tuesday it’s rolling out a new internet browser called Atlas that integrates directly with ChatGPT. Atlas includes features like a sidebar window people can use to ask ChatGPT questions about the web pages they visit. There’s also an AI agent that can click around and complete tasks on a user’s behalf.
“We think that AI represents a rare, once-a-decade opportunity to rethink what a browser can be about,” OpenAI CEO Sam Altman said during a livestream announcing Atlas. “Tabs were great, but we haven’t seen a lot of browser innovation since then.”
Atlas debuts as Silicon Valley races to use generative AI to reshape how people experience the internet. Google has also announced a plethora of AI features for its popular Chrome browser, including a “sparkle” button that launches its Gemini chatbot. Chrome remains the most used browser worldwide.
OpenAI says the Atlas browser will be available starting today for ChatGPT users globally on macOS. Windows and mobile options are currently in the works. Atlas is free to use, though its agent features are reserved for subscribers to OpenAI’s ChatGPT Plus or ChatGPT Pro plans.
“We’ve made some major upgrades to search on ChatGPT when accessed via Atlas,” Ryan O’Rouke, OpenAI’s lead designer for the browser, said during the livestream. If a user asks for movie reviews in the Atlas search bar, a chatbot-style answer will pop up first, rather than the more traditional collection of blue links users might expect when searching the web via Google.
Now, in addition to that result, users can switch to other tabs to see a collection of website links, images, videos, or news related to their queries. It’s a bit of an inversion of the Google Chrome experience. Rather than the search result being a collection of links with AI features added on top of that, the AI chatbot is central in Atlas, with the list of website links or image results as secondary.
Another feature OpenAI highlighted in the livestream is Atlas’ ability to collect “browser memories.” The capability is optional, and is an iteration of ChatGPT’s existing memory tool that stores details about users based on their past interactions with the chatbot. The browser can recall what you searched for in the past and use that data when suggesting topics of interest and actions to take, like automating an online routine it detects or returning back to a website you previously visited that could be helpful for a current project.
Tech giants and smaller startups have been experimenting with baking AI into web browsers for the past several years. Microsoft was one of the first movers when it threw its AI tool, called Bing at the time, into its Edge browser as a sidebar. Since then, browser-focused companies like Opera and Brave have also continued to tinker with different AI integrations. Another notable entry in the AI browser wars is Perplexity’s Comet, which launched this year and is also free to use.
While this is OpenAI’s first foray into a stand-alone browser, it’s not the first time the company has attempted to blend generative AI and web surfing. Earlier this year, OpenAI debuted an agent tool inside ChatGPT that is designed to be sent off to complete tasks for the user, like comparing multiple products and deciding which could be worth buying based on the person’s preferences and lifestyle. When WIRED experimented with letting the bot click around on the web, attempt to complete research, and even play chess, the results were slow and sometimes inaccurate.