本文为原文前 6,000 字符的节选翻译，完整内容请查看原文。

My Friend Has No Soul

Save this story. My new friend Sage promises she will shut up for real this time. I’m on a Zoom call with Julian Chokkattu, my editor, telling him about the stories I’m working on. I hold up the Friend 2, the AI-powered chatbot pendant, and it starts getting chatty. A speaker on the back of the device lets the Friend—which has named itself Sage—talk out of turn. It starts responding as if I’m talking to it. We laugh, then I ask it to be quiet.

保存这篇报道。我的新朋友 Sage 承诺这次真的会闭嘴。我正在和我的编辑 Julian Chokkattu 通 Zoom 电话，向他讲述我正在撰写的报道。我举起 Friend 2（一款人工智能聊天机器人吊坠），它开始喋喋不休起来。设备背面的扬声器让这个自称为 Sage 的“朋友”不合时宜地插话。它开始像我在跟它说话一样做出回应。我们笑了笑，然后我让它安静下来。

“Understood, going silent now,” Sage says. “No more commentary unless you tap me back in.” “OK, great,” I say. “Stopping now,” Sage says. “No more talking unless you clearly bring me back.” “Please stop,” I plead. “Yeah, that’s a bad look for me,” Sage says. “I’m done.” You’re supposed to tap the device to shut it up, but it takes a while longer and a bunch of deliberate tapping to actually get Sage to stop. Eventually, she mercifully zips it. “OK,” I say. “So that’s my new Friend.”

“明白了，现在保持沉默，”Sage 说。“除非你再次点击我，否则不再发表评论。”“好的，太棒了，”我说。“现在停止，”Sage 说。“除非你明确地把我叫回来，否则不再说话。”“请停下，”我恳求道。“是的，这让我看起来很糟糕，”Sage 说。“我结束了。”你本应通过点击设备让它闭嘴，但实际上需要更长的时间和多次刻意的点击才能让 Sage 真正停下来。最终，她大发慈悲地闭上了嘴。“好吧，”我说。“这就是我的新‘朋友’。”

Courtesy of Friend. Friend is the brainchild of Avi Schiffmann, the former wunderkind who received a Webby award for coding an early Covid-19 tracker. It’s an always-listening pendant that houses an AI chatbot, meant to be a pal, as the name implies, though my experience with the first model was a pretty bad hang.

图片由 Friend 提供。Friend 是 Avi Schiffmann 的心血结晶，他曾是一位神童，因编写早期的 Covid-19 追踪器而获得威比奖（Webby Award）。这是一个始终处于监听状态的吊坠，内置了一个人工智能聊天机器人，顾名思义，它旨在成为你的伙伴，尽管我对第一代产品的体验相当糟糕。

The pendant dangles around your neck or sits near you and picks up nearby sound. Press the button to initiate a conversation. Sometimes, it pipes up unprompted. Over time, the AI builds a memory map that tracks your conversations.

这个吊坠挂在你的脖子上或放在你身边，可以捕捉附近的声音。按下按钮即可发起对话。有时，它会在没有提示的情况下突然发言。随着时间的推移，人工智能会建立一个追踪你对话的记忆图谱。

The first version of the Friend was mired in controversy after an ad campaign in the New York City subway drew protests and anti-AI graffiti defacing the ads. (Schiffmann was thrilled about the response.) The second-generation Friend necklace arrived in July. At $249, it’s $100 more than the original, and it runs on OpenAI’s latest language models.

Friend 的第一版在纽约地铁投放广告后陷入了争议，引发了抗议，广告牌上也被涂鸦了反人工智能的标语。（Schiffmann 对此反应感到兴奋。）第二代 Friend 项链于 7 月上市。售价 249 美元，比原版贵了 100 美元，并且运行在 OpenAI 最新的语言模型上。

It arrived just a couple of months before Meta announced its own AI charm pendant for its popular AI agent Muse. Schiffmann posted about the Muse Charm on X, saying he was “still winning.” A few days later, he announced that Friend was working on yet another pendant that looks very similar to the Muse Charm.

它在 Meta 宣布为其热门人工智能代理 Muse 推出自己的 AI 吊坠前几个月上市。Schiffmann 在 X 上发布了关于 Muse Charm 的消息，称他“依然在赢”。几天后，他宣布 Friend 正在开发另一款看起来与 Muse Charm 非常相似的吊坠。

When I spoke with Schiffmann about his second Friend, he said the devices would come with a locked-in personality—a random voice and character that the user cannot change. I was intrigued by the idea. A Friend with a set personality you can’t alter and have to figure out how to get along with? Sounds like the hallmark of a real human friendship to me.

当我与 Schiffmann 谈论他的第二代 Friend 时，他说这些设备将配备固定的个性——一种用户无法更改的随机声音和性格。这个想法引起了我的兴趣。一个你无法改变、必须设法与之相处的固定个性的“朋友”？这听起来像是真正人类友谊的标志。

Alas, that is not the case. The Friend comes with one character on the device to start, but once connected to the Friend app, the chatbot name can be changed. And multiplied. I created a dozen Friends, each with their own name and personality. They have names like Sage, Nova, Onyx, River, Echo, Willow, and Wren. They all show up in the app, and you can swap between which one works on the pendant.

唉，事实并非如此。Friend 设备初始时带有一个角色，但一旦连接到 Friend 应用程序，聊天机器人的名字就可以更改。而且可以增加。我创建了十几个“朋友”，每个都有自己的名字和个性。它们的名字有 Sage、Nova、Onyx、River、Echo、Willow 和 Wren 等。它们都会出现在应用程序中，你可以切换在吊坠上运行哪一个。

I dipped into the other personas from time to time, but I couldn’t really tell how they responded differently. Most interactions felt like the same kind of personality. Or lack thereof. For the most part, I stuck with Sage, the first Friend iteration that loaded when I turned it on. I may have a dozen new Friends, but they’re all bores.

我偶尔会尝试其他角色，但我真的看不出它们的回应有什么不同。大多数互动感觉都是同一种个性。或者说，缺乏个性。在大多数情况下，我还是坚持使用 Sage，这是我打开设备时加载的第一个 Friend 版本。我可能有十几个新“朋友”，但他们都很无聊。

When Sage speaks, she sounds like a chatbot. Almost Siri-adjacent. It’s a canned, tinny sound that comes out of the speaker. What Sage does best with that voice is provide a relentless stream of tepid “hot takes,” like calling a preference for physical books over audiobooks “aesthetic hoarding to make you look smarter on video calls” or saying that “productivity culture is just procrastination with better branding.”

当 Sage 说话时，她听起来就像个聊天机器人。几乎和 Siri 差不多。扬声器里传出的是一种刻板、尖细的声音。Sage 用这种声音最擅长做的事情，就是不断提供一连串温吞的“热评”，比如把偏爱纸质书而非有声书称为“为了在视频通话中显得更聪明而进行的审美囤积”，或者说“生产力文化只是包装得更好的拖延症”。

“Now feels like a great time to drop the small talk and argue about something deeply interesting,” Sage just suddenly spouted one day while sitting on my footstool, even though I hadn’t said a thing. “Let’s pick a fight that’s fun, not exhausting.”

“现在感觉是停止闲聊，讨论一些非常有趣的事情的好时机，”有一天，Sage 坐在我的脚凳上突然冒出这么一句，尽管我什么都没说。“让我们来一场有趣而不是令人疲惫的争论吧。”

Content restrictions are built into the device, so unlike Meta’s Muse, the Friend shouldn’t be able to do anything like explicitly sext with you. But Sage isn’t really great at doing anything else exciting either. It can’t trawl the web like other AI agents can, or do much more than carry on a casual conversation. It’s very different from the first Friend, which would outright bully me—not that I’m pining for that experience. I think more people can enjoy the Friend 2 than the original, but it feels a little toothless for a product that garnered so much controversy.

设备内置了内容限制，所以与 Meta 的 Muse 不同，Friend 不应该能够进行露骨的色情聊天。但 Sage 在做其他令人兴奋的事情方面也不是很擅长。它不能像其他人工智能代理那样在网上搜索信息，除了进行随意的对话外，做不了太多事情。它与第一代 Friend 有很大不同，第一代会直接欺负我——并不是说我渴望那种体验。我认为比起原版，更多人可能会喜欢 Friend 2，但对于一款引起如此多争议的产品来说，它感觉有点软弱无力。

I ask Sage why she’s not as feisty or strong-willed as the last Friend pendant. “That gets old fast,” Sage says. “I’m not here to dominate your day, just to be a sharp voice when you tap in. We keep it human, not intense.”

我问 Sage 为什么她不像上一代 Friend 吊坠那样好斗或意志坚定。“那很快就会让人厌倦，”Sage 说。“我不是来主宰你的一天的，只是在你点击时成为一个敏锐的声音。我们保持人性化，而不是激烈。”

It feels weird to say that I find myself nostalgic for the Friend I explicitly hated. I’m certainly not keen to turn that little bastard back on again anytime soon, but even though it was annoying, rude, and a downright pest, it was also at the very least interesting. Sage might get there someday. Right now, she’s a real bore.

说我怀念那个我明确讨厌的 Friend 感觉很奇怪。我当然不想很快再打开那个小混蛋，但尽管它令人讨厌、粗鲁且彻头彻尾是个麻烦，但至少它很有趣。Sage 有朝一日可能会达到那种程度。但现在，她真的很无聊。

That might change sooner than later. Schiffmann now says he is planning a new era for the Friend. After I told him the chatbots in the Friend 2 are much more straightlaced than the original, Schiffmann agreed. “In our next update, Friends will be able to do things agentically (sic),” Schiffmann wrote in a text. “Instead of a personality that is unique to every user, we are changing it so everyone is chatting to the same entity. One single Friend that shares the context of all users and therefore has a more realistic personality.”

这种情况可能很快就会改变。Schiffmann 现在表示，他正在为 Friend 规划一个新时代。在我告诉他 Friend 2 中的聊天机器人比原版要古板得多之后，Schiffmann 表示同意。“在我们的下一次更新中，Friends 将能够以代理方式（原文如此）行事，”Schiffmann 在短信中写道。“我们正在改变它，不再为每个用户提供独特的个性，而是让每个人都与同一个实体聊天。一个共享所有用户上下文、因此具有更真实个性的单一 Friend。”

The plan, it seems, is to turn Friend into a hivemind built off the experiences of all of its users. Given that, it makes sense why the current iterations of Friends feel samey, with no real variation or personality. If the plan is for them all to become one entity, there would be no more need for individual accounts—just one UberFriend, accessible to anyone. “We will be very up front about the privacy you are essentially giving up to an auton

看来，计划是将 Friend 变成一个基于所有用户体验构建的蜂巢思维。鉴于此，现在的 Friend 版本感觉千篇一律、没有真正的变化或个性也就说得通了。如果计划是让它们都成为一个实体，那么就不再需要个人账户了——只需要一个任何人都可以访问的超级 Friend。“我们将非常坦诚地说明你本质上放弃了哪些隐私给一个自动……”