«

开放数字市场,让AI为您选购与议价。

qimuai 发布于 阅读:17 一手编译


开放数字市场,让AI为您选购与议价。

内容来源:https://news.microsoft.com/signal/articles/open-digital-markets-agentic-economy/

内容总结:

微软研究团队呼吁构建开放AI代理经济,打破“数字围墙”

想象一下,未来你的AI助手不仅能聊天,还能主动“逛”数字市场,为你采购食品、预订机票,甚至与房东协商租赁条款。这种由AI代理代表用户与企业代理进行交互、自主完成复杂任务的场景,正从科幻走向现实。

微软研究院团队在最新发表于《ACM通讯》的论文中指出,构建一个“开放代理经济”是人工智能发展的最优路径,能为企业和个人创造最大价值。目前,AI代理多存在于“围墙花园”中——例如,用户只能在航空公司网站内使用其专属AI改签机票。而在开放代理经济中,将形成一个去中心化的“代理网络”,不同AI代理可以跨平台自由交互与合作,类似万维网的开放互联模式。

论文作者之一、微软研究员大卫·罗斯柴尔德表示,当前少数科技巨头已占据用户大量数字时间,它们有动力将用户锁定在自有生态中,但这可能抑制创新。“如果我们不现在开始讨论开放生态,惯性将把我们推向‘围墙花园’版本,那意味着社会整体福祉和机会的减少。”

研究团队认为,开放代理经济将使市场运行更高效:消费者能轻松匹配最佳服务商并自由切换,企业竞争焦点将从“吸引眼球”的广告投放,转向通过提升产品真实竞争力来赢得用户偏好,即从“注意力经济”转向“偏好经济”。

实现这一愿景需多方协同:需建立广泛采用的助理代理、服务代理及标准化交互协议;需构建类似“国际空间站”的中立数字市场,而非由单一公司主导的平台;还需技术公司、标准组织、服务商、政府及监管机构共同参与治理,以确保市场保持开放、竞争与安全。

罗斯柴尔德强调:“为确保开放市场的安全与繁荣,我们需要在技术层面构建一个既给予代理充分约束、又保障安全可控的空间。这是一项值得以开放研究与合作方式推进的重要权衡。”

该研究呼吁业界共同探索开放生态的技术框架与治理模式,避免AI时代的经济机会被封闭系统所垄断。

中文翻译:

预计阅读时间:5分钟

开放数字市场,让AI替你"购物"和"谈判"

想象这样一个世界:你的数字助手不仅能通过聊天回答问题,还能被派往数字市场采购食品杂货、预订航班,甚至为你协商公寓租赁条款。这些由AI驱动的智能体可以代表你与企业的智能体互动,全程无需你亲自动手。

这样的未来并非科幻。微软研究团队在最新发表于《ACM通讯》的论文中指出,这种开放的智能体经济将是人工智能发展的最优路径,能为企业和个人创造最大化的机遇。

何为开放智能体经济?

当前智能体经济初具雏形,我们接触的多数AI智能体仍处于"封闭花园"模式。例如改签航班时,你只能与航空公司官网的AI对话。而在开放智能体经济中,研究者设想将形成"智能体网络"构成的去中心化生态系统,AI智能体可以跨平台自由交互,突破单一网站的藩篱。

研究者将开放智能体经济比作万维网的理想形态——任何智能体都能相互交易。个人助理智能体将扮演类似浏览器的角色,服务商智能体则相当于各类网站。

论文作者之一、微软研究员大卫·罗斯柴尔德表示,团队选择探讨这一议题,正是担忧少数平台可能构建封闭生态阻碍创新。他指出:"当前少数公司垄断了我们的数字时间,它们会竭力将用户禁锢在自家平台内。"

"现在开启开放智能体经济的讨论至关重要,"罗斯柴尔德强调,"否则发展的惯性与便利性将把我们推向封闭花园模式,那意味着社会整体福祉和发展机遇的减损。"

为何需要开放智能体经济?

微软研究团队在论文中论证,让多元AI智能体在开放市场中运作是最佳发展路径。这种模式能使市场运行更顺畅,降低用户切换服务的门槛,并以去中心化方式让更多人获取数字工具与服务。

合著者、项目经理马特·沃格尔认为,开放智能体经济将形成双赢格局:"消费者能精准匹配需求并自由切换服务商,企业竞争焦点也将从广告投放转向产品优化。"

罗斯柴尔德进一步阐释:"我们预见从'注意力经济'向'偏好经济'的转型。品牌投入将真正用于提升产品品质与用户认知,而非单纯争夺曝光。"

如何构建开放智能体经济?

研究者指出,构建开放智能体经济需要多项关键技术突破与结构创新。首要条件是普及智能体应用并建立程序化交互协议。当个人与企业普遍使用AI智能体作为代表时,这些智能体必须具备自主协商能力。

交互应在中立的数字市场中进行——类似国际空间站的多方合作模式,而非任何单一企业控制的平台。为此需要建立标准化框架与协议,确保智能体既能相互发现又可安全交互。

尽管消费者将主导个人AI助手的训练,但研究者强调需要多方协同共建:科技公司、标准组织、企业服务商、政府及监管机构共同参与治理,才能保障新兴市场的开放性、竞争性与安全性。

罗斯柴尔德总结道:"为了在开放市场中实现安全繁荣的未来,我们需要开发更精密的技术框架——既给予智能体充分活动空间,又施加必要约束。这种权衡需要以开放研究、开放开发的方式推进。我们相信,为更开放的未来而奋斗是值得的。"

更多深度解析请参阅《智能体经济》专题。相关研究成果可访问微软研究博客关于Magentic Marketplace的开源智能体市场模拟环境。

题图由da-kuk/Getty Images提供。

英文来源:

– The estimated reading time is 5 min.
Opening digital markets so AI can shop — and negotiate — for you
Author
Imagine a world where you have a digital assistant that can do more than just respond to your questions in a chat. In this future, you could send your assistant out to a digital marketplace to order your groceries, book a flight, or even negotiate the terms of your apartment lease.
These AI-powered agents could interact with agents from companies on your behalf and advocate for you — all without you needing to lift a finger.
That future isn’t just for the world of science fiction. In a paper published Thursday in Communications of the ACM, Microsoft researchers say this kind of open agentic economy is the most beneficial way for AI to move forward, maximizing opportunity for both businesses and individuals.
What is an open agentic economy?
Currently, as the agentic economy takes shape, many of the AI agents we interact with exist in what’s called a “walled garden.” If you need help rescheduling a flight, for example, you might chat with the AI agent on the airline’s website. But in an open agentic economy, the researchers say, there’d be a “web of agents” that form a decentralized ecosystem where AI agents can interact freely with each other without being confined to the “walled garden” of just one website.
They compare the open agentic economy to the promise of the World Wide Web, in that any agent could transact with any other. Assistant agents would play a role similar to web browsers, and service agents similar to websites.
One of the paper’s authors, Microsoft researcher David Rothschild, says he and his colleagues chose to write about this topic because they’re concerned that “walled gardens” will pop up on a few major platforms and could stymie innovation. Rothschild points out that some companies have already built large platforms that they will want to keep existing users in.
“There are a very few companies that have captured all of our digital time,” he explains. “They have an enormous percentage of our attention, and they will do everything they can to keep you siloed into their platforms.”
Rothschild says starting the discussion of an open agentic economy now is critical.
“Otherwise, momentum and ease are going to push us into that version of walled gardens,” he says. “And that would mean less overall welfare and opportunity for society.”
Why have an open agentic economy?
In the paper, Microsoft researchers say that letting many different AI agents operate in an open market is the best path forward. In this setup, the agents help markets run more smoothly, make it easier for people to switch between services, and give more people access to digital tools and services in a decentralized way.
Co-author and Program Manager Matt Vogel says he believes an open agentic economy will be beneficial for both people and companies.
“Consumers can find the best business that matches their needs and very easily switch between them,” Vogel says. “They’re not stuck and locked into one of them.”
And businesses will move from attempting to get their product in front of people via advertising to improving the product, Rothchild says.
“We foresee a movement from what we’ll call the ‘attention economy’ to the ‘preference economy,’” he explains. “The hope is that there is still money being spent by brands to make them more successful, but they’re doing it in a way that really both improves the product and improves our understanding of the product rather than simply getting in front of us.”
How to get to an open agentic economy
The researchers say that to form an open agentic economy, several key technological and structural developments will be required. First, the widespread adoption of assistant and service agents and programmatic inter-agent communication. Once people and businesses adopt to using AI agents as their representatives, those agents must be able to communicate, unscripted, with each other.
This would be done in a neutral digital marketplace, akin to how the International Space Station was created in cooperation with businesses and governments, rather than on any one platform owned by a single company.
For the agents to do their jobs, the researchers say there would need to be standardized frameworks and protocols so the agents can both discover each other and conduct secure interactions.
While consumers would adopt and train their personal AI assistant agents, researchers say, several groups must be involved in building and governing the open agentic economy: technology companies, standards bodies, businesses/service providers, governments and regulatory bodies. The researchers contend this would help ensure the new marketplaces stay open, competitive and secure.
“To ensure a safe and prosperous future with an open marketplace, more technology needs to be developed to create a very close space in which the agents are highly constrained, but that also adds to some safety and some control,” Rothschild says. “This is a tradeoff that we think is important to confront in an open way — an open research way and an open development way. We believe that it’s worth it to fight for the more open future.”
Learn more in “The Agentic Economy.” For more information on related work, check out the Microsoft Research Blog post on Magentic Marketplace, an open-source simulation environment for studying agentic markets.
Lead image courtesy of da-kuk/Getty Images.

微软AI最新进展

文章目录


    扫描二维码,在手机上阅读