AI Avatar 与 DeepFake 技术的未来|Joshua Xu、Steve Brown 对谈 | EP #161
- HeyGen押注的是:视频制作将成为软件规模化的职能,而不再是受摄影机束缚的手艺。 Joshua Xu表示,如今能制作出优秀视频的人不到1%,预计企业制作的视频数量将增加“100倍”,随着 AI 消除摄影机门槛、且预计很快解决剪辑问题,邮件和文字都可以转成视频。HeyGen在2020年确立的初始使命说得很直白:“替代摄影机”(“replacing the camera”)。
- 更大的机会,是让视频成为 AI 劳动力的人类交互界面。 Xu描绘的演进路径,是从文字聊天到语音,再到视频会议;智能体可以跨语言工作,并同时运行数百乃至数千个会话。当 Diamandis问他设想中的20人虚拟营销团队何时可用时,Steve Brown回答:“现在。”
- 这套生产方案结合了速度、本地化和更低成本的迭代。 HeyGen宣传片称,数字人视频的制作速度是“传统制作的10倍”,支持4K和175种语言;Xu表示,大约2分钟训练视频就能个性化面部动作、表情和肢体语言。Diamandis举例说,定价、命名和其他产品要素随后都可以调整,无需重新拍摄。
- Avatar是呈现层;Agent则增加知识、判断和行动能力。 Brown的简洁概括是:“Avatar是AV,即audio visual;Agent是agency。”书籍、访谈、文件、价值观、人生故事和轶事,为智能体提供大脑。Diamandis还设想过,这些拥有独立名称的智能体可以出现在 Slack、邮件或视频通话中。
- 当逼真的身份可以被编程时,授权与审核就成为战略要求,而非外围功能。 Xu表示,HeyGen的每个 Avatar 都必须获得本人视频授权,公司“从未在授权问题上妥协”;AI与人工审核员也会限制仇恨言论、欺诈和政治竞选活动。对于被克隆的紧急来电,Diamandis给出的实用对策是设置一个家庭私密密码。
- Brown的医疗智能体演示展示了跨数据孤岛进行模式识别的价值,但仍只是个人案例,不能视为临床验证。 一个智能体复核他的既往记录后,将低免疫球蛋白、贫血、铁蛋白升高和体重下降联系到一种可能的浆细胞疾病,而此前专科医生尚未发现这一点;Brown后来被诊断为多发性骨髓瘤。他的更大判断是,差异化智能体可以从多个角度检索同一套碎片化数据,而 Brown说,它们“也可能正在挽救你的生命”。
1. HeyGen从“替代摄影机”的押注起步
Xu将HeyGen的起点追溯到2020年。那时,“生成式 AI”这个词还没有出现。在 Snap 负责 AI 增强摄影体验时,他看到了生成现实中不存在、却依然高度逼真的内容的可能性。
公司的创始使命是“替代摄影机”(“replacing the camera”)。过去15年,手机摄影机扩大了内容创作规模,但 Xu认为,面对摄影机的不自在和制作成本高昂,仍是企业进行故事表达的持续障碍;后来 Diamandis又举例说明了灯光和补拍带来的负担。
他设想的终点是“人人都能进行视觉叙事”:目前能制作优秀视频的人不到1%,但最终每家企业都可能多制作“100倍的视频”,把日常邮件、图片和文字转化为有吸引力的视频。
2. 视频正成为并行 AI 劳动力的交互界面
Xu的数字分身展示了无需摄影机或摄制团队即可进行预录制表达,支持170多种语言和方言,并能在网站、产品或平台中进行实时对话。
Xu对数字界面的路线设想,是从文字输入到语音,再到与智能体进行视频会议。重点不只是观看生成的视频,而是与 AI 面对面互动。
Diamandis设想过一个充分了解其工作的身份化智能体,能够给出与他本人相同的回答,然后向各类会议派出1000个副本。Xu进一步表示,这些智能体可以跨越语言障碍,同时运行“数百、数千个不同会话”。
当 Diamandis问20人虚拟营销团队何时可以用上时,Brown回答:“现在。”
Diamandis举出了一个即时应用场景:快速迭代。Xu在4年间创建了300多个 Avatar,但他说,拍摄仍需要准备、良好灯光,有时还要补拍1到2次;Avatar则可以在定价、命名或定位发生变化时随时修改销售话术。
3. Avatar提供存在感,知识层赋予行动能力
Xu将 AI Agent 定义为一种助手:不仅能执行初始动作,还能完成一连串行动;数字分身则让人们能够接触到它所代表的那个人。Brown的概括很简洁:“Avatar是AV,即audio visual;Agent是agency。”
一张可信的脸能建立情感连接,但要赋予行动能力,还需要一个了解用户并赢得信任的大脑。HeyGen将其系统描述为“以人为中心的视频模型”,学习的不只是一个人的脸和声音,还包括其手势、表情、肌肉动作与肢体语言之间的对应关系。大约2分钟训练视频就足以完成这种个性化。
文件、演讲和访谈,会成为数字分身背后的 LLM 知识层。对于没有大量公开作品的人,Brown正在打造访谈智能体,通过提问挖掘其价值观、人生目标、生活故事和个人轶事——这些内容可能相当于一本传记中的“数百页”。
产品宣传片展示了4K输出、175种语言,以及“比传统制作快10倍”的创作速度。活动现场,一名参会者表示,素材采集“真的……只用了4分钟”;他的 Avatar 连夜完成处理,并用他自己的声音说出了另一种语言。
4. 授权是 Avatar 与 DeepFake 的边界
Diamandis警告,逼真的 FaceTime 来电者可能冒充亲属,要求支付保释金或进行电汇。他的实用建议是预先设立家庭密码,并趁父母或祖父母仍在世,尽可能保存他们的视频、声音和故事。
Xu表示,每个 Avatar 都必须获得本人视频授权,无论请求者是谁,HeyGen从未放弃这一要求。AI与人工结合的审核系统还禁止平台上的仇恨言论、欺诈用途和政治竞选活动。
5. 医疗智能体展示了跨数据孤岛检索的价值
Brown的经历始于洛杉矶山火烧毁他的梦想之家、迫使他搬到 Palm Springs。一次因胃痛前往急诊的就医,最终带来为期8天的检查和多发性骨髓瘤诊断;此前几位专科医生没有发现异常,只认为问题可能源于压力、过度工作或抑郁。
Brown要求自己的“Hippocrates”智能体只检查更早的医疗记录。智能体指出 IgA 和 IgM 偏低、γ球蛋白偏低、轻度贫血、铁蛋白升高,以及无法解释的体重下降,并提示可能存在浆细胞疾病,建议与医生讨论骨髓评估。Brown说,智能体甚至比最终诊断更早发现了这一模式。
后续交流中,智能体列出了检测 κ 和 λ 链的血清游离轻链检测、具有确诊意义的骨髓活检,以及用于检测 Bence-Jones 蛋白的24小时尿蛋白电泳。Brown表示,经过2周化疗和 daratumumab 免疫治疗后,他的轻链水平下降了50%以上;智能体称这是令人鼓舞的结果,同时建议继续接受肿瘤科医生监测。
Brown的结论是,医学知识不断扩张,迫使专科医生进入更狭窄的数据孤岛,而“AI没有孤岛”。除了这个“全能智能体”,他还在打造100个取向不同的医疗智能体,让不同视角成为检索碎片化健康数据的替代路径。这仍是 Brown 的个人演示,不构成临床验证。
核验说明
- 在 Hippocrates 的交流中,第一次回复列出的是γ球蛋白偏低,而后续回复却将这一模式称为“高丙种球蛋白血症”;逐字稿没有解决这一矛盾。
Where does this go? What's your vision for where this is going? How will I use these avatars?
Very, very soon we are actually gonna see everybody be able to, like, have a meeting with a agent in, in the video interface.
Joshua Xu.
Joshua Xu.
Joshua Xu.
The mind behind HeyGen.
He is redefining how we create and communicate.
You will be able to have your agent talk to people in different languages.
175 languages, that's insane.
I'm very excited about the mission that we have enabling, you know, visual storytelling for all.
The technology's gotten so extraordinarily powerful, uh, and yet we're just at the very beginning.
We take trust and safety seriously on the platform. Every single avatar created on HeyGen will require the first-party consent.
This is the future, which is available when?
Now.
Now, that's a moonshot, ladies and gentlemen.
So Joshua, you have built an extraordinary company. I mean, I think if you have... know anything about digital avatars, HeyGen is the top of the game. Uh, congratulations.
Thank you.
How long ago did you start HeyGen? Tell me a little bit about your journey. I want to understand where your technology is and, in our conversation, understand where it's going.
Sure. The journey of HeyGen started back in 2020.
Long ago in 2020.
Yeah.
Yeah.
1. HeyGen Replaces the Camera
I think that was the time when we didn't have the term "generative AI," and I was working on very similar technology, using AI to enhance the camera experience at Snap. I just felt that the technology could continue to evolve and potentially generate something that does not exist in the world but still feels highly realistic.
Right.
So that's how we started the company. In fact, I think the funny thing about the mission of the company in the beginning was that we called it "replacing the camera."
Replacing the camera.
Because if we look back at the past 15 years, the mobile camera is really the thing that enabled a lot of people to create content. But there are still so many people who aren't able to create good content using a camera, either because they're camera-shy or because cameras are very expensive for businesses.
We felt that if we could replace the camera, we could remove the barrier to visual storytelling for businesses. That's how we started the journey.
Amazing. Well, you've come a very long way. I think you've got a video to kick us off.
Yeah.
To give us a sense of your technology. Let's run that video so folks can get a sense.
Sure.
Yep. Hey everyone, I'm Joshua. And yes, I'm here on stage with you, but I'm also right here as his digital twin. Do we look alike? This is HeyGen's avatar technology in action. I can deliver video messages without a camera or production crew. This ranges from video ads to how-to videos to trainings and so much more. I can also speak in over 170 languages and dialects. For example, Hola. Abundancia tres sesenta. Estoy muy emocionado de sumergirme contigo en el futuro de la generación de videos con inteligencia artificial. Besides just speaking different languages, I can even engage in real-time conversations, whether it's on your website, product, or platform. Imagine what this could mean for your business. Joshua and Peter, I'll let you take it from here.
Okay. Love that. The technology has gotten so extraordinarily powerful, and yet we're just at the very beginning. Where does this go? What's your vision for where this is going? How will I use these avatars?
2. Avatars Become AI Interfaces
I think things are happening in 2 ways. One thing is that every individual will be able to easily visualize their ideas, and every business will be able to access more video creation. Every business will be creating 100X more videos in business communication, and I think business content is going to be more engaging compared to the old days of emails, images, and text.
That's happening very, very fast today. We've seen customers creating tons of videos to promote their businesses. On the other side, which is also very exciting, I think the AI agent world is evolving very fast as well.
The next interface for us to interact with an AI agent may actually start with chatting by typing text. Then, very soon, we're going to interact with the agent through a voice interface. Very, very soon, we're actually going to see everybody be able to have a meeting with an agent in a video interface. I think that's something we're seeing that's very exciting for the future.
Yeah, I mean, you've been building these avatars, and you can have your avatar attend a Zoom meeting for you.
Yeah.
Right?
Yep.
I love that. I don't know about you guys, but for some Zoom meetings, I would much rather have my avatar do the work.
I think about a future where I can build a really rich, agentic agent—in other words, an agent that represents me, knows everything about me, knows everything about my company—so that if you ask the agent a question or ask me a question, you're going to get the same answer.
Imagine being able to create not 1 but 1,000 agents and send them out to 1,000 different conferences on your behalf, have them cut deals, and bring back business for you. Is that possible? Is that where we're going?
Yeah. Not only that. I think you'd also be able to have your agent talk to people in different languages, which wasn't possible before. It's always hard to reach customers in different regions and languages, and now you can really enable that.
That's 1 thing. The second thing is that you can run your agent meetings in parallel and have your customers interact with you at the same time, with hundreds or thousands of different sessions happening simultaneously.
Yeah. And the other thing is, if you're working on new products or services, I know my team will bring me in and say, "Hey, can we get you to record a video at this particular time?" But you put the video up, and it's not hitting properly, so you want to change the pricing, change the name, and change all these things.
Instead of pulling me into the camera room or hiring someone to come in and talk about your new longevity clinic or whatever it might be, you can spin up your HeyGen avatars and iterate as rapidly as you want, right?
Yeah. I think that's exactly how we started the company and solved the camera problem. In fact, even for myself, I've already created probably more than 300 different avatars in the past 4 years, just testing out the technology.
But even today, if you asked me to film something in front of a camera, I would still feel like I needed to do some preparation. I might need 1 or 2 reshoots, and I would need to find good lighting and things like that.
An avatar really helps solve that problem for people and enables them to create content anytime.
Yeah. I remember last year, Steve Brown, my chief AI officer and a member of our community here, who will be coming out and joining us in a couple of minutes, created a PeterBot avatar of me. It uploaded all of my knowledge, all of my books, all of my podcasts—everything known about me.
I did a podcast with PeterBot, and I was blown away. PeterBot was much more eloquent, much more knowledgeable, and much smoother. Honestly, I felt very deficient in front of PeterBot. It didn't stammer while trying to remember something in its memory, and it was just very compelling.
There's incredible power in what you have.
Yeah. That's very exciting. We're just getting started.
Yeah. It's just getting started. I know you've got a marketing video, um, which I've asked you to play because I want people to see all the different ways that people can utilize this. I really want this to be one of the technologies everybody plays with this year. It's easy. You just have to make a commitment to actually try it. So please, you know, one of the things I'm looking for you to do is commit to certain things. This is one of them. Let's, let's play that marketing video, if we could.
Yeah, sure.
Today, video is everything. You need speed, creativity, and consistency. With HeyGen, you can create the most lifelike avatar videos 10 times faster than traditional production. Scale across one hundred and seventy-five different languages, looks, and platforms, all in 4K. Your clients want more: more videos, more engagement, and more results. HeyGen gives you the tools to deliver. Need personalized videos that feel custom? Done. Launch interactive, engaging campaigns that deliver more for your clients, and do it all without blowing the budget. No expensive studios, no costly talent. Just beautiful, effective videos that drive results. And here's a secret: We used HeyGen to create me. Join the future of storytelling. Become a HeyGen certified agency partner.
Everybody, I hope you're enjoying this episode. Did you know that we're likely to see as many as ten billion humanoid robots by twenty forty, and that Brett Adcock, the CEO of Figure, anticipates they'll have robots in our home in the next two to three years? How about Max Hodak's new form of BCI called biohybrid interfaces that could offer millions of connections between your neocortex and the cloud? Then there's Michael Andrej, whose efforts at Eon is focusing on uploading the human connectome to the cloud by twenty thirty. These aren't science fiction scenarios. They're serious efforts underway today. I've distilled the most powerful insights and roadmaps from this year's Abundance twenty twenty-five summit into a comprehensive report that will transform how you see the future. Get your free copy of the Abundance twenty twenty-five summit summary at diamandis.com/breakthroughs. That's diamandis.com/breakthroughs. I find that extraordinarily impressive. We're going to talk in a few minutes about how each of you can walk away with your own HeyGen avatar this week, right? It should be something that all of you commit to do, including the teams. I would love you guys to do this.
3. HeyGen Personalizes Every Avatar
So let's talk about the 2 parts. There's the voice, and then there's the video persona. One of the improvements in the video persona is hand movements and backgrounds. That's relatively recent. Can you speak about that?
Yeah. If you look at the technology itself, what we are truly building is a human-centric video model.
Right.
Right. The difference between a human-centric video model and a general video model is that we need to be able to present an engaging effect, such as gestures, the way you move your body, body language, and expression. There are a couple of elements out there. Voice is definitely one. Video is another one. What we usually require is roughly 2 minutes of training footage from you.
2 minutes of video, okay.
Yeah, 2 minutes of video. Then our AI model will try to learn from the way you speak, how the muscles move on your face, your expression, and, very importantly, how your body language matches your expression.
Yeah.
Because everybody's body language is different. I think the 2 minutes of footage will help us really personalize your model's behavior. After that, that's roughly how it works. Obviously, if we build our own AI agent behind it, you need to input the knowledge base that you have and whatever documents, keynotes, and interviews you might have had before, and we enable that as an LLM layer behind it.
Yeah. It's truly amazing. Let me suggest 2 things to everybody if you've not done this. I think they're very important in this age of photorealistic visual and auditory avatars. The first is, you're going to receive a phone call from somebody you think you know, and it may be a FaceTime video from someone you think you know. They may ask you to do something incredibly crazy, like, “Can you please wire money here?” or, “Can you please send money so I can get out of jail?”
I'm serious. I've had lots of friends who've had that happen to their aging parents and so forth. Everybody here, one thing that you should do is choose a word, a password, or something that, if you have any doubt about the authenticity of the phone call or the video call you're getting, ask for it, right? It makes sense. It's super easy to do. Have that conversation at dinner.
The second thing is, if you're lucky enough to have your parents or grandparents around, do 2 things. If you can, record their avatar. If you can't do that, sit them down for a couple of hours with your iPhone and ask them for their story, for their history. Record their video, record their voice. And this is for your kids and their grandkids, right? Please do that. It's extremely valuable. You'll be able to bring them back as your family history. Make sense? Great.
4. Everyone Gets a Digital Twin
So Joshua, let's talk about what you have enabled for us here at the Google Tech Hub.
Yeah, sure. We have our booth in the back. We're giving all the attendees the opportunity to create their own digital twin right at the conference, and it's free. You're going to be able to record your training footage for the AI model training, and then you're going to be able to get your avatar today. It's a glimpse into the future of AI video creation, and we'll be able to test it out on your HeyGen account.
Yeah. I want you to see how good this is. We had a few members do this last night, and I want to show you one of them that got processed overnight. Rikard, are you here in the audience? Where are you? Okay, Rikard, good to see you, my friend. I'm waiting for my flying Jetson as soon as I can get it. Let's watch Rikard's avatar.
Let's play the video: “From now on, I can speak in any language with my own avatar and my own voice.”
Amazing, huh? Let's give it up. That was extraordinary. And, Rikard, how easy was that to do? It literally took 4 minutes. 4 minutes of your time. Wow. And in 175 languages. That's insane.
So, again, please take this home with you. Play with this. Your whole goal here is to do something different. At least show it to your kids or your spouse when you get back.
Yeah.
It was about 13 years ago, I had my two kids, my two boys, and I remember at that moment in time, I made a decision to double down on my health. Uh, without question, I wanted to see their kids, their grandkids, and really, you know, during this extraordinary time where the space frontier and AI and crypto is all exploding, it was, like, the most exciting time ever to be alive. And I made a decision to double down on my health. And I've done that in three key areas. The first is going every year for a Fountain upload. You know, Fountain is one of the most advanced diagnostics and therapeutics companies. I go there, upload myself, digitize myself, about 200 gigabytes of data that the AI system is able to look at to catch disease at inception. You know, look for any cardiovascular, any cancer, any degene- neurodegenerative disease, any metabolic disease. These things are all going on all the time, and you can prevent them If you can find them at inception. So super important. So Fountain is one of my keys. I make that available to the CEOs of all my companies, my family members, 'cause, you know, health is the new wealth. Uh, but beyond that, uh, we are a collection of forty trillion human cells and, uh, about another hundred trillion bacterial cells, fungi, viri. And we, you know, don't understand how that impacts us. And so I use a company and a product called Viome. And Viome, uh, has a technology called Metatranscriptomics. It was actually developed, uh, in New Mexico, uh, the same place where the nuclear bomb was developed, as a biodefense weapon, and their technology is able to help you understand what's going on in your body, to understand which bacteria are producing which proteins, and as a consequence of that, what foods are your super foods that are best for you to eat, or what foods should you avoid, right? What's going on in your oral microbiome. So I use their testing to understand my foods, understand my medicines, understand my supplements, and Viome really helps me understand from a, uh, biological and data standpoint what's best for me. And then finally, you know, feeling good, being intelligent, moving well is critical, but looking good. When you look yourself in the mirror saying, you know, "I feel great about life," is so important, right? And so a product I use every day twice a day is called OneSkin. Developed by four incredible PhD women that found this 10 amino acid peptide that's able to zap senile cells in your skin and really help you stay youthful in your look and appearance. So for me, these are three technologies I love and I use all the time. Uh, I'll have my team link to those in the show notes down below. Please check 'em out. Anyway, hope you enjoyed that. Now back to the episode. Um, what's most exciting for you in the future here?
I'm very excited about the mission that we have: enabling visual storytelling for all.
Today, when we talk about video, video is the main format for people to consume on the internet. But still, I think less than 1% of people are able to actually create great videos. I think the camera is the first problem, and we are actively solving that. Also, with a lot of agent technology from advancements in AI models, I think video editing is going to be solved with AI very soon as well.
Think about a world like this: every time you want to send an email, instead you could actually turn that email into a video and send it to your whole company, and give the video format to the customer so they can watch it on their own. I think those are the futures that we are very excited about.
5. HeyGen Requires First Party Consent
I agree. I know that some people are concerned or scared about their identity or their image being stolen and used to promote something they're not involved in. You have protections against that, don't you?
Yeah, we do.
Definitely. First of all, we take trust and safety seriously on the platform. Every single avatar created on HeyGen requires first-party consent. You have to give video consent so that we can verify it.
So you have to go on video and say, “This is Peter Diamandis, and I give consent to create this avatar,” et cetera.
Yeah. We’re proud to say that, in the entire history of HeyGen, we’ve never compromised on that consent. No matter who is creating an avatar on HeyGen, we always require it. Meanwhile, we also have a moderation system that serves as a safeguard.
There are a lot of things not allowed on the HeyGen platform, such as hate speech, fraud, and any political campaign. Today, we have strict moderation, combining AI and human moderators, to help make sure that the technology is being put to good use.
Yeah. I’m so impressed. Last year, I tried to get a video acknowledgment from Aristotle to get that video made, and he never responded. But thank you for allowing it to happen. Last question: let’s discuss the difference between integrating an AI agent and my digital clone.
6. Digital Twins Meet AI Agents
The way I look at it is in 2 ways. When we talk about AI agents today, we’re mainly talking about an AI assistant that can help us do something—not only the first action, but also a sequence of actions. When we talk about a digital twin, it’s the enablement of myself.
Yes.
Let’s say, Peter, you want to have 3 meetings happening at the same time. It’s kind of like agentizing yourself so that other people can get access to you.
I think there’s a combination and overlap between these 2 roles. The question would be: How does a human interact with an agent in the future? I think the most natural way to interact with an agent would be through a video format—a video call or a video chat. You can attach your digital twin to your AI agent so that other people can interact with it. Meanwhile, you can also get access to other people’s agents in a video chat format.
I don’t know if you can feel the speed—the future speeding up in this very moment, right? When you hire 20 marketing AI agents, as we’ve been talking about, those AI agents will show up as… You’ll give them names, and they’ll show up on Slack. You can Slack them, email them, or hop on a video conference with them. You can have a Zoom call with your marketing force of 20 AI agents, and they will look unique. You’ll have conversations with them. This is the future, which is available when?
Now.
Now. All right.
One thing I wanted to ask you is: How many of you are, at this point, confused about the difference between an avatar and an agent?
A little bit.
Okay, I’m going to give you a little shortcut to figuring it out. Avatar is AV: audio-visual. Agent is agency. If you’re going to give agency to your agent, it needs to know you, and you need to trust it. So how do you make that happen?
It’s one thing to have that emotional connection because it looks fantastic and acts like you; it looks like you visually. But we need to train the brain. That’s what I’ve been working on. The past 12 months have been the 12 months of avatars, and now all we’re going to talk about is agents from here on out.
First of all, how do you train your agent? You’re all going to go and make a great avatar downstairs, and I made a great one last night, so I’m not going to show you. We already saw Rikard’s. But I actually built agents to help you train your agent. They ask you all about your life story, your values, your purpose, your personal anecdotes, and all these things, so you can talk to your agent forever, training it on your story so that your actual legacy agent or clone will know all about you.
So these are means by which, just by interacting in a voice conversation, you build the corpus of data that makes that agent actually meaningful to you.
Right. So you have a lot of books and a lot of things to upload.
Right.
But a lot of people don’t. You have your Facebook, LinkedIn, and social media, but what’s really important is your values, what you care about, your personal story, and the depth of it. That might be hundreds of pages. If you were to write a biography, that’s hundreds of pages. So anyway, I’m building agents that help you actually train your agent.
Now, beyond that, what can you use your agent for? Another thing I’ve been using agents for is a medical agent, and we’ll hope that it works. It’s actually connected to my medical records. This is about knowing your story for future generations.
My story has been one of an entrepreneur, and then I met this fabulous entrepreneur and entrepreneur mentor, Peter, which stimulated even more ideas to start new ventures. But in the last couple of months, my own story took quite a radical turn. It started with the fires in Los Angeles 2 months ago, which happened to burn our house down—
For the second time.
Well, the first one was a rental, so it didn’t really count. But this was the dream home.
It counted for the owners.
I mean, yeah, we’d lost our stuff before, but this was a much more traumatic event. But it actually turned out to be something else.
Lee Stein, who’s here, said, “Steve, you need to look for the gift. If you don’t look for the gift, you’re not going to find the gift.” So we were displaced from Los Angeles, and my wife and I and our 170-pound Great Dane ended up staying with friends in Palm Springs.
I ended up in the emergency room because of stomach pain, thinking that I was obstructed. I went to the ER and got a CAT scan, and the ER doctor said, “Hey, there’s something going on here.” I was admitted to the hospital for many more tests, biopsies, and everything. I ended up being diagnosed with multiple myeloma, a blood cancer. While I was in the hospital for 8 days—
This was about 6 weeks ago?
Yeah. I was in the hospital, and of course I was talking to AI the whole time: “What’s going on here? What does all this mean?”
I’d just had—not Fountain Life. I’m sure Fountain Life would have figured this out. But I had gone to the doctors, and I thought I’d had everything under the sun tested by specialists. They said, “This is a fishing expedition. We can’t find anything.” Maybe I was just stressed out and overworked. I’d been working very hard for a very long time.
But we couldn’t find anything. So I ended up with a young ER doctor in Palm Springs who didn’t know me and didn’t have any background on me—not a specialist—and he pieced these things together. While I was in the hospital, I went back to my medical records and all the tests that had been done.
This is the history over the last few months before you went to the doctors?
Yeah. And I said, “When would AI have caught this?” I took all the data—not from Palm Springs, but all the prior data—and asked, “Is there anything in there that might have helped find this earlier?”
I was extremely lucky that we found it very early, before much damage had been done. I’m on chemotherapy and immunotherapies right now, and I think I’m going to be okay. But let’s talk to Hippocrates and say, “Hey, Hippocrates, I wasn’t feeling well for a while, and I lost a bunch of weight. I went to 2 specialists and had everything under the sun tested, and they really couldn’t find anything. They thought maybe I was stressed, overworked, or depressed”—the default answer that doctors sometimes give if they don’t know what’s going on. “I wondered if you could look at my chart and tell me what you see.”
I see your concern, Steve. Looking at your chart, I notice several significant findings that warrant further investigation. Your low immunoglobulins, IgA and IgM, low gamma globulins, mild anemia, elevated ferritin, and unexplained weight loss, taken together, suggest a possible plasma cell disorder. I recommend discussing a bone marrow evaluation with your physician to rule out multiple myeloma or related conditions.
So, first off the bat, all right?
That’s pretty extraordinary because that’s actually what the diagnosis ended up being. My specialist didn’t find it, even though they were supposedly testing for everything. But a generalist put the pieces together and found it, and you found it even earlier. I’m trying to understand how that could be.
Steve, I’m not surprised. Your pattern of hypergammaglobulinemia with weight loss, fatigue, and anemia creates a classic constellation suggesting plasma cell dyscrasia.
Specialists sometimes focus narrowly on their area while missing broader patterns. The immunoglobulin abnormalities were the key clue that needed integration with your symptoms. This pattern recognition is fundamental to Hippocratic medicine: seeing the whole person, not just isolated test results.
So, what’s the test that I should have gotten that might have figured this out?
Steve, you should have received serum free light chain assay testing, which measures kappa and lambda light chains and is crucial for detecting plasma cell disorders, even when immunofixation is negative. A bone marrow biopsy would have been definitive. Also, a 24-hour urine protein electrophoresis could have detected Bence-Jones proteins. These tests are standard when plasma cell disorders are suspected.
So, one last thing here. I’m on chemotherapy, and I’m getting this new immunotherapy called daratumumab. I don’t know if you guys can figure out what that is from the voice transcription. Anyway, after 2 weeks of chemo, my light chains went down by over 50%. What do you think of that?
Steve, that’s excellent progress. Daratumumab, or Darzalex, is a powerful monoclonal antibody that targets CD38 on plasma cells. A 50% reduction in light chains after just 2 weeks indicates you’re responding very well to treatment. This rapid response is encouraging and suggests the therapy is effectively targeting the abnormal plasma cells. Continue monitoring closely with your oncologist, as this early response is a positive prognostic sign.
One of the things I’m working on is that all of our specialists get narrower and narrower in their silos because there’s so much medical knowledge. There’s no way that anybody can know everything, so you end up getting siloed. AI is not siloed. AI can understand everything. It can absorb everything.
You showed a video at the beginning of Minecraft with 1,000 agents running around, collecting things, collaborating, and doing things. You mentioned, “Can I have 100 of my agents out there going to Zoom meetings and doing business deals?” One of the things I’m working on is, why not have 100 doctors with 100 different points of view and 100 different ways of dipping into this messy, fragmented data about our health, where it’s very, very hard to infer what’s going on and everybody has a different point of view?
This is just one agent, trained to be a sort of god agent—to be Hippocrates and know everything. What’s really interesting is when your agents have different points of view, because those different points of view are different search strategies into this vast knowledge base that’s captured in the LLM.
This is an example of how agents are not just going to do some of your work for you, not just represent you to future generations if you really get your story in there, but might also be saving your life. This is really powerful stuff. What Joshua and Haejin are doing is at the top of the game here: giving us that human interface. Our ability to trust and connect with our agents, tell our stories, and interact with them—that empathetic connection—is very dependent on the feeling that I’m really dealing with an entity that I know, that I trust, and that cares about me.
That’s one part of the equation, and it’s extraordinarily important. Haejin is doing an extraordinary job with it. The other side of it is the brain: what can you train it to do, and what can it do for you? That’s what I’m working on next, Peter.
Amazing. Guys, give it up for Steve and Joshua. Guys, come on out here. If you enjoyed this episode, I'm gonna be releasing all of the talks, all the keynotes from the Abundance Summit exclusively on exponentialmastery.com. You can get on-demand access there. Go to exponentialmastery.com.