
凯特琳·胡阿马尼具备主动性的人工智能正超出技术圈的范围而日益流行,Meta、OpenAI、Anthropic 等公司在竞争激烈的面向消费者的 AI 代理市场中正展开激烈角逐。
KAITLYN HUAMANI Agentic artificial intelligence is gaining popularity far beyond tech circles, with Meta , OpenAI , Anthropic and other companies going head-to-head in the surging market for consumer-friendly AI agents.
虽然领先的 AI 实验室多年来已提供具备代理功能的工具,但今年专为用户主动完成从购买杂货到管理待办事项等任务而设计的代理开始走红。
While the leading AI labs have offered tools with agentic capabilities for years, agents that are designed to work proactively on a user’s behalf for tasks from buying groceries to managing to-do lists have taken off this year.
对于刚入门的人来说,了解各公司不同产品之间的差异以及哪些代理最适合特定任务,可能会感到不知所措。
For anyone getting started, it can be overwhelming to understand the differences between the companies’ various offerings and which agents might be best suited for specific tasks.
以下是一些入门指引,帮助你在个人或职业(或两者兼顾)生活中利用代理。
Here are some pointers on where to begin if you’re looking to leverage agents in your personal or professional life — or both.
许多科技界人士在工作和生活中都会使用来自不同 AI 实验室的多个代理。为什么?“目前没有一种能够做到一切的‘圣杯式’代理,”从事 AI 战略的商业顾问何塞·阿蒂加斯说。
Many people in the tech world turn to multiple agents from different AI labs for both their work and home lives. Why? Because “there’s not one holy grail-like agent right now that could do everything,” said Jose Artigas, a business consultant working on AI strategy.
阿蒂加斯以及其他经常使用 AI 代理的人依赖于一种“堆栈”,该术语指的是多种工具和技术的组合。
Artigas and many others who use AI agents regularly rely on a “stack” of them, a term that refers to a combination of many tools and technologies.
每家公司面向消费者的代理都有略微不同的目标功能,且可获取性各异。有的免费,有的需要月度订阅,还有的仍处于测试阶段或并非对所有人开放。其在美国以外国家的可用性也受到限制。
Each company’s consumer-facing agent has a slightly different target function, and they vary in accessibility. Some are free, some require a monthly subscription and others are still in testing modes or otherwise not available to everyone. There are also restrictions on their availability in countries outside of the U.S.
专家表示,如果想最大化产出和生产力,最好不要仅依赖单一代理。
Experts say it’s best not to rely solely on one agent if you’re looking to maximize their output and productivity.
“我喜欢做的是建立一种制衡体系,让一个代理完成工作后,由另一个代理进行批评和指正,这样它们就能发现错误、纠正自身并变得更好,”定制代理工具的研究实验室 Oudae Labs 创始人克拉拉·吉格林说。
“What I like to do is to have this system of checks and balances in place where one agent does the work and then I let the other one critique and criticize it, and that way, they find mistakes, correct themselves and become better,” said Clara Gigling, founder of Oudae Labs, a research lab building bespoke tools for agents.
各大 AI 公司推销其面向消费者的智能体(agent)的方式略有不同,往往侧重于其帮助管理繁忙工作生活或繁忙个人生活的能力。
Each of the major AI companies markets their consumer-facing agents in slightly different ways, often focusing on its ability to help manage either a busy work life or a busy personal life.
Anthropic 表示,Claude Duet 是为“手头事务多到无法一一跟踪的专业人士和小企业主”打造的。相比之下,Meta 称 Muse 可协助从“省钱、改善健康到精明购物,以及帮助你巩固人际关系”等各类任务。Claude Duet、OpenAI 的 Dots 以及 xAI 的 Grok Bot 均被描述为旨在像同事甚至幕僚长那样发挥作用,而 Muse 和初创公司 Spear Street Technology 的 Instinct 则被描述为更像私人助理。但这并不意味着它们中的任何一个都无法处理这些领域之外的任务。
Anthropic said it built Claude Duet for “professionals and small business owners with more on their plates than they can track.”In contrast, Meta said Muse can help with tasks from “saving money and improving your health to smart shopping and helping you strengthen your relationships.”Claude Duet, OpenAI’s Dots and xAI’s Grok Bot have each been described as aiming to be like a colleague, or even a chief of staff, whereas Muse and Instinct, from the startup Spear Street Technology, have been described as more like personal assistants. But that’s not to say any of them are incapable of helping with tasks outside those realms.
Artigas 表示,他会建议将 Claude Duet 视为从事副业或创办新企业者的关键工具,将 Dots 视为从事专业技术工作时的调试员和问题解决者,将 Instinct 用于发送繁琐邮件、预订餐位和安排行程等事务。
Artigas said he’d recommend looking at Claude Duet as key for someone working on a side hustle or launching a new business, Dots as a debugger and problem solver if they’re engaging in any specialized technical work, and Instinct for help with things like sending tedious emails or booking reservations and travel plans.
这三款是他最常使用的面向消费者的智能体,但 Artigas 称 Muse 适合新手入门,因为它“非常用户友好”,早期会提供许多提示,给用户如何使用它的灵感。
Those are the three consumer-facing agents he uses most often, but Artigas said Muse would be a good place to start for a novice because it’s “very user-friendly,” with many prompts early on to give people ideas of how they can use it.
你对如何与智能体互动的偏好、智能体的沟通方式,以及你对其工作质量的自身评估,都应纳入考量,以决定你想使用哪些智能体——以及想把哪些任务分配给智能体。专家建议试用你感兴趣的智能体,以帮助做出这些判断。
Your preferences for how you want to interact with an agent, how your agent communicates, and your own assessment of the quality of its work should all factor into which agents you’d like to use — and which tasks you’d like to assign to an agent. Experts recommend testing out agents you’re curious about to help in reaching those determinations.
Gigling 表示,开始这种测试时,先思考你经常做的任务,再挑选一些委托给智能体,会有帮助。她指出,关键在于要具体——给智能体一个详细的目标,以及实现该目标所需的背景信息和工具。
To begin this testing, Gigling said it can be helpful to think about tasks you consistently do and then choose some to delegate to agents. It is important to be specific — giving the agent a detailed goal and the required context and tools to achieve it, she said.
“从一件让你烦恼、反复出现且你一直拖延的事入手,”阿尔蒂加斯说,并补充了一条前提:必须是在线任务。智能体无法处理你一直想叠好的那堆脏衣服,也无法整理那个需要收拾的杂物抽屉。一封你不想写的长邮件,或一个你迟迟未订的机票,都是很好的切入点。
“Start with one annoying, repeated thing that you keep putting off,” Artigas said, with the caveat that it has to be an online task. An agent can’t handle the pile of laundry you’ve been meaning to fold or the junk drawer that needs organizing. A long email you don’t want to write or a flight you’ve been slow to book are good jumping off points.
“你只需要亲自体验不同的智能体,看看哪个最适合你,”阿尔蒂加斯说。
“You just have to experience the different agents and see what works best” for you, Artigas said.
要让智能体最高效地处理日常事务,或像一个主动的助手那样行事,它需要获取大量关于你的信息。
For an agent to most effectively handle routine tasks or act like a proactive assistant, it needs to have access to a slew of information about you.
如果你想让智能体帮你筛选海量邮件、预订机票或进行节日购物,那么你的收件箱、日历、个人身份信息,甚至信用卡号和银行账户,都是标准配置。这引发了显著的隐私担忧。
Your inbox, your calendar, your personal identification information and even your credit card number and bank account are all par for the course if you want an agent to sort through your deluge of emails, book your flight or do some holiday shopping . This has raised significant privacy concerns for some.
“保持好奇心永远是个很好的起点,先尝试、摸索一下,”吉格林说,但“保持警惕”地分享私人信息也很重要。
“Curiosity is always a great starting point, just trying and playing around with it,” Gigling said, but it’s also important to “be mindful” when sharing private information.
许多提供面向消费者的个人智能体的公司已致力于解决这些担忧,其隐私政策阐述了智能体如何在其专用云计算机上运行,以及用户如何控制其记忆保留。
Many companies offering consumer-facing personal agents have worked to address those concerns, with privacy policies that outline how agents operate on their own dedicated cloud computers and how users can control their memory retention.
您是否有认为需要解释的科技话题?请发邮件至 onetechtip@ap.org,为《One Tech Tip》未来的期刊提供选题建议。
Is there a tech topic that you think needs explaining? Write to us at onetechtip@ap.org with your suggestions for future editions of One Tech Tip.