BuildSpeak每日 builder 文摘
今日归档生词本关于
🎙 播客Training Data· 2026 年 6 月 2 日· 7,102 词 · 约 36 分钟

Knowing What Your Customers Want, All the Time: Listen Labs' Alfred Wahlforss

SPACE 播放 / 暂停·←→ 上一句 / 下一句
Speaker 100:00 - 00:27
Our goal is to get to a billion people in our audience and then to be able to stratify and know what exactly is this person an expert on. And it might be, you know, even something like sneakers. You have some people who are influencers and kind of early adopters. And if you're able to find that audience and interview them first, the insights are much more valuable. And we can learn across all of the interviews that we do.
Speaker 100:00 - 00:27
我们的目标是把受众规模扩展到十亿人,然后能够对他们进行分层,并准确知道这个人究竟擅长什么。甚至可能是像 sneakers 这样的东西。你会有一些人是 influencers,也算是 early adopters。如果你能先找到这类受众并先采访他们,得到的洞察就会更有价值。而且我们也能从我们做过的所有访谈中持续学习。
Speaker 100:27 - 00:34
We build profiles of people as we do more interviews in the platform, and then we can search and find the right person.
Speaker 100:27 - 00:34
随着我们在平台上进行越来越多的访谈,我们会为人们建立 profile(画像),然后就能搜索并找到合适的人。
Speaker 200:51 - 01:19
Okay, today we're sitting down with Alfred Wolfforst, founder and CEO of Listen Labs. Listen is an AI first customer research platform that can run thousands of voice interviews simultaneously. You launched about a year ago, you now serve 20% of the Fortune 500, including iconic brands like Microsoft, Anthropic, Sweetgreen, NBC, and others. And, Constantine, I'm very excited to sit down with you today and talk about market research and how it's getting transformed with AI.
Speaker 200:51 - 01:19
好,今天我们请到了 Listen Labs 的 founder and CEO Alfred Wolfforst。Listen 是一个 AI first 的客户研究平台,能够同时运行数千场语音访谈。你们大约一年前上线,现在已经服务了 20% 的 Fortune 500,其中包括 Microsoft、Anthropic、Sweetgreen、NBC 等知名品牌。Constantine,我今天非常期待和你坐下来聊聊市场调研,以及它正在如何被 AI 改变。
Speaker 101:19 - 01:20
Yeah. Thank you for having me.
Speaker 101:19 - 01:20
好的。感谢邀请我来。
Speaker 201:20 - 01:25
Maybe just to get started. So you are building an AI enabled platform that scales market research. What does that mean?
Speaker 201:20 - 01:25
也许我们先从这里开始。你们正在打造一个由 AI 赋能、能够规模化 market research(市场调研)的平台。这具体是什么意思?
Speaker 101:25 - 01:46
Yeah. So we have this AI agent that can understand your customers better than you can. And the way we do that is by talking to them. So to give you an example, you can ask a question like, how can you improve KRSUS onboarding? And then Listen will create an interview guide, which is an instructions for the agent to make the interviews.
Speaker 101:25 - 01:46
是的。我们有这样一个 AI agent,能比你自己更了解你的客户。而我们做到这一点的方式,就是去和他们交谈。举个例子,你可以问一个问题,比如,怎样改进 KRSUS onboarding?然后 Listen 会创建一份 interview guide,也就是给 agent 用来执行访谈的说明。
Speaker 101:46 - 02:12
And then we have an audience. We have 30,000,000 participants. We can find pretty much anyone from an oncologist to a software engineer, and we'll go and actually talk to them and have hundreds of those interviews and then analyze the data, give you recommendations. And now the final step that we're just launching in a couple months is simulation. So after you've done tens of thousands of interviews in the platform, can you predict how your customers will answer questions in the future?
Speaker 101:46 - 02:12
然后我们还有一个 audience。我们有 30,000,000 名参与者。我们几乎可以找到任何人,从 oncologist 到 software engineer;接着我们会真正去和他们交谈,做数百场这样的访谈,然后分析数据,给你建议。而我们将在未来几个月刚刚推出的最后一步是 simulation。也就是说,当你已经在平台上完成了成千上万次访谈之后,你是否能预测你的客户未来会如何回答问题?
Speaker 102:12 - 02:22
Put it another way, as we get closer to AGI, it will be easier to build things, but the hard part will know what to build. And that's what we're building at Listen.
Speaker 102:12 - 02:22
换句话说,随着我们越来越接近 AGI,构建东西会变得更容易,但真正困难的部分将是知道该构建什么。而这正是我们在 Listen 所做的事情。
Speaker 202:22 - 02:24
Awesome. Do you have any favorite customer stories?
Speaker 202:22 - 02:24
太棒了。你有没有什么最喜欢的客户案例?
Speaker 102:25 - 02:29
Yeah. So Chubbies is one of our customers.
Speaker 102:25 - 02:29
有啊。Chubbies 是我们的客户之一。
Speaker 202:29 - 02:30
Like the short spreads?
Speaker 202:29 - 02:30
就是那个卖短裤的?
Speaker 102:30 - 02:32
Yeah. They've been like one of our early customers.
Speaker 102:30 - 02:32
对。他们算是我们比较早期的客户之一。
Speaker 202:32 - 02:33
What they use you for?
Speaker 202:32 - 02:33
他们用你们做什么?
Speaker 102:33 - 03:06
They use us for everything. So a lot of marketing testing for testing shirts to understand what products perform well and what doesn't. And one of my favorite examples is they discovered that chest hair interface really poorly with one of the materials they have. So it's like really uncomfortable to wear one of their shirts and they changed the shirt and it became like radically more comfort comfortable. So we saw, you know, the small things to the big things, Manscaped, changed their Super Bowl ad with insights from from Lissanne.
Speaker 102:33 - 03:06
他们什么都用我们来做。所以有很多 marketing(市场营销)测试,比如测试衬衫,以了解哪些产品表现好、哪些不好。然后我最喜欢的一个例子是,他们发现胸毛会和他们某一种面料产生很差的 interaction(交互影响)。所以穿他们其中一款衬衫会特别不舒服,后来他们改了这款衬衫,舒适度就有了非常大的提升。所以我们看到的既有小事也有大事,Manscaped 就根据来自 Lissanne 的洞察,修改了他们的 Super Bowl 广告。
Speaker 103:06 - 03:06
So
Speaker 103:06 - 03:06
所以
Speaker 203:07 - 03:09
Never heard of that, but I'm not gonna ask.
Speaker 203:07 - 03:09
这个我倒是没听过,不过我就不追问了。
Speaker 303:10 - 03:13
Huge. So you you got the men's hair market covered.
Speaker 303:10 - 03:13
厉害。所以你们把男性毛发护理市场都覆盖了。
Speaker 103:13 - 03:14
Yes. That's our niche.
Speaker 103:13 - 03:14
是的。那就是我们的细分领域。
Speaker 303:14 - 03:18
From shaping to clothing. That's right. Wow.
Speaker 303:14 - 03:18
从塑形到服饰。没错。哇。
Speaker 103:18 - 03:20
We do other things. Skimps is one of customers this morning.
Speaker 103:18 - 03:20
我们也做别的事情。Skimps 是我们今天早上的客户之一。
Speaker 303:20 - 03:21
You heard it on Trader.
Speaker 303:20 - 03:21
你是在 Trader 上听到的。
Speaker 203:22 - 03:51
Don't know what you're talking about, but I know context clues. That's awesome. I'd love to understand, as you framed it, as we get closer to this AGI future, one of the questions I have is, traditionally, I've always been very skeptical, actually, of surveys because people get paid to take surveys. So you already got a selection bias issue. The things that people say they would do or the way that they describe how they would behave is different from how they actually behave in practice.
Speaker 203:22 - 03:51
我不知道你在说什么,但我能通过上下文线索理解。那太棒了。正如你刚才所说,随着我们越来越接近这个 AGI 未来,我很想理解的一点是:我一直以来其实都对 survey(问卷调查)非常怀疑,因为人们是拿钱去做 survey 的。所以这里天然就有 selection bias(选择偏差)的问题。人们说自己会怎么做,或者他们描述自己会如何表现的方式,往往和他们在实际中真正的行为并不一样。
Speaker 203:51 - 04:05
And so I guess I come from the school of thoughts where actual just telemetry in the real world matters so much more than asking people about what they would do. And so I'm curious what you think of that and how you think AI or Listen Labs can help bridge that gap.
Speaker 203:51 - 04:05
所以我想,我属于这样一种观点流派:真实世界里的实际 telemetry(行为遥测数据)比起去问人们“他们会怎么做”要重要得多。所以我很好奇你怎么看这一点,以及你认为 AI 或 Listen Labs 能如何帮助弥合这个差距。
Speaker 104:05 - 04:41
Yeah. And so we've done a lot of research on this. One of the things we've done with surveys, for example, is we went back to the same person and asked them a multiple choice survey again, and they were radically inconsistent. So even if you go back to the same person and ask them a survey question in a multiple choice fashion, they're much more inconsistent. But we did the same thing with listen when you actually have to think and you have to really reason through your answer and then you're much more consistent with at least how you answer the same question.
Speaker 104:05 - 04:41
是的。我们在这方面做了很多研究。比如说,关于 survey,我们做过的一件事是:我们回头去找同一个人,再次让他做一份 multiple choice(多项选择)survey,结果他们前后非常不一致。所以即使你回到同一个人那里,用选择题的方式问同一个 survey 问题,他们的回答也会更加不一致。但我们也用 listen 做了同样的事:当你真的需要思考、需要认真推理自己的答案时,人们至少在回答同一个问题时会一致得多。
Speaker 104:41 - 05:13
And then we're constantly tracking, for example, with Chubbies, when we test their different charts, we couple of months later look back and see how did that perform with the actual sales data. And I think it depends on the different use cases. I agree that AB test is kind of the holy grail, but in practice it becomes really difficult to get right because you need a very large volume of users. And it's it's really useful to have some kind of input than no input at all.
Speaker 104:41 - 05:13
另外,我们也在持续跟踪。比如和 Chubbies 合作时,当我们测试他们不同的 charts(图表/方案)后,几个月之后我们会回头看:它在实际 sales data(销售数据)中的表现到底如何。我认为这取决于不同的 use case(使用场景)。我同意 AB test(A/B 测试)某种程度上算是 holy grail(终极标准),但在实践中要把它做好真的非常困难,因为你需要非常大的用户量。而且,与其完全没有输入,拥有某种形式的输入终究还是很有价值的。
Speaker 305:13 - 05:31
Does Listen do voice to text, as in the actual customer who is answering the survey, can speak their answer and then you guys transcribe it? Does it also do text to voice? It's a two way conversation. What does listen start with, and what does it finish with for the user experience?
Speaker 305:13 - 05:31
Listen 会做 voice to text 吗?也就是说,实际回答问卷的客户可以直接说出答案,然后由你们来转写?它也会做 text to voice 吗?这是一个双向对话。对于用户体验来说,Listen 一开始是从什么开始,最后又是以什么结束的?
Speaker 105:31 - 05:54
Yeah. So it's essentially a Zoom call that you have with the agent. So you're on video, and you can also detect their emotions. So that's another way to bridge the gap between what they say and how they actually think and feel. So it looks at your eyes the way you say it, and that's kind of much closer to how you actually behave in the real world.
Speaker 105:31 - 05:54
会的。所以本质上,这就像你和 agent 进行的一次 Zoom 通话。你会出现在视频里,同时系统也可以检测他们的情绪。所以,这是弥合“他们说了什么”和“他们实际上怎么想、怎么感受”之间差距的另一种方式。它会看你的眼神、你表达时的方式,而这就更接近你在真实世界中的实际行为。
Speaker 305:54 - 06:12
And have you seen, per Sonya's point, that actually having the person's face and their emotions and their voice and whatnot yields more engagement, truthfulness? Have have we been able to have any studies or or at least data to point in that direction?
Speaker 305:54 - 06:12
那你们有没有发现,像 Sonya 提到的那样,把一个人的脸、情绪、声音等等都纳入进来,实际上会带来更高的参与度、真实性?我们有没有做过任何研究,或者至少有没有一些数据能指向这个方向?
Speaker 106:12 - 06:48
Yeah. Specifically with advertising, it's a huge benefit because you might have people say on a Likert scale, which is like five questions that you click, are you extremely likely to buy this product versus when you might have very high scores on a survey question like that, but when someone also reacts very enthusiastically, going to be like perform much higher. And we've seen that those ads then perform better in performance marketing, for example, on on Meta and and LinkedIn.
Speaker 106:12 - 06:48
有的。特别是在 advertising 方面,这个优势非常大,因为你可能会让人们在 Likert scale(李克特量表)上作答——比如五个可点击的问题,像“你是否非常可能购买这个产品”——但相比之下,即使某人在这类问卷题上的分数很高,如果他同时表现出非常强烈、非常热情的反应,那通常意味着实际表现会高得多。我们也看到,这类广告随后在 performance marketing 中的表现确实更好,比如在 Meta 和 LinkedIn 上。
Speaker 306:48 - 06:59
And can you if you're the customer and you commission this and you get all this response, can you actually click in and if you ever wanted to watch the interview to get that level of granularity? Yeah. So we built
Speaker 306:48 - 06:59
如果你是客户,你委托做了这件事,也拿到了所有这些反馈,那么你是否真的可以点进去,在你想要那种细粒度信息的时候,直接观看访谈?可以。所以我们构建了
Speaker 106:59 - 07:13
the platform around traceability so that for every data point, you can always click and then look at the video or see the quote. So you know that AI is not just hallucinating kind of where it's coming from. That's awesome.
Speaker 106:59 - 07:13
这个平台时就是围绕 traceability(可追溯性)来设计的,因此对于每一个 data point(数据点),你都可以随时点进去查看视频,或者看到对应的引述。这样你就知道 AI 不是在幻觉式地编造它的来源。太棒了。
Speaker 307:13 - 07:14
Makes sense.
Speaker 307:13 - 07:14
明白了。
Speaker 207:14 - 07:16
How'd you come up with the idea to build this?
Speaker 207:14 - 07:16
你当初是怎么想到要做这个的?
Speaker 107:17 - 07:37
So my co founder and I actually built a consumer app and That did what? That went viral. It was called a Bfake. So create an AI avatar of yourself. It was an early version of the chatty bitty images, and you could fine tune stable diffusion and put yourself in that world.
Speaker 107:17 - 07:37
所以我和我的 co-founder(联合创始人)当时实际上做了一款面向消费者的 app,它后来怎么样了?它火了,直接 viral(爆红)。它叫 a Bfake。你可以创建一个自己的 AI avatar(AI 虚拟形象)。那算是 chatty bitty images 的一个早期版本,你还可以 fine-tune stable diffusion,然后把自己放进那个世界里。
Speaker 107:37 - 07:57
And that ended up going super viral. And overnight we had 20,000 users. And we were also kind of experimenting with different ways of using AI. So we built this AI interview for ourselves because we had a bunch of questions of how we had we had a ton of churn. So we wanted to understand why how they thought about our positioning, different use cases.
Speaker 107:37 - 07:57
后来它就传播得非常快,一夜之间我们有了 20,000 名用户。与此同时,我们也在试验各种不同的 AI 用法。所以我们给自己做了一个 AI interview(AI 访谈),因为我们当时有很多问题想搞清楚——我们的 churn(用户流失)很高。所以我们想理解用户为什么流失、他们怎么看我们的 positioning(定位),以及不同的 use case(使用场景)。
Speaker 107:57 - 08:00
And it was really useful for us. And that's how we got started.
Speaker 107:57 - 08:00
结果它对我们非常有用,我们也就是这样开始的。
Speaker 208:01 - 08:16
Maybe just walk us through how the industry is changing before and after Listen Labs. Historically, let's say you're somebody with an app with 20,000 users. You understand how users are using the app, what they want next, why they're churning. Historically, how do people go about doing that?
Speaker 208:01 - 08:16
也许你可以带我们梳理一下,在 Listen Labs 出现前后,这个行业是怎么变化的。比如说,过去如果你有一个拥有 20,000 用户的 app,你想了解用户是怎么使用这个 app 的、他们接下来想要什么、他们为什么流失。从历史上看,人们通常是怎么做这件事的?
Speaker 108:16 - 08:50
Yeah. So what we discovered was that there are these survey tools that are pretty old school, like Qualtrics, But then there's also this very large services industry because it becomes harder and harder, especially if you wanna do market research where you wanna talk to your prospective customers, not your current customers. It becomes harder and harder to do that as you scale. So that's a multi billion dollar industry. And it's what they do is come up with questions to ask, which is an academic subject in of itself.
Speaker 108:16 - 08:50
是的。我们发现,一方面有一些 survey tools(问卷工具)已经非常 old school(传统)了,比如 Qualtrics;但另一方面,也存在一个非常庞大的服务行业,因为这件事会变得越来越难,尤其是如果你做的是 market research(市场调研),你想交谈的是 prospective customers(潜在客户),而不是现有客户。随着规模扩大,这件事会越来越难。所以这是一个数十亿美元级别的行业。他们所做的事情之一,就是设计要问的问题,而这本身就是一门很学术的课题。
Speaker 108:50 - 09:26
It's actually really hard to know, like, how do you ask questions to your point that get to how someone actually will behave? You can't just ask, like, how much are you willing to pay for this? There's different methodologies that work better than others. To finding the audience, how do you source the participants to then like analyze hundreds of these calls. And in the traditional industries like CPG, even in Microsoft, they spend tens of millions of dollars on focus groups to bring people in a room and interview them, and we can help speed that up much faster.
Speaker 108:50 - 09:26
实际上这真的很难。比如说,像你提到的那样,怎样提问,才能真正触及一个人实际会如何行动?你不能只是直接问一句:你愿意为这个付多少钱?有些 methodology(方法论)就是比另一些更有效。再到 audience(受众)筛选,你要如何找到这些参与者;然后还要去分析数百通这样的通话。在传统行业里,比如 CPG,甚至在 Microsoft,他们会花几千万美元做 focus groups(焦点小组),把人召集到一个房间里进行访谈,而我们可以帮助把这个过程大幅加速。
Speaker 209:26 - 09:42
Okay. So that's the old world of how this used to be done. Maybe describe the new world. And then it seems to me that there were obvious kind of first order benefits. Like it's probably much more scalable, probably much more cost effective, but there's probably also less obvious benefits.
Speaker 209:26 - 09:42
好的。所以这就是过去这个领域的老做法。那也许你可以讲讲新世界是什么样的。在我看来,第一层的好处似乎很明显:它大概更容易 scale(扩展),也大概更有 cost-effectiveness(成本效益);但可能也还有一些没那么显而易见的好处。
Speaker 209:43 - 09:50
Maybe just talk about some of those benefits of what is it like when you actually do AI first market or customer research?
Speaker 209:43 - 09:50
不如谈谈其中一些好处:当你真正以 AI first(AI 优先)的方式来做 market research(市场调研)或 customer research(客户研究)时,体验到底是什么样的?
Speaker 109:50 - 10:11
Yeah. So most decisions that gets made are not based on the customer input. Right? And the reason for that is it's just a lot of friction to even talk to customers. So when you can lower the barriers of talking to customers, you end up making much smarter decisions.
Speaker 109:50 - 10:11
是的。所以,大多数做出的决策其实都不是基于客户输入(customer input)。对吧?原因在于,哪怕只是和客户聊一聊,这件事本身的阻力都很大。所以,当你能降低与客户交谈的门槛时,你最终就会做出聪明得多的决策。
Speaker 110:11 - 10:36
So the speed advantage is actually huge. For us, you can get input within five minutes from real people, and it's a really magical experience when you see hundreds of people populate in your, like, interview. And so that's one thing. And because it's asynchronous, it's also much more affordable. So you can pay people much less than if you would had to run, like, synchronous interviews.
Speaker 110:11 - 10:36
所以,速度优势实际上非常大。对我们来说,你可以在五分钟内就从真实的人那里获得输入;而当你看到几百个人涌入你的、类似访谈(interview)的流程时,那真的是一种很神奇的体验。这是一方面。而且因为它是 asynchronous(异步)的,所以成本也低得多。因此,相比必须进行那种 synchronous(同步)访谈的情况,你可以用低得多的报酬来支付受访者。
Speaker 110:37 - 10:53
So actually that's an interesting thing that people often ask us like, do people even like being interviewed by an AI? And the objective answer is yes, because you can pay them less to talk to an AI than to talk to a actual interviewer.
Speaker 110:37 - 10:53
所以,其实有个很有意思的问题,人们经常会问我们:人们真的喜欢被 AI 访谈吗?而客观答案是喜欢,因为与其和真正的访谈员交谈,你可以用更低的报酬让他们和 AI 交谈。
Speaker 210:53 - 10:54
Why is that?
Speaker 210:53 - 10:54
为什么会这样?
Speaker 110:55 - 11:12
I think it's mostly because it's asynchronous and people are very busy, but then also Lower Yeah. Lower pressure. You can kinda go on and off. And we've also found that people are more honest talking to an AI. We've had people really open up.
Speaker 110:55 - 11:12
我觉得主要是因为它是 asynchronous(异步)的,而人们都很忙;但另外也是因为压力更小。对,压力更低。你可以断断续续地进行。我们还发现,人们在和 AI 交谈时会更诚实。我们遇到过一些人真的会敞开心扉。
Speaker 111:12 - 11:32
It's a very therapeutic experience because it's a nonjudgmental entity that's really interested in you. And we can also have sensitive like interviewing kids, how they react to different products. And so I think that's another advantage as well that people can be brutally honest talking to the AI.
Speaker 111:12 - 11:32
这是一种很有治疗感的体验,因为对方是一个不带评判的实体,而且真的对你感兴趣。我们还可以处理一些比较敏感的场景,比如访谈孩子,了解他们对不同产品的反应。所以我觉得这也是另一个优势:人们在和 AI 交谈时可以非常坦率。
Speaker 211:32 - 11:39
Okay. So historically, for example, I was to research on the kids market, very, very hard to access that market. Is that a regulatory thing? Is it a scheduling thing?
Speaker 211:32 - 11:39
好的。比如说,过去如果我要研究儿童市场,要接触这个市场会非常、非常难。这是监管方面的问题吗?还是排期安排的问题?
Speaker 111:40 - 11:48
Yeah. You need parental consent. Know, kids are really busy. They go to school. They have extracurricular activities.
Speaker 111:40 - 11:48
对。你需要父母同意。你知道的,孩子们也非常忙。他们要上学,还要参加课外活动。
Speaker 111:48 - 12:17
How do you find time with them? And you need to find the right kind of kids. Like one of the things we realized is the audience is extremely important, and that's actually where we spend 80% of our engineering resources. Every company is driven by a power law in customer segmentation. So even a product like Sweetgreen, which you would think is for everyone, the right audience is typically urban, high household income, mostly female.
Speaker 111:48 - 12:17
你怎么找到时间和他们接触?而且你得找到合适类型的孩子。比如我们意识到的一件事是,受众极其重要,而实际上这正是我们投入 80% engineering(工程)资源的地方。每家公司在 customer segmentation(客户细分)上都受 power law(幂律)驱动。所以即便是像 Sweetgreen 这样的产品,你可能会觉得它是面向所有人的,但它真正合适的受众通常是城市居民、家庭收入高、并且大多是女性。
Speaker 112:17 - 12:35
And by the way, they need to know what seed oils are, which only like 1% of the population does. And then you find that some people go to Sweetgreen every single day and that's 80% of their revenue. So if you can find that segment, the research is so much more actionable.
Speaker 112:17 - 12:35
顺便说一句,他们还得知道 seed oils(种子油)是什么,而整个人口里大概只有 1% 的人知道。然后你会发现,有些人每天都会去 Sweetgreen,而这部分人贡献了他们 80% 的收入。所以如果你能找到那个细分人群,研究结果就会变得更可执行。
Speaker 312:35 - 12:56
Yeah. There's probably a network effect to it as well where when you get of a certain scale and people use it, you can access the same kind of person that otherwise might be really difficult to access, or maybe it's a scale economy, something along the lines of accessing those really, really specific people that are really valuable for the type of product that you're trying to introduce.
Speaker 312:35 - 12:56
对。而且这里可能也存在 network effect(网络效应):当你达到一定规模、并且人们在使用它时,你就能接触到那类原本非常难以接触的人;或者也许这算是一种 scale economy(规模经济)。总之,大意就是你能够接触到那些非常非常特定、而且对你要推出的产品类型来说极其有价值的人。
Speaker 112:56 - 13:24
Yeah. It's really about you know, we our goal is to get to a billion people in our audience and then to be able to stratify and know what exactly is this person an expert on. And it might be, you know, even something like sneakers. You have some people who are influencers and kind of early adopters. And if you're able to find that audience and interview them first, the insights are much more valuable.
Speaker 112:56 - 13:24
对。核心其实是,我们的目标是让自己的受众规模达到 10 亿人,然后能够对他们进行分层,并准确知道这个人究竟在哪方面是 expert(专家)。这甚至可能是像 sneakers(球鞋)这样的领域。你会有一些人是 influencers(影响者),属于 early adopters(早期采用者)。如果你能够先找到这类受众并采访他们,得到的洞察会有价值得多。
Speaker 113:25 - 13:36
We can learn across all of the interviews that we do. So we build profiles of people as we do more interviews in the platform, and then we can search and find the right person.
Speaker 113:25 - 13:36
我们可以从自己做过的所有访谈中持续学习。所以,随着在平台上进行越来越多的访谈,我们会建立人物画像,然后就能搜索并找到合适的人。
Speaker 313:36 - 13:49
So someone might say in a totally unrelated interview, I'm a total sneakerhead, and you can keep that in the database on that person. Then when Nike or what have you is launching their new product line, you can offer that person up. That's right. That's amazing.
Speaker 313:36 - 13:49
所以,有人可能会在一次完全无关的访谈里说:“我是个十足的 sneakerhead(球鞋迷)”,你就可以把这个信息保存在数据库里、挂在这个人名下。这样等到 Nike 或类似公司要发布他们的新产品线时,你就可以把这个人推荐出来。没错。太厉害了。
Speaker 113:49 - 13:58
And that was not possible to to do before because it was usually, like, separate entities, and it would be a very manual process where you would have an email list and you
Speaker 113:49 - 13:58
而这在以前是不可能做到的,因为过去通常都是彼此分开的实体,而且流程会非常手工化:你会有一个 email list(邮件列表),然后你
Speaker 313:58 - 14:00
would just spam email. I've been on the receiving end
Speaker 313:58 - 14:00
就只能群发 spam email(垃圾邮件)。我自己就曾是这种邮件的接收者。
Speaker 114:00 - 14:29
of Yeah. They're terrible. And one of the problems with that is that you then need to have a extensive screening process. So you have something called an incidence rate, which can be ten percent, which means only one in ten people gets qualified to even take the interview. And that causes significant churn on these databases because it's really annoying be greened out 10 times to even get paid the first time.
Speaker 114:00 - 14:29
对,是的。它们很糟糕。而其中一个问题是,你随后就需要一个非常繁琐的筛选流程。所以会有一个叫 incidence rate 的指标,可能只有 10%,这意味着 10 个人里只有 1 个人连参加访谈的资格都能拿到。这会导致这些数据库出现显著 churn(流失),因为一个人要先被筛掉 10 次,才能第一次拿到报酬,这真的很让人恼火。
Speaker 214:29 - 14:39
Why do these brands even need you for access if, let's take Sweetgreen. Sweetgreen knows who that eighty percent is. Can't they just reach them? Don't they have direct relationships with them already?
Speaker 214:29 - 14:39
那这些品牌为什么还需要通过你们来获取触达呢?比如拿 Sweetgreen 来说。Sweetgreen 知道那 80% 的人是谁。难道他们不能直接联系这些人吗?他们不是本来就已经和这些人有直接关系了吗?
Speaker 114:40 - 15:00
Yeah. So they can, and we do that as well. We connect to their CRM and they can send that out. But then the really interesting part is how do you talk to prospective customers? People who also may not be kind of current power law users and how do you compare those two?
Speaker 114:40 - 15:00
对,他们可以,我们也会这样做。我们会连接到他们的 CRM(客户关系管理系统),他们可以通过 CRM 把这些发出去。但真正有意思的部分是,你要怎么和 prospective customers(潜在客户)沟通?以及那些可能并不是当前高频核心用户的人,你要怎么触达?还有,你又该如何比较这两类人?
Speaker 115:00 - 15:17
And then also what we found is that the CRM is typically really unorganized. And sometimes there's also like regulatory issues. If you're at Google, you can't just send emails to people who use Gmail and it gets much easier to use an external third party.
Speaker 115:00 - 15:17
另外我们还发现,CRM 通常都非常混乱。有时候也会有 regulatory issues(监管问题)。如果你在 Google,你不能直接给 Gmail 用户群发邮件,而使用外部第三方就会容易得多。
Speaker 315:17 - 15:33
And you run the risk of spam, which can get you totally blocked. I have seen that at some of our companies over the years where, you know, you do outbound and then eventually you're in the Google filter, and next thing you know, you're in Microsoft purgatory. I guess, going through you guys, you don't have to deal with that.
Speaker 315:17 - 15:33
而且你还要承担被当成 spam(垃圾信息)的风险,这可能会让你被彻底封锁。这些年我在我们的一些公司里见过这种情况:你做 outbound(外呼/主动触达),然后慢慢地你就进了 Google 的过滤器,接下来不知不觉又掉进 Microsoft 的 purgatory(收件黑洞)里。我想,如果通过你们来做,就不用处理这些问题了。
Speaker 115:33 - 15:34
Yeah. Exactly.
Speaker 115:33 - 15:34
对,完全是这样。
Speaker 315:34 - 15:35
It's cool.
Speaker 315:34 - 15:35
挺酷的。
Speaker 215:35 - 15:45
What does this mean for the MacKinzeys or whoever is in the world that are you know, the people that are building the 100 slide decks that, you know, reach 3,000 people to reach some set of insights? Did you
Speaker 215:35 - 15:45
这对世界上的那些 MacKinzeys,或者不管叫什么的那类人,意味着什么?就是那些会做 100 页 slide decks(幻灯片报告)的人,他们为了得到某些洞察,要触达 3,000 个人。你有……
Speaker 315:45 - 15:46
do that, Sonia? Wasn't that a former life?
Speaker 315:45 - 15:46
做那个,Sonia?那不是前世的事了吗?
Speaker 215:47 - 15:50
You know, no, Constantine, but I'm glad that that's what you think of me.
Speaker 215:47 - 15:50
你知道,不是的,Constantine,不过我很高兴你是这么看我的。
Speaker 315:51 - 15:52
Isn't that what banking does?
Speaker 315:51 - 15:52
那不就是 banking 在做的事吗?
Speaker 215:54 - 15:58
We hired consultants. We used to hire these people.
Speaker 215:54 - 15:58
我们雇了 consultants。我们以前常常雇这帮人。
Speaker 315:58 - 15:59
Got it. Got it.
Speaker 315:58 - 15:59
明白了。明白了。
Speaker 216:00 - 16:02
So I was, you know, was a layer on top of the layer on top
Speaker 216:00 - 16:02
所以我当时,那个,算是盖在一层之上的又一层之上的一层
Speaker 316:02 - 16:02
Okay. Of
Speaker 316:02 - 16:02
好的。的
Speaker 216:03 - 16:11
Got it. Okay. I was even more redundant. But what does it mean for all these people? Do they still have a role to play in this new future?
Speaker 216:03 - 16:11
明白了。好的。我就更显得多余了。但这对所有这些人意味着什么?他们在这个新的未来里还会有可发挥的角色吗?
Speaker 116:11 - 16:27
Yeah. I think AI is changing all roles very quickly. And we work a lot with Bain, for example. So they use us to speed up their traditional processes. And I think they still have a role to play.
Speaker 116:11 - 16:27
对。我觉得 AI 正在非常迅速地改变所有岗位。比如说,我们和 Bain 有很多合作。所以他们会用我们的产品来加速他们的传统流程。我觉得他们仍然有自己的角色可以发挥。
Speaker 116:27 - 16:52
I think traditional services and being able to then implement these changes is still extremely valuable, but a lot of margins are going to drop and you have to make sure you kind of unbundle a lot of your services to maybe allow for AI agents to to help solve some of the problems that you would go to traditional consulting firms before?
Speaker 116:27 - 16:52
我认为,传统服务以及随后能够去实施这些变化,仍然是极其有价值的,但很多利润率都会下降,你必须确保把自己的很多服务做某种程度的 unbundle(拆分),也许这样才能让 AI agents(AI 代理)帮助解决一些过去你会去找传统咨询公司处理的问题?
Speaker 316:52 - 17:15
Maybe I'm an optimist here, but why wouldn't it be more? Why wouldn't I, if I'm running a business, say, oh, great. I wanna find five new areas to expand to now that I have AI and these tools, and I will pay you, Bain, or what have you, the same dollars you use LISN, and just explore those new areas and tell me where to where to build? Is that overly optimistic?
Speaker 316:52 - 17:15
也许我是个乐观派,但为什么不会更多呢?如果我是一个企业经营者,为什么我不会说,哦,太好了。既然我现在有了 AI 和这些工具,我想再找五个新的扩张领域,我会付给你,Bain,或者类似的机构,同样的钱,就像你使用 LISN 一样,然后去探索那些新领域,告诉我该去哪里建设?这是不是过于乐观了?
Speaker 117:15 - 17:26
No. I I think it's one of those areas where the ceiling is very high. You can kind of learn and more about your you can build more things. And so I think I think you're right.
Speaker 117:15 - 17:26
不。我觉得这是那种上限非常高的领域之一。你可以不断学习更多关于你自己的东西,也可以构建更多东西。所以我觉得你说得对。
Speaker 317:26 - 17:31
Yeah. I was I was thinking the chest hair shirt thing. I'll probably I'll
Speaker 317:26 - 17:31
对。我刚刚还在想 chest hair shirt 那件事。我可能会——我会——
Speaker 217:31 - 17:34
probably I love that twenty minutes and you're still thinking about chest hair. That's insane.
Speaker 217:31 - 17:34
可能会。我太喜欢这一段了,二十分钟过去了,你还在想着胸毛。这太离谱了。
Speaker 317:34 - 17:45
There's so many little things that I'd love to tell the companies that I am a consumer of. Like, the smallest little even the way they laced these shoes, I'd love to give that feedback.
Speaker 317:34 - 17:45
有太多很小的细节,是我很想告诉那些我作为消费者会使用的公司的。哪怕最微小的地方,甚至连他们给这双鞋系鞋带的方式,我都很想给出那样的反馈。
Speaker 217:45 - 17:47
Yeah. This is why you're a venture capitalist.
Speaker 217:45 - 17:47
对。这就是为什么你是个 venture capitalist。
Speaker 317:47 - 17:49
Details. Details.
Speaker 317:47 - 17:49
细节。细节。
Speaker 117:51 - 17:55
We hope to live in a world that finally works the way people want. That'd be great. Please.
Speaker 117:51 - 17:55
我们希望生活在一个终于能按人们期望那样运转的世界里。那会很棒。拜托了。
Speaker 217:55 - 18:12
Are you seeing any pricing compression already hit the industry? Like, I would imagine if I am Bain's customer, I'm thinking, well, you're able to do this survey a lot more efficiently now with AI than before AI. Who's getting the benefit of that economic surplus?
Speaker 217:55 - 18:12
你觉得定价压力、也就是价格压缩,是否已经开始冲击这个行业了?比如,我会想象如果我是 Bain 的客户,我会觉得,嗯,你现在用 AI 做这类调研,比 AI 出现之前高效多了。那这部分经济剩余带来的好处,最终是谁在拿走?
Speaker 118:12 - 18:18
So because you're able to do it faster, I would argue you should be able to charge more for it.
Speaker 118:12 - 18:18
所以正因为你能做得更快,我反而会认为你应该为此收更高的价格。
Speaker 218:18 - 18:19
And is that what's actually playing out?
Speaker 218:18 - 18:19
那实际情况也是这样发展的吗?
Speaker 118:19 - 18:54
We have done some studies where we're able to charge hundreds of thousands of dollars to speak to 20 doctors across eight countries. So maybe over the long term, like the individual interview will become more affordable. But I think you'll be doing kind of two orders of magnitude more of research. And I think what's really exciting is also simulation, which is something we're building now where you're able to unlock the 99% of use cases where you would never have time to talk to real people?
Speaker 118:19 - 18:54
我们做过一些研究项目,比如我们能够就“与八个国家的 20 位医生交流”这件事收取几十万美元。所以也许从长期来看,单次 individual interview(单独访谈)会变得更便宜。但我认为,你做的 research(研究)规模会扩大大约两个数量级。还有一件我觉得特别令人兴奋的事是 simulation(模拟),这也是我们现在正在构建的东西——借此你就能解锁那 99% 的 use case(用例):在这些场景里,你原本根本不可能有时间去和真人交谈。
Speaker 318:54 - 19:40
I think that's so awesome in part because there are so many areas where they don't even listen to the customer. Like medicine, there are a million little problems with the medical system I hear about all the time. And these are You know, their doctors are busy, important people, but it feels like the companies haven't even invested the time in figuring out where all those paper cuts are. And the doctors are really busy, so they're not gonna go schedule an appointment and have some long conversation and meet with some group. But if they could do it at any time, like in an app on their phone as part of the normal homepage app, and give feedback on their EHR or something in the operating room or something along those lines, that seems like a life saving use case for listen over time.
Speaker 318:54 - 19:40
我觉得这特别棒,部分原因是有太多领域他们甚至都不听 customer(客户/用户)的声音。比如医疗,我一直不断听到医疗系统里有无数细小的问题。你知道,医生都是很忙、很重要的人,但感觉这些公司甚至都没有投入时间去弄清楚那些让人不断受小伤似的痛点到底在哪里。而医生又真的很忙,他们不会专门去约个时间,进行一场很长的谈话,再去和某个团队见面。但如果他们可以在任何时候完成这件事,比如就在手机上的 app 里,作为平时常用首页 app 的一部分,顺手就能对他们的 EHR 提反馈,或者在手术室里也能做类似的事,那听起来就像是一个随着时间推移足以挽救生命的“持续倾听” use case(用例)。
Speaker 119:40 - 20:04
Yeah. I think what I'm really excited about as well is taking all those small things and then telling another agent to go and solve that problem. And we're getting pulled in in this direction by some of our customers where they will have a churn interview and then they will connect them. If you find a bug, for example, they'll connect that to another coding agent to go and solve the problem.
Speaker 119:40 - 20:04
对。我同样非常兴奋的一点是,把所有那些小问题收集起来,然后告诉另一个 agent(智能体)去把那个问题解决掉。我们的某些客户正在把我们往这个方向拉:他们会先做一场 churn interview(流失访谈),然后把结果接起来。比如说,如果你发现了一个 bug,他们就会把它连接到另一个 coding agent(编码智能体),让它去把问题解决掉。
Speaker 320:04 - 20:05
That's cool.
Speaker 320:04 - 20:05
这很酷。
Speaker 220:05 - 20:36
Let's talk about generative agent simulation. Like it seems like the entire industry has gone from market research one point zero, call 100 people one by one, collate them manually, market research two point zero, AI native, where AI designs the question track, is able to talk to thousands of people simultaneously, synthesize the answers. It seems like we're maybe moving to market research three point zero with generative agent simulation. What do you make of that? And I both see the dream of it.
Speaker 220:05 - 20:36
我们来聊聊 generative agent simulation(生成式 agent 模拟)。感觉整个行业好像已经从 market research 1.0——一个个打电话给 100 个人、再手动整理——发展到了 market research 2.0,也就是 AI native(AI 原生)模式:由 AI 设计提问路径,能够同时和成千上万人对话,并综合他们的回答。现在看起来,我们也许正在迈向借助 generative agent simulation 的 market research 3.0。你怎么看这件事?我自己也能看到其中的那种愿景。
Speaker 220:37 - 20:54
I see how synthetic data has changed, for example, self driving cars. And then I also am inherently skeptical of it. Like, is a bunch of synthetic data just remixing what's already in the pre training sets? And are you actually learning anything useful or with alpha there? And so I'd love to hear your take on it and how you guys are taking on the three point zero.
Speaker 220:37 - 20:54
我看到 synthetic data(合成数据)已经如何改变了,比如说,self-driving cars(自动驾驶)。但与此同时,我本能上也对它持怀疑态度。比如,一大堆 synthetic data 会不会只是把 pre-training sets(预训练数据集)里已有的内容重新 remix(重组)一遍?你真的能从中学到任何有用的东西,或者获得 alpha(超额收益/优势)吗?所以我很想听听你的看法,以及你们是怎么切入这个 3.0 的。
Speaker 320:54 - 20:56
And maybe what is it too,
Speaker 320:54 - 20:56
或者也可以先说说,它到底是什么,
Speaker 120:56 - 21:10
to start? Yeah. So the way we are building simulation is by interviewing a single person. Say if I interview Constantine for one hour, I can probably start to predict your preferences to something.
Speaker 120:56 - 21:10
作为开头?对。所以我们构建 simulation(模拟)的方式,是先去采访一个单独的人。比如说,如果我采访 Constantine 一个小时,我大概就可以开始在某种程度上预测你的偏好。
Speaker 321:10 - 21:12
Fascinating insights about chest hair.
Speaker 321:10 - 21:12
关于胸毛的精彩洞见。
Speaker 121:13 - 21:43
And it turns out that LMs are quite good at this as well. So you can essentially try to feed in as much information as possible on a single individual. Then in some cases, we're able to get 95% accuracy to predict how they will answer certain questions. Now the problem becomes things are changing all the time and chaos theory tells us it's really hard to predict the future. Otherwise, we would be on on Wall Street and and making a ton of money.
Speaker 121:13 - 21:43
而事实证明,LMs(语言模型)在这件事上也相当擅长。所以,本质上你可以尝试尽可能多地输入某一个个体的信息。然后在某些情况下,我们能够以 95% 的准确率预测他们会如何回答某些问题。现在问题在于,一切都在不断变化,而 chaos theory(混沌理论)告诉我们,预测未来是非常困难的。否则的话,我们就会去 Wall Street,然后赚很多很多钱了。
Speaker 121:44 - 22:02
So what what we how we think about it is you need to hydrate these audiences. And the way we do that is by all of the interviews that are running through Listen. So we have a very strong network effect. We've done a million interviews so far, and that's grown exponentially since we reported that number.
Speaker 121:44 - 22:02
所以,我们的看法是,你需要去 hydrate 这些受众群体。而我们实现这一点的方式,就是利用所有通过 Listen 进行的访谈。所以我们有非常强的 network effect(网络效应)。到目前为止,我们已经做了 100 万次访谈,而且自从我们报告那个数字以来,这个数量一直在指数级增长。
Speaker 322:02 - 22:02
Wow.
Speaker 322:02 - 22:02
哇。
Speaker 122:03 - 22:22
And we're able to train audiences on on those interviews. So you can imagine a future where you can ask a question and listen, like how do software engineers think about cloud code? Interesting. And then listen, we'll say, well, I already talked to a thousand software engineers this week. Let me predict how they're gonna answer that question.
Speaker 122:03 - 22:22
而且我们能够用这些访谈来训练受众模型。所以你可以想象这样一个未来:你提出一个问题,然后去“听”他们会怎么回答,比如,软件工程师是如何看待 cloud code 的?很有意思。接着系统会说,好吧,我这周已经和一千位软件工程师聊过了。让我来预测他们会如何回答这个问题。
Speaker 122:22 - 22:30
But the tricky part is knowing what things can you answer and what can't you answer because
Speaker 122:22 - 22:30
但棘手的部分在于,你要知道哪些事情是你能回答的,哪些事情是你不能回答的,因为
Speaker 222:30 - 22:31
And how do you do that?
Speaker 222:30 - 22:31
那你是怎么做到的?
Speaker 122:31 - 22:54
Yeah. We try to be very explicit to the model of what is the domain of knowledge they have and then see how much can you expand that domain. That's kind of the fundamental idea. And we can back test how well the simulation works with what's in the kind of in our training data set. We remove one of questions.
Speaker 122:31 - 22:54
对。我们会非常明确地告诉 model(模型)它拥有哪些知识 domain(领域),然后再看你能把这个 domain 扩展到多大。这算是最核心的思路。我们还可以用训练 data set(数据集)里的内容做回测,看看这种 simulation(模拟)的效果到底有多好。我们会拿掉其中一个问题。
Speaker 122:54 - 23:08
And then see like, okay, how accurately did you predict that? And then you can add in nonsensical things like what's the name of their dog or something like that. And then you can say, like, is the model able to understand that you can't predict that? That's really cool.
Speaker 122:54 - 23:08
然后再看,比如说,好,刚才那个问题你预测得有多准确?接着你还可以加入一些毫无意义的东西,比如他们的狗叫什么名字之类。这样你就可以判断,model 是否能够理解这类事情是它无法预测的。那真的很酷。
Speaker 223:08 - 23:13
What sorts of things can you are you finding that you can predict well versus can't?
Speaker 223:08 - 23:13
你们发现,哪些类型的事情是可以预测得比较好的,哪些又是不行的?
Speaker 123:13 - 23:39
One of the most useful things and is message testing. So that's the idea of, like, how what what's the tagline on the billboard? Or I was actually using it this weekend. So I have created a a panel of of our customer base, and I had to come up with the title for a talk at a conference. And it's like a small thing, but it actually does matter because it will increase conversion if people show up.
Speaker 123:13 - 23:39
最有用的事情之一是 message testing(信息测试)。也就是比如说,广告牌上的 tagline(标语)该怎么写?其实我这个周末就在用它。所以我建立了一个由我们客户群体组成的 panel(样本小组),而我得为一次 conference(会议)上的演讲想一个标题。这看起来像是件小事,但实际上确实很重要,因为如果人们愿意到场,它就会提高 conversion(转化率)。
Speaker 123:39 - 23:53
And I came up with a 100 different titles for my talk and inputted that into our simulation. And then Oh, wow. The the top talk was, like, twice better than the next one. And Wow. Cool.
Speaker 123:39 - 23:53
我当时给我的演讲想了 100 个不同的标题,然后把它们输入到我们的 simulation(模拟)里。结果,哇,排名第一的标题大概比第二名好出两倍。哇,很酷。
Speaker 123:53 - 24:10
And I I like I don't know if it's correct, but it certainly felt correct. And it was really helpful to have guidance in making that decision. And I also think, like, even if it's wrong, it's just nice to make to have some help in making a decision. It's also nice to outsource your decisions.
Speaker 123:53 - 24:10
我觉得——我也不知道它是不是真的正确——但它确实让人感觉是对的。而且在做这个决定时,有这样的指导真的很有帮助。我还觉得,就算它是错的,有人帮你一起做决定,本身也挺不错的。把决策外包出去的感觉也很好。
Speaker 224:10 - 24:13
And how does it compare to just asking ChatGeePC same thing?
Speaker 224:10 - 24:13
那它和直接把同样的问题问 ChatGPT 相比怎么样?
Speaker 124:13 - 24:37
Yeah. So then I inputted the same questions into ChatGPT and actually had one, I had another talk I did that was not so successful and I inputted a competitor's or another talk that was more successful. And I showed both of them to ChattyBT and both of them to our simulation. And in ChattyBT, it picked the wrong one. And in our simulation, picked the right one.
Speaker 124:13 - 24:37
对,所以我又把同样的问题输入给了 ChatGPT。实际上,我还有另一个我做过但没那么成功的演讲,我又输入了一个竞争对手的、或者说另一个更成功的演讲。我把这两个都给 ChattyBT 看了,也把这两个都给了我们的 simulation。结果在 ChattyBT 那边,它选错了;而在我们的 simulation 这边,它选对了。
Speaker 124:39 - 24:54
You know, it's early for us. We're going to release this in a couple of months, but it seems like it's performing better than the general models. And the models are trained on the average person. Yeah. And you want to build for a very specific niche.
Speaker 124:39 - 24:54
你知道的,对我们来说现在还早。我们会在几个月后发布这个产品,但目前看起来,它的表现比那些通用模型更好。而那些模型是基于普通人的平均数据训练出来的。对,你真正想构建的是面向一个非常具体的 niche(细分群体)。
Speaker 124:54 - 24:58
And that's how we can kind of essentially train the models to follow that niche.
Speaker 124:54 - 24:58
这也就是我们如何在某种意义上训练这些模型去遵循那个 niche 的方式。
Speaker 224:58 - 25:41
And just to push on this, because I think it's so fascinating, like can't you kind of force the models into a specific niche or personality like, Hey, ChatGPT, you're a 35 year old, really grumpy software engineer that likes using your terminal. And then it does then take on the preferences of that niche. Sort of my mental model at least. And so I'm actually surprised that ChatGP wasn't able to arrive at the right answer and then bootstrapping off real user data was, because ultimately it all is kind of a reflection of real user data, right? And so actually, what is the intuition for why kind of SAM only on pre trained data isn't sufficient?
Speaker 224:58 - 25:41
我想继续追问一下,因为这点真的很有意思:你难道不能把模型硬“设定”进某个特定的 niche 或 personality(人格)里吗?比如说,“嘿,ChatGPT,你是一个 35 岁、脾气很差、喜欢用 terminal(终端)的软件工程师。”然后它不就会开始呈现出那个 niche 的偏好吗?至少这是我脑中的模型。所以我其实很惊讶,为什么 ChatGP 没法得出正确答案,而基于真实用户数据进行 bootstrapping(自举)却可以,因为归根结底,这一切某种程度上不也都是对真实用户数据的反映吗?所以,为什么仅靠只在 pre-trained data(预训练数据)上的 SAM 不够用,你们背后的直觉到底是什么?
Speaker 125:41 - 26:11
Yeah. So we've tried many different inputs and that certainly performs a little bit better than just vanilla ChatGPT. But what performs much better is we try credit card spend, kind of behavioral data, purchasing behavior. But what we found was the best dataset is interviews because it's more kind of allows you to go off tangents. It understands, you can ask behavioral questions.
Speaker 125:41 - 26:11
对,我们试过很多不同的输入,这种方法的表现确实比原版的 vanilla ChatGPT 稍微好一点。但效果好得多的是,我们会尝试用信用卡消费、行为数据、购买行为这类信息。不过我们发现,最好的数据集其实是访谈,因为它更能让你延展到一些支线话题。它理解得更好,而且你还可以问行为类问题。
Speaker 126:11 - 26:26
So also it can't just be any interview. Like the way you design the questions is also really important. And the intuition, I think, is that the models don't have clean data on how a specific persona acts and
Speaker 126:11 - 26:26
所以,也不能随便找个 interview(访谈)就行。比如你怎么设计问题也非常重要。我觉得背后的直觉是,模型并没有关于某个特定 persona(人物画像)如何行动、以及
Speaker 326:26 - 26:38
how they think? It's anecdotal, but it makes perfect sense. Because if you wanna understand someone, what better way to understand them than asking them a lot of questions? That's why we're all here. Guess it's kind of the purpose of this type of format.
Speaker 326:26 - 26:38
他们如何思考的干净数据?这更多是经验性的说法,但其实非常有道理。因为如果你想理解一个人,还有什么方式能比不停地问他们很多问题更好呢?这也是我们今天都在这里的原因。我想,这某种程度上就是这种形式存在的目的。
Speaker 326:39 - 27:13
And if you have enough people that follow a certain group as opposed to the average, that can tell you a lot about other things that they might not have explicitly said. You know, all of AI is this generalization of some sort of compressed data of some sort. And so if you have this compression in a slightly different part of this hyperspace that you say, now complete this orbital of what everybody is thinking in this category of person, you know, Listen can fill that out because it has enough interviews. Yeah. Do you think that you will offer that package as a product?
Speaker 326:39 - 27:13
而如果你掌握了足够多跟随某个特定群体、而不是平均人群的人,那这些信息就能告诉你很多他们可能没有明确说出来的其他事情。你知道,所有 AI 本质上都是对某种压缩数据的泛化。所以如果你在这个 hyperspace(高维空间)的某个稍微不同的位置上有了这种压缩,然后你说:现在把这个类别的人脑中“所有人在想什么”的这条轨道补全出来,Listen 就能把它补全,因为它已经有足够多的 interviews(访谈)了。对。你觉得你们会把那个打包成一个产品吗?
Speaker 327:13 - 27:42
As in, if I wanted to understand my customer, and for me, for us, our customers, founders, I they're very different, though. So extremely different people. If I wanted to understand my customer, could you do active interviews, the normal ListenLab interviews, have a thousand or 10,000 cumulatively, and then offer a little special purpose ListenLabs bot that then I can use instantaneously for any ad hoc question?
Speaker 327:13 - 27:42
也就是说,如果我想理解我的 customer(客户),对我来说,对我们来说,我们的客户是 founders,不过他们彼此差异很大。真的非常不同的人。如果我想理解我的 customer,你们能不能先做主动访谈、做普通的 ListenLab interviews,累计做到一千次或一万次,然后再提供一个小型的、专用的 ListenLabs bot(机器人),让我之后针对任何临时性的 ad hoc 问题都能即时使用?
Speaker 127:42 - 27:53
Yeah. That's exactly what we what we have. Okay. So that's that's what we call augmented responses. The cool part of this as well is that it can also live in your coding agents or your other agents.
Speaker 127:42 - 27:53
对,这正是我们现在有的东西。好。所以那就是我们所说的 augmented responses(增强式响应)。这里另一个很酷的地方是,它也可以直接存在于你的 coding agents(编码 agent)或者其他 agents 里面。
Speaker 127:54 - 28:09
So I think in the future, you will want to have almost a human API where the agents are able to call the preferences of your users to be able to know, like, what to build, how to do it, or who to invest in, or how to help them the best.
Speaker 127:54 - 28:09
所以我觉得未来你会希望拥有一种几乎像 human API(人类 API)一样的东西,让这些 agents 能够调用你的用户偏好,从而知道该构建什么、该怎么做、该投资谁,或者怎样才能最好地帮助他们。
Speaker 328:09 - 28:21
Today, is it all rag? Is it fine tuned? Is it something else? How do you take those conversations and then combine them with, you know, the models that you're doing the rest of the Listen Labs with?
Speaker 328:09 - 28:21
现在的话,全部都是 RAG(检索增强生成)吗?还是 fine-tuned(微调)?还是别的什么?你们是怎么把这些对话和你们在 Listen Labs 其余部分使用的那些模型结合起来的?
Speaker 128:21 - 28:32
Yeah. We are doing post training, typical RAG as well. There's, like, a bunch of different techniques. Some of them are proprietary, but yeah.
Speaker 128:21 - 28:32
对,我们会做 post-training(后训练),也会用典型的 RAG。大概有一堆不同的技术,其中一些是 proprietary(专有)的,不过,嗯,是的。
Speaker 328:33 - 28:37
Right. We'll do customer interviews and all your engineers report back.
Speaker 328:33 - 28:37
对。我们会做客户访谈,然后让你们所有工程师回来汇报。
Speaker 228:37 - 28:51
I'm curious what you think of multi agent systems and their role in helping us kind of iteratively use, at inference time, iterate to a better answer. Is that part of how you're doing simulation or not?
Speaker 228:37 - 28:51
我很好奇你怎么看 multi agent systems,以及它们在帮助我们以某种迭代方式、在 inference time(推理时)不断迭代出更好答案这件事上的作用。这是否是你们做 simulation(模拟)的方法之一,还是并不是?
Speaker 128:51 - 29:07
Yeah. Like the way we do simulation is essentially you have one person that you model really, really well, and then you scale it up with a thousand people. So you have a representative sample, it's essentially multi agent.
Speaker 128:51 - 29:07
对。比如我们做 simulation 的方式,本质上是先把一个人建模得非常、非常好,然后再把它扩展到一千个人。这样你就有了一个 representative sample(代表性样本),本质上这就是 multi agent。
Speaker 229:07 - 29:10
But you're not having those thousand people debate each other. That's what I'm asking.
Speaker 229:07 - 29:10
但你们并不会让那一千个人彼此辩论。我的问题是这个。
Speaker 129:10 - 29:19
Oh, yeah. No. We we don't have that yet, but that's a good help? Potentially. But there are other competitors that are doing that approach more.
Speaker 129:10 - 29:19
哦,对。没有。我们还没有这么做,不过这会是个不错的帮助吗?有这种可能。但也有其他竞争对手更多是在采用那种方法。
Speaker 129:19 - 29:39
I think the worry is that, again, chaos theory tells us that when things kind of compound, it becomes really hard to predict how the things are gonna interact with each other. And it's something we definitely should explore more, but I'm a little bit skeptical of the Maybe
Speaker 129:19 - 29:39
我觉得令人担心的是,还是那句话,chaos theory(混沌理论)告诉我们,当事情开始层层叠加时,就会变得很难预测这些事物彼此之间将如何相互作用。这绝对是我们应该进一步探索的方向,但我对这种做法还是有一点怀疑,也许——
Speaker 229:40 - 29:54
the analogy I'd make is like the AI council approach of send the same query out to three different LLMs and then have one LLM act as judge and synthesizing them. I do think on average you get a slightly better response. Yeah. Cool. So where else do you see yourself going from here then?
Speaker 229:40 - 29:54
我会打的类比,有点像 AI council 的做法:把同一个 query(查询)发给三个不同的 LLM,然后再让一个 LLM 充当裁判并加以综合。我确实认为,平均来说你会得到一个稍微更好的回答。对。不错。所以接下来你还觉得自己会往哪些方向发展?
Speaker 229:56 - 30:08
You're going from market research two point zero to market research three point zero now with kind of generative simulation. Do you expect that three point zero takes over as the majority of queries over time? And then what else is ahead?
Speaker 229:56 - 30:08
你们现在是从 market research 2.0 迈向 market research 3.0,也就是某种 generative simulation(生成式模拟)。你预计 3.0 会不会随着时间推移接管大多数 query?然后,再往后还有什么?
Speaker 130:08 - 30:46
Yeah. I think you'll still need human input, but I think there would be many more use cases that are now opened up where you can get customer input. So for the large decisions, if you're doing a Super Bowl ad or things like that, you will still need to run real interviews. But for the smaller things like what should be the tagline for your billboard, if it's a small billboard, then you can use simulation to answer that. And I still think that there's a lot of alpha on the core product as well to improve.
Speaker 130:08 - 30:46
对。我觉得你仍然会需要人的输入,但我认为现在会打开更多新的 use case(使用场景),让你能够获得客户输入。所以对于重大决策,比如你要投放一个 Super Bowl 广告之类的,你仍然需要去做真实访谈。但对于一些更小的事情,比如一个小型 billboard 的 tagline(宣传标语)该怎么写,那你就可以用 simulation(模拟)来回答这个问题。而且我也认为,在核心产品上仍然有很多 alpha(超额价值/优化空间)可以继续挖掘和改进。
Speaker 130:47 - 31:13
I mean, when we started, the core idea was just making the interview less annoying to go through. Like we had an eval that looked at repetitive questions or looked at is the AI even able to follow the instructions? And with GPT-four, sometimes we would ask the same question a 100 times. Yeah. And in the beginning, that eval was, 20%.
Speaker 130:47 - 31:13
我是说,我们刚开始做的时候,核心想法其实只是让整个访谈过程没那么烦人。比如我们当时有一个 eval(评测),会去看有没有重复提问,或者看 AI 到底能不能遵循指令。用 GPT-four 的时候,有时我们会把同一个问题问上 100 次。对。而在最开始,那个 eval 的结果只有 20%。
Speaker 131:13 - 31:34
Now we've been able to climb that eval to be 85%. But now we created a new eval that's much more advanced. So it's able to understand what are you doing on your screen when you're screen recording, or can you skip questions that are not relevant anymore? And now we're back at, like, 20%, which I think is one of the values that
Speaker 131:13 - 31:34
现在我们已经能把那个 eval 提升到 85%。但现在我们又创建了一个更高级得多的新 eval。所以它能够理解你在进行 screen recording(录屏)时,屏幕上到底在做什么,或者它能不能跳过那些已经不再相关的问题。而现在我们又回到了大概 20%,我觉得这正是其中一个价值所在——
Speaker 331:34 - 31:35
That's great. Yeah.
Speaker 331:34 - 31:35
这很棒。对。
Speaker 131:35 - 31:46
Vertical AI companies can have is that they have this proprietary eval that they can use and essentially climb that eval. And that's your advantage as a vertical AI company.
Speaker 131:35 - 31:46
Vertical AI 公司所能拥有的一个优势,就是它们有这种 proprietary eval(专有评测体系)可用,并且本质上可以不断把这个 eval 往上爬升。而这就是你作为一家 Vertical AI 公司的优势。
Speaker 331:46 - 31:50
Keep pushing forward, better data, harder problems, better data, repeat. Yeah.
Speaker 331:46 - 31:50
持续往前推进,更好的数据,更难的问题,更好的数据,重复这个循环。对。
Speaker 231:50 - 31:59
It seems to me like you're in the middle of a very interesting infinity loop. Right? Because, I mean, fundamentally, company is figure out what to build, build it.
Speaker 231:50 - 31:59
在我看来,你们正处在一个非常有意思的 infinity loop(无限循环)中间。对吧?因为归根结底,公司做的事情就是先弄清楚该做什么,再把它做出来。
Speaker 331:59 - 32:00
Yes.
Speaker 331:59 - 32:00
对。
Speaker 232:00 - 32:01
Figure out what to build, build it.
Speaker 232:00 - 32:01
弄清楚该做什么,再把它做出来。
Speaker 132:01 - 32:02
Write code and talk to users.
Speaker 132:01 - 32:02
写代码,并和用户交流。
Speaker 232:02 - 32:19
Exactly. Yeah. Write code is gonna come And the build it is coming up rapidly up and exponential. And the figure out what to build is the thing that you are pushing forward. And then it's not only even outside of product and engineering, the broader loop is actually strategy execution.
Speaker 232:02 - 32:19
没错,是的。写代码这件事会到来,而且把它构建出来的能力正在迅速提升,并且呈指数级增长。而真正需要你去推动向前的,是弄清楚该构建什么。甚至这还不只是产品和工程内部的事,更大的闭环其实是战略执行。
Speaker 232:20 - 32:38
And so much of what AI is enabling us to do is it's making execution faster, cheaper, better, all these things. And the thing that you guys fundamentally are positioning yourselves to do as a company is the strategy part from what to build to what to say. Is that a fair synthesis?
Speaker 232:20 - 32:38
所以,AI 正在使我们能够做到的,很大一部分就是让执行变得更快、更便宜、更好,诸如此类。而你们这家公司从根本上要把自己定位去做的,是战略这一部分——从构建什么,到说什么。这样概括公平吗?
Speaker 132:38 - 32:53
Yeah. And I think we'll when we have that one person billion dollar company, we'll be part of that loop. So we'll have a coding agent and listen and then run that in the loop. And we'll have these autonomous organizations and
Speaker 132:38 - 32:53
对。而且我认为,当我们迎来那种一人十亿美元公司时,我们会成为那个闭环的一部分。所以我们会有一个 coding agent(编码 agent),再加上 listen,然后让它们在这个闭环里运行。我们还会有这些 autonomous organizations(自治组织),以及
Speaker 332:53 - 33:36
And make even sense to the big companies, though, like back to this idea of you can implement things faster. Let's say you have an agent. I mean, if you can be a big company and we're talking in software because software is native to us, but in software, if you can talk to a customer, figure out a bug, create a PR, have a coding agent, close it, ship it, customer's happy, that seems like a really important left hand side of the equation, find the bug from an actual human. But I imagine it's the same thing in a big atoms company. Like, if you're consumer packaged goods, if you're clothing, if you're any of those things, I imagine that's even more important because you have to figure it out because once you actually do the thing, it's done.
Speaker 332:53 - 33:36
不过,这对大公司来说同样也说得通,还是回到这个观点:你可以更快地实施事情。假设你有一个 agent(智能体)。我的意思是,如果你是一家大公司——我们这里谈的是软件,因为软件对我们来说是原生的——但在软件里,如果你能和客户交流,找出一个 bug,创建一个 PR,让一个 coding agent 处理它,把它关闭,发布上线,客户也满意,那看起来就是这个方程左侧非常重要的一部分:从真实的人类那里发现 bug。但我想,在大型 atoms company(实体公司)里也是一样。比如你做 consumer packaged goods(消费包装品),做服装,或者做任何这类东西,我想这甚至更重要,因为你必须先把它搞清楚,因为一旦你真的把东西做出来了,事情就定了。
Speaker 133:36 - 33:50
Yeah. Exactly. Like, Procter and Gamble, when they're launching in a new market, that can be tens of millions of dollars, if not more. And you have to make sure that that is right when you launch. That's one of the reasons why they're the customers have listened.
Speaker 133:36 - 33:50
对,完全正确。比如 Procter and Gamble 在一个新市场推出产品时,那可能要花掉数千万美元,甚至更多。而且你必须确保在推出时这件事就是对的。这也是他们会成为 listened 客户之一的原因。
Speaker 333:50 - 34:08
Who has done this historically really well? Like, who are the companies that are admired in history, have done a great job of listening to their customers, either in the, you know, consumer space or in the software space?
Speaker 333:50 - 34:08
历史上有哪些公司真正把这件事做得很好?比如说,哪些在历史上受人尊敬的公司,不管是在消费领域还是软件领域,都非常擅长倾听客户?
Speaker 134:08 - 34:31
I think Procter and Gamble is is kind of the archetype of best market research organization where they're essentially marketing companies that are trying to figure out what are niches that people really care about and then build specific brands to solve those problems. I mean, one example is the Tide washing machine
Speaker 134:08 - 34:31
我认为 Procter and Gamble 算是最佳市场研究组织的一种原型,他们本质上是营销公司,努力弄清楚人们真正关心的是哪些细分需求,然后打造特定品牌去解决这些问题。举个例子,就是 Tide washing machine。
Speaker 334:31 - 34:32
Yeah. Pods.
Speaker 334:31 - 34:32
对。Pods。
Speaker 134:33 - 35:14
Pods that they were able to figure out that it was really uncomfortable to use the washing liquid. And they discovered that people wanted something that was much more easy to use. And through customer interviews, they found this insight, made this new type pod and became really successful. Another example, which is in the Acquired podcast when they talk about Mars, they did the first one of the first market research studies in the 1950s where M and M's were originally designed for being in the used in the army because they were like a sweet treat. Don't melt in pocket.
Speaker 134:33 - 35:14
他们发现,人们已经意识到使用洗衣液其实很不方便。随后他们又发现,人们想要一种更容易使用的东西。通过客户访谈,他们得到了这个洞察,做出了这种新型 pod,并且取得了非常大的成功。另一个例子出自 Acquired podcast 讨论 Mars 的那期:他们在 1950 年代做了最早的一批市场研究之一,而 M and M's 最初其实是为军队使用而设计的,因为它们算是一种甜点零食,而且放在口袋里也不会融化。
Speaker 135:15 - 35:32
They discovered through market research that another great segment was young kids. And they then decided to pivot the entire advertising strategy to focus on this because it doesn't melt and ruins your furniture, for example, and things like that.
Speaker 135:15 - 35:32
他们通过市场研究发现,另一个非常好的细分人群是小孩子。于是他们决定把整个广告策略都转向这一点,比如强调它不会融化、不会把你的家具弄脏之类的特点。
Speaker 235:32 - 35:42
As we progress towards this, you know, Listen Labs future vision of the world, what are the things that you're confident will work? And what are the things that you're still not sure about?
Speaker 235:32 - 35:42
随着我们朝着这个方向推进,也就是你知道的 Listen Labs 对未来世界的愿景,你确信哪些事情会奏效?又有哪些事情你仍然拿不准?
Speaker 135:42 - 36:22
I'm confident that in the future you'll still need to have human input because even if you have a perfect rational being, like AGI, humans are still irrational. Totally. And they will still want kind of be chaotic in their nature where they all of a sudden get obsessed with a new product, a new TikTok trend that shows up and you have to change your entire marketing strategy towards that. And so I think that will remain a really huge part of how we do things. I think I'm still uncertain of what level simulation will play.
Speaker 135:42 - 36:22
我确信,未来仍然需要 human input(人工输入),因为即便你拥有一个完美理性的存在,比如 AGI,人类依然是不理性的。完全没错。而且他们天性中仍然会带有某种混乱性——会突然迷上一款新产品,或者突然冒出一个新的 TikTok 趋势,于是你就不得不围绕它调整整个营销策略。所以我认为,这仍然会是我们做事方式中非常重要的一部分。我现在仍然不确定的是 simulation(模拟)会在其中扮演多大角色。
Speaker 136:22 - 36:31
And I'm confident that it will work for certain questions, but we'll see how, like, how good
Speaker 136:22 - 36:31
而且我确信,对于某些问题它会奏效,但我们还要看看,比如说,
Speaker 336:31 - 36:48
the models get to predicting human behavior. I mean, I'd imagine that it's actually even more important the better AI gets to have the delta because the competition if companies are about serving people, which I think we can all agree on. Like, at the end of the every company is about serving humans.
Speaker 336:31 - 36:48
这些 model(模型)在预测人类行为方面到底会变得多好。我的意思是,我甚至会想,AI 越强,拥有这种 delta(差异化优势)反而越重要,因为如果公司的竞争本质上是围绕服务人来展开的——我想这一点大家都同意——那么归根结底,每一家公司都是在服务人类。
Speaker 236:48 - 36:50
Constantine, a resident humanist.
Speaker 236:48 - 36:50
Constantine,一个常驻 humanist(人文主义者)。
Speaker 336:50 - 37:12
I've I've I've I'm a humanist. Absolutely. But if companies are about serving people, because that's what, that's why we're all working, is to help someone else in some way, and intelligence gets better and better and better, and you kinda have what the human wants here and the intelligence is approaching that asymptote, then the delta in that asymptote, which is what is in a human's mind that isn't in the AI's mind, only becomes more important.
Speaker 336:50 - 37:12
我我我是一名 humanist(人文主义者)。当然是。但如果公司存在的意义是服务人,因为这正是、这就是我们所有人工作的原因——以某种方式帮助别人——而 intelligence(智能)变得越来越好、越来越好、越来越好,同时你大概可以把人的需求看作在这里,而 intelligence 正在逼近那条 asymptote(渐近线),那么在那条 asymptote 上的 delta(差值)——也就是人类头脑里有而 AI 头脑里没有的东西——就只会变得更加重要。
Speaker 137:12 - 37:26
Yeah. And, yeah, one of the things that we've also realized is that there's, you know, a lot of talk around what's the motive of these AI vertical AI companies. Yeah. And Yeah. What's your motive?
Speaker 137:12 - 37:26
对。还有,对,我们也意识到的一件事是,关于这些 AI、vertical AI companies(垂直领域 AI 公司)的动机,外界有很多讨论。对。然后,对。你的动机是什么?
Speaker 137:26 - 37:27
What is our motive?
Speaker 137:26 - 37:27
我们的动机是什么?
Speaker 337:27 - 37:30
We've got network effects and scale economies. Yes. We we Those are nice.
Speaker 337:27 - 37:30
我们拥有 network effects(网络效应)和 scale economies(规模经济)。对。我们,我们——那些都很不错。
Speaker 237:30 - 37:32
I feel like we're on an episode of Acquired
Speaker 237:30 - 37:32
我感觉我们现在像是在录一期 Acquired
Speaker 337:32 - 37:35
right now. Hey. I'm feeling it. I'm feeling it right now. It's it's a good book.
Speaker 337:32 - 37:35
一样。嘿,我有那种感觉。我现在真的有那种感觉。这,这本书很好。
Speaker 337:35 - 37:38
Recommend it. Seven Powers.
Speaker 337:35 - 37:38
推荐它。Seven Powers。
Speaker 137:38 - 37:49
Yeah. Seven Powers. Love love Seven Powers. Yeah. On on the modes, I mean, we have the clear modes, which are the network effects on the panel where you have supply and demand dynamic.
Speaker 137:38 - 37:49
对。Seven Powers。很喜欢,很喜欢 Seven Powers。对。说到 moat(护城河),我的意思是,我们有很明确的 moat,也就是 panel(平台)上的 network effects,在那里你会有 supply and demand(供需)动态。
Speaker 137:49 - 38:24
We also have the network affects the data moat. As we do more interviews, you get better simulation. And then the product is very sticky because you have all these interviews in your platform and you don't wanna lose that. You wanna track things over time. But even like the simplest things, I think, in terms of product advantages, like one of the first things that Brian Shire said, one of our Sequoia partner, was that founders want to build something that's complex, but customers want something that's stupid, simple, and it just works.
Speaker 137:49 - 38:24
我们还有网络效应会影响 data moat(数据护城河)。随着我们做更多访谈,你会得到更好的 simulation(模拟)。然后这个产品会非常 sticky(粘性很强),因为你所有这些访谈都在你的平台里,你不想失去这些内容。你还想持续追踪这些事情的变化。不过即使是最简单的层面,我认为从产品优势来看也是如此。比如我们的一位 Sequoia partner,Brian Shire,最早说过的一句话就是:founder 想打造复杂的东西,但客户想要的是那种蠢到不能再简单、而且就是能用的东西。
Speaker 138:25 - 38:52
They don't want to configure their own workflow. They don't wanna sit and build a custom software. And just one example of this is creating the interview guide is really difficult. It's actually an academic subject. And it's one of the reasons why you have services firms because they know what methodology to use if you want to understand pricing or brand perception, these kind of things.
Speaker 138:25 - 38:52
他们不想自己去配置 workflow(工作流)。他们也不想坐下来搭一套 custom software(定制软件)。举个例子,创建 interview guide(访谈提纲)其实非常困难。这实际上是一个学术课题。也正因为如此,才会有 services firms(服务公司)存在——因为如果你想理解 pricing(定价)或者 brand perception(品牌认知)这类问题,他们知道该用什么 methodology(方法论)。
Speaker 338:52 - 38:53
You don't want to lead the witness.
Speaker 338:52 - 38:53
你不能去引导受访者。
Speaker 138:53 - 39:14
You don't want to lead the witness. And it's really hard to get that right. In the beginning, we just used the vanilla LM models and the customers would create interviews, they would get the data back, and then they'd come back to us really frustrated saying like, what is this? Like, I can't use this data for anything. And we took the blame for that.
Speaker 138:53 - 39:14
你不能去引导受访者。而这件事其实很难做好。刚开始的时候,我们只是用了最基础的 vanilla LM models(原始语言模型),客户会创建访谈、拿回数据,然后非常沮丧地回来对我们说,这是什么东西?这些数据我根本没法用。对此我们也承担了责任。
Speaker 139:15 - 39:32
Now we've trained it to follow the best practices so that you always get good data out of the interviews. And I think that's the advantage you have as a vertical AI company that you can essentially train this agent to follow best practices in the work that you do.
Speaker 139:15 - 39:32
现在我们已经把它训练到会遵循 best practices(最佳实践),这样你总能从访谈中得到高质量数据。我认为这正是 vertical AI company(垂直 AI 公司)的优势所在:你基本上可以把这个 agent 训练成遵循你所在工作领域中的最佳实践。
Speaker 239:32 - 40:01
So I wanna go back to the concept of tide pods that you had mentioned earlier. I think it's really interesting. And so much of market research, as I understand it today, is almost more inviting people to pass judgment on ideas that you feed them. But it seems to me that one of the Hallucinations can be a bug. They can also be a feature with And generative do you think we're gonna see user research actually evolve into live product ideation?
Speaker 239:32 - 40:01
所以我想回到你之前提到的 Tide Pods 这个概念。我觉得这很有意思。据我今天的理解,market research(市场研究)在很大程度上更像是在邀请人们对你喂给他们的想法做判断。但在我看来,Hallucinations(幻觉)有时可以是 bug(缺陷),也可能在 generative(生成式)场景里成为一种 feature(特性)。你觉得我们会不会看到 user research(用户研究)真正演变成 live product ideation(实时产品创意生成)?
Speaker 240:01 - 40:19
I could almost imagine AI inventing solutions as customers are going about their interview process, even helping visualize those solutions. Are your customers doing that already? Or do you think we're going to have the moment where AI can create a Tide Pods idea in a market interview anytime soon?
Speaker 240:01 - 40:19
我几乎可以想象,AI 会在客户进行访谈的过程中一边发明解决方案,甚至还帮助把这些解决方案可视化。你的客户现在已经在这么做了吗?或者你觉得我们很快会迎来这样一个时刻:AI 能够在市场访谈中创造出一个像 Tide Pods 这样的点子?
Speaker 140:19 - 40:50
Yeah, I think that's really exciting. Today, they do that manually, use AI to generate images of different concepts and feed that into the interviews. But I think specifically also with simulation, it becomes really powerful. So we now have an MCP as well so that you can feed that into Claude and then you can tell Claude like, hey, run, listen in the loop, and then come up with a bunch of ideas for how to market something or different concepts.
Speaker 140:19 - 40:50
会的,我觉得这真的很令人兴奋。今天,他们还是手动这么做:用 AI 生成不同概念的图片,再把这些内容放进访谈里。但我认为,尤其是结合 simulation(模拟)时,这会变得非常强大。所以我们现在也有一个 MCP,这样你就可以把它接入 Claude,然后你可以对 Claude 说,嘿,跑一个 listen in the loop(环路监听),然后想出一堆关于如何营销某个东西的点子,或者不同的概念。
Speaker 340:50 - 40:50
And
Speaker 340:50 - 40:50
还有,
Speaker 140:50 - 40:54
then you can have it run like that. I'm even thinking in
Speaker 140:50 - 40:54
那你就可以让它那样运行。我甚至在想,
Speaker 240:54 - 41:04
the course of an interview as somebody's complaining about, you know, the tide is I It's not very portable. For the AI to be, you know, live brainstorming with you solutions, not just soliciting.
Speaker 240:54 - 41:04
在面试过程中,当有人在抱怨,比如说,这个东西——它的可移植性不太好——让 AI 直接和你进行实时头脑风暴,一起想解决方案,而不只是单方面收集信息。
Speaker 141:04 - 41:05
Yeah. Is what it
Speaker 141:04 - 41:05
对。这就是它
Speaker 341:05 - 41:08
could look like, an image generator too. Yeah. That'd be cool, Sonya.
Speaker 341:05 - 41:08
可能呈现的样子,再加上一个 image generator(图像生成器)也可以。对。那会很酷,Sonya。
Speaker 141:08 - 41:11
Yeah. I think it's a good idea. You should be on our product team.
Speaker 141:08 - 41:11
对。我觉得这是个好主意。你应该加入我们的产品团队。
Speaker 241:12 - 41:29
Awesome. Well, Alfred, we really love what you're building. Thank you for taking the time to share insights both on the Brother market, which I think is just so fascinating, and also what it takes to be building in the application layer right now. We really admire the business that you've built, thank you for your continued partnership.
Speaker 241:12 - 41:29
太棒了。Alfred,我们真的很喜欢你正在打造的东西。感谢你抽时间分享见解,不仅谈到了 Brother market——我觉得这真的非常有意思——也谈到了当下在 application layer(应用层)进行构建需要具备什么。我们真的很钦佩你打造出的这门生意,也感谢你一直以来的合作。
Speaker 141:29 - 41:29
Thank you so much.
Speaker 141:29 - 41:29
非常感谢。
原文 ↗https://www.youtube.com/watch?v=Rumft-rsEu4
BuildSpeak — 关于本项目BUILT IN PUBLIC · 跟随 builders 而非 influencers