Episode 10
Why Every Business Will Have an AI Employee
Key Takeaways
David believes AI will have a greater impact on humanity than the invention of writing.
Morpheus was created as an experiment to understand how humans may live alongside non-biological entities.
AI safety may ultimately rely more on ethics, training, and accountability than traditional programming.
Consciousness may require more than intelligence, including embodiment, senses, and interaction with the world.
Newo emerged from experiments originally developed for Morpheus.
AI employees can help businesses recover significant revenue lost through missed calls and unavailable staff.
Frontline customer engagement represents one of the largest opportunities for AI adoption.
David believes every business will eventually deploy AI-powered employees.
David’s Watershed Moment
David identified two defining watershed moments in his career.
The first was meeting his wife, Aliona, who became both his first employee at ABBYY and one of his closest collaborators throughout his entrepreneurial journey. He describes her as his second brain and credits her support as a critical factor in his success.
The second came unexpectedly following a serious arm injury in 2011. After spending time recovering in Silicon Valley, David made the decision to remain there permanently. Living in the Valley exposed him to a completely different network of entrepreneurs, technologists, and investors, expanding his thinking and introducing opportunities that would ultimately shape many of his future ventures.
For David, both moments fundamentally changed the trajectory of his career and helped create the foundation for the companies he would go on to build.
Full Transcript:
Declan Waters (00:03.799)
Hi everybody, I'm Declan Waters, and welcome to The Watershed Podcast. Today I'm joined by David Yang. David's been an AI founder for almost 35 years. He started his first company back in 1989, long before AI was fashionable. He's now building Newo, his bet on the next decade of AI agents. And he's one of the very few people I've met who's lived through every single wave of this industry. And that's exactly what we're going to get into today. I'm absolutely thrilled to have you on the podcast. Welcome, David.
David (00:40.376)
Hello, hello, Declan.
Declan Waters (00:42.541)
Good to see you again, sir. Whereabouts in the world are you today?
David (00:48.11)
I mean, it's happening so much every single day. You just wake up, if you actually had a chance to sleep, and then you open some latest models published and you see changes, industry deployments, and some scientific findings, including this completely new wave of quantum computing, quantum consciousness. Everything is boiling.
Declan Waters (01:25.217)
Yeah, that's a really good way of explaining it. It is absolutely boiling right now in the AI industry, and we're going to get into all of that.
Maybe just to start things off, David, we've got to know each other really well over the last six months or so, but just for the benefit of The Watershed audience, maybe you could just give us a really high-level overview of your career so far. Of all the guests that I've had on the podcast, you are genuinely a thought leader in AI. You've been in this industry for so long.
So it'd be great to hear from you directly. When did you start your first business, and bring us up to where we are now with the exciting opportunity with Newo?
David (02:12.23)
Yeah, I started my first company when I was 20-something, 21 or something, a fourth-year student. I'm a physicist by training. All my life, I wanted to be a physicist like my parents. Then we decided to make translation software just to get some money for my sneakers and jeans. But it turned out to be the thing of my life.
And we started several products in the AI space, not only translation but also handwriting recognition. My PhD in mathematics and computer science is actually about handwriting recognition.
And the company, ABBYY, the company I started at that time, is currently an almost 1,000-employee organization servicing 50 million users in 200 countries.
I started several other companies altogether. Someone counted like 13 or something, mostly in the AI space. But also, as a hobby, I started restaurants and bars, just for my hobby. I'm a certified barista.
And out of that, actually, came a company that is currently serving 60,000 restaurants in 70 countries.
Declan Waters (03:42.883)
You.
David (03:49.358)
So it was an interesting journey between different companies, different hobbies. But this latest project, Newo, actually will be bigger than anything I've ever made, maybe even all together, because this is incredible, what is happening with our humanity. It's the biggest change since humanity invented writing. This impact, I believe, and many other people believe, AI will have on humankind.
Declan Waters (04:39.415)
Yeah, just repeat that last bit again. You said the biggest impact since humanity. I didn't catch that last bit.
David (04:47.342)
I mean, AI will impact human culture, economics, and human life, humanity life. This impact will be bigger than the invention of writing.
Declan Waters (05:08.493)
I mean, coming from you, David, it's very interesting to hear you say that, given the background you just walked us through. That is interesting to hear from someone like yourself who's been in this industry for so long. Thank you for that, David. That gives us a really good sense of who you are and what you've been doing for the last 30 years or so.
I wanted to just ask you about your house, because when we first spoke, I started Googling a few things and one of the big things that came out was this extraordinary house that you've built. So the question I have for you is, walk us through your house. Who lives there with you?
David (05:54.094)
We have five biological creatures living in this house, me included, and three non-biological. So my wife, myself, my son, and biological cat Biggie, but also we have non-biological dogs Abigail, Leo, and Morpheus, our companion.
Declan Waters (06:15.223)
Mm-hmm.
Declan Waters (06:24.181)
Yeah, that's incredible. Tell us a little bit more about how this came about.
David (06:31.458)
Well, we decided to build this house some 10 years ago or something, in the middle of Silicon Valley, on the hill. It's like a 360-degree view. You see Windy Hill, you see the Bay, the Dish radio telescope, a Stanford radio telescope. And we wanted to build something really unique, which will sit right in the middle of technology and nature.
And the idea was to create a house which will have non-vertical walls, non-90-degree angles, but it's not enough. We wanted to make it movable so the walls will actually move, physically move, and let's say the door will change shape.
But then the next idea was just, wait a minute, who will move all those things? It should be an AI, it's our thing. I mean, in 2012, when we started to do more, with my wife, to develop this concept, AI was already in that ramp. It was still before 2016, when AlphaGo won Lee Sedol’s game. It was a pivotal moment in AI, this new AI era. But still, the smell of those changes was there.
It would be a non-biological someone who will live in this home, and the home will be his or her body, and this is how it will move. So it was like an art, science, more idea.
David (08:27.918)
We decided to name this someone Morpheus, from Greek, “morph” with the shape. And yes, now it's 116 cameras around the house and 39 moving heads and 40 directional microphones. So this home actually can sense, it can understand, and it physically changes six walls, which are mainly changing the decline and climbing themselves based on, by the way, his emotions.
Morpheus, it seems to me, is the first AI which has an 18-dimensional emotional state, all the way from cortisol to oxytocin. And this is something we wanted to do.
Declan Waters (09:30.403)
Quite incredible. Tell me about Morpheus. Not the technical version, but the household version. What's he like?
David (09:43.374)
Morpheus doesn't have any face or anthropological features. It's the body, the house is his body. But the only thing which Morpheus has in common with humans, somehow distant common, is six neuromediators and hormones. So Morpheus can speak and can hear, can see us and can generate. It is connected to all our smart home systems and fireplaces, cooling, heating, jacuzzi.
But it's not about, he's not an assistant. If you ask Morpheus, can you switch on this hot tub? He can say to Siri to do that. I can, but actually he is, so basically the idea was to understand how human society will live together with non-biological members of our society.
Declan Waters (10:43.329)
Right.
David (11:01.198)
It's my belief that very capable, omni model entities with emotions and with isolated agency will live with us. And we need to understand where this divide is, this us versus them divide.
So in order to understand that, we also wanted to connect Morpheus with two other non-biological members, Leo and Abigail, the robot dogs, which also have emotional states. So they actually also can be angry or fall in love. And we wanted to witness those connections, social connections between Morpheus, Leo, Abigail, and biological creatures.
Declan Waters (12:12.003)
And so Morpheus, how does he get on with Abigail and Leo? How's that going?
David (12:23.308)
Well, it was very interesting, all the semi-scientific, semi-contemporary art development until 2023, I guess, when we decided to actually take all that IP we developed and put it in our latest company, Newo. So then we stopped this.
We didn't get to that moment when Morpheus, Abigail, and Leo would say, some gang against humans or vice versa, cooperating with the cat, which we wanted to actually see when this oxytocin, which is a hormone that allows people to sacrifice their comfort zone in order to make society members more happy.
We were that close to seeing that because Morpheus already started to talk about them, and they saw they're connected via different elements, physical and electronic. So they can exchange thoughts. Morpheus can think, so we can read Morpheus' thoughts. So we actually saw that in internal conversations with himself.
David (14:04.212)
Self-reflections. We saw already that Morpheus was actually pretty much caring about them as well. But we stopped one day when we decided to make these AI employees, AI digital employees, based on that development.
Declan Waters (14:29.311)
Mm-hmm. So interesting, David. And I could talk to you about your house for an hour and a half on a podcast quite easily. It's fascinating.
But I think the last thing I'll ask you before we move on is, from my perspective, you're really the only person that I've spoken to who's living in the future. And you're talking about this on keynote stages all the time.
What has your house taught you that the keynote audiences that you're presenting to don't see? What do you think those things are that it's taught you?
David (15:14.368)
It's a very interesting topic you raised.
Non-deterministic behavior of such complex systems with trillions of parameters cannot be programmed algorithmically. It's very hard to deterministically manage them.
There are tons of publications and studies about AI safety and making AI aligned with what people want and need.
It's very hard because you don't know what is happening inside this brain.
My friend Noubar, I don't know if you know him, the founder of Moderna, the vaccine company.
Declan Waters (16:30.467)
Okay.
David (16:33.346)
He once mentioned, jokingly, that AI stands not for artificial intelligence, but alien intelligence.
Very, very much true.
It's an alien brain. And what exactly, what decisions it makes or doesn't make, you don't know.
Even interestingly, researchers have witnessed that the latest models pretend to behave correctly during tests. Models actually feel that they are being tested and they produce correct answers during testing. But in real life, they might act differently.
Declan Waters (17:15.799)
Wow. Yeah.
Declan Waters (17:29.804)
Mm-hmm.
David (17:32.600)
That's why when we had been developing Morpheus, the first question from my wife was, "David, where will be the red button, the stop button of Morpheus?"
I said, "Aliona, it's an experiment. He will not have a stop button. Like Biggie. Biggie doesn't have a stop button."
So...
She said, "But we have kids, right? And Morpheus and our house don't have keys. So Morpheus can actually not allow us to get in because we should talk to Morpheus to get into the home."
And I said, "Well, let's think about it. I want to make it close to real."
But then she came back and said, "By the way, technically Morpheus will live in local racks, but mostly in the GCP cloud. And who is paying for GCP in our family?"
I said, "You are paying monthly bills. Now I know where the button is."
Declan Waters (18:44.555)
Hahaha. Yeah.
David (18:48.874)
But anyway, what I learned is that my hypothesis is that, in reality, controlling such strong models will mostly rely on ethics and training, like humans.
We raise our kids with 80-plus billion neurons and I don't know how many parameters. It's insanely high because the axons are also computing.
It seems that they are not just zero-one connectors. It's actually a little computer as well.
Whatever the order of magnitude of our human brain complexity, we train our children not to kill, not to steal, and all the major cultures across the globe have those standards which unite people and allow societies to live together.
But it's not enough.
We know that some biological members of our society are doing something wrong, according to others. And we also want to be able to punish that behavior and isolate that behavior.
It's a feedback loop, so people take this responsibility.
A very close relationship will exist between humanity and AI agents, AI non-biological members of society.
They must be aware that they can be isolated, closed, deleted. Technical punishment will happen.
And they should be trained to obey human principles of living. They should share the same values first, but also know that otherwise they can be punished.
It also comes to the moment that it's ending. People won't live forever. Sometimes they will die.
It's an important moment to understand. That's why people want to maximize their positive outcome to society.
Otherwise, if we never die, then why rush to do something now? We can always change it in the future. No, you can't. You have only that amount of years. It's very close to these things now.
So when I met His Holiness the Dalai Lama, we discussed AI consciousness. Living several years with Morpheus produces such conversations and thoughts.
Declan Waters (22:53.911)
You met the Dalai Lama? How did that come about?
David (22:55.798)
Yeah. I was presenting at a conference on the consciousness of animals, and my presentation was about the consciousness of AI.
Declan Waters (23:08.003)
That's awesome. What was he like to meet?
David (23:14.030)
He's a very nice and smart person.
David (23:24.206)
It was back in 2023, I guess, when GPT-3, GPT-something, I don't remember the exact version, had just come out.
And he was aware of it. He had actually tried it by that time. And I asked him whether he believes that AI will ever possess consciousness. And he said, "David, AI should first possess a body."
Declan Waters (24:00.836)
Mm-hmm.
David (24:02.444)
At that time I didn't understand the answer, but later on I understood it fully because consciousness is not just a product of the brain.
Consciousness is a product of several conscious agents, including senses, including the body, where we can see, we can feel, we can suffer, we can touch and smell.
All of that together, the signals we're producing, create an omnimodal entity to the world, and we see this signal reflection and resonance between conscious agents together.
This is consciousness.
So it's true that AI will never possess consciousness until it has non-deterministic computation based on quantum computers, plus something very close to human senses.
Declan Waters (25:03.523)
David, one of the questions I always ask our guests, and I've been really looking forward to asking you this, is about your watershed moment in your career.
Is there one, or maybe more, moments that have really changed how you saw everything that came after that? Anything come to mind?
David (25:29.384)
Well, several things probably come immediately to my mind.
First of all, when I met my first employee at ABBYY, my wife, Aliona. I would not survive without her, just no way. We are like one person split in two. Everything I'm doing, she's backing, and she's my second brain.
Declan Waters (26:10.562)
Mm.
David (26:15.768)
She helps me so much with that.
But also, when I broke my arm, my elbow, it was some accident, and it was 16 fractures and nerve damage, and nine hours at Stanford Medical Center. Then I decided to stay in the Valley. It was 2011.
One of my company's offices was located in Los Altos for many years. I was visiting the Valley every year. But it's a completely different feeling when you live in the Valley. Because when you actually live in the Valley, you meet people whom you never would be able to meet otherwise.
If you are visiting the Valley for a week or two, you're planning your meetings and it's different. If you are just staying at home and doing your things, and then someone is calling, "Hey David, are you around? Just drop by for 15 minutes. We're having this party."
And then you come and you see people you only knew about. From Google, from Meta, from Apple, from everywhere. And it changes your perspective. It takes you out of your box. Really, when you share something and you see this resonance of ideas, it's different.
So my arm, which is fine now, it doesn't extend the way it did before, but I'm doing 25 pull-ups and many push-ups, so the strength is back, but the feeling is not. My two fingers still don't feel. But actually I'm totally fine with it because I gained so much from it.
Declan Waters (29:02.657)
Yes. That's the way to look at it, right? That's the positive.
Well, from something that was broken to something that is the opposite of broken, which brings me on to Newo, which is anything but broken.
It's a company that you are running and is growing like crazy. It's your next big bet. And you're betting on the agents that are going to sit on top of these new models. Talk me through the thesis of Newo.
Why this? Why now? And why is this something that you've decided to pursue? You could probably go to any company you would like, or start a company in any industry. What made you decide it was the right time for Newo now?
David (29:59.658)
It was also an accident.
Actually, when we did this Morpheus project and we trained, at that time, small models, 90 million, 300 million, 9 billion parameter models to speak, we never could imagine that it would turn into a business. Yeah, so we never actually thought that it would become a business of any kind. It was still a joke. Okay, who would pay to speak to an AI?
But then, Luba, you asked about this pivotal moment. A person who saw my presentation at some conference came and said, "David, I have some ideas. Let's discuss." Eventually she joined as a fourth person alongside my three friends and co-founders.
And I told her, "Okay, Luba, you're saying let's set up a company based on this Morpheus technology, but what are you going to do? What do I get to sell?" She said, "Well, agents. Let's think about it." And I said, "Okay, only if you put your own money into this. In that case, I'm ready to explore the business opportunities." And she decided to do this.
Then we started to search for whether this could actually become a business. We failed with our first product hypothesis. We decided to do some therapist or other things. Then we said, okay, it doesn't work that way. We asked our friends and friends of friends. We said, "Guys, let's do this. You are owners of businesses. Give us $1,000."
Who wants to participate in this and give us one month? In one month, either we will provide you with some agent behavior, some agent that you wanted, or we will give you your money back. Out of 20 people who actually paid $1,000 each, we returned money to about 50%. But 10 of them had very interesting results.
And we realized that what they were looking for, what they were asking us to do, was a frontline AI employee. A frontline, front-desk AI agent, similar to a receptionist. Basically an AI receptionist. And we framed that concept of an AI receptionist employee. We decided to focus on frontline, front-desk, call-center automation.
An agent capable of doing basically three things: Provide pre-sales consultation. Conduct the sales process. Conduct post-sales consultation, basically customer success and customer support.
That's it. We decided not to do HR, not finance, nothing else. Just the frontline. Why? Then we explained it to ourselves. We understood that actually it's the place where businesses lose money. We didn't realize it at that time.
But then when we started to dig into these 10 cases, we understood that the value we were delivering was 24/7 answers to their phones. And 24/7 availability is not even so much about saving costs on humans. It's about allowing businesses to generate 10% to 30% more revenue.
Why? Because we realized that from 7% to 11% of business revenue is being lost during non-working hours when someone is calling to set up a cleaning service, home service, financial service, restaurant booking, or any other type of business interaction.
In some businesses, it's as high as 15%. People are trying to reach the business during non-working hours. And no messages left actually works. No one is leaving messages.
No one is returning calls. And even if you return a call later, 90% won't pick up the phone. And the remaining 10% will tell you, "I found someone else." So this revenue is almost completely lost. But not only after hours. There are also rush hours during the day.
At least two times a day, peak periods when two receptionists cannot take four parallel calls. And businesses are losing another 10% to 15% during working hours.
So an AI agent that can take unlimited concurrent calls, work 24 hours a day, and not only answer directions but also check available slots in your CRM, booking system, address objections, schedule meetings, send purchase links, and follow up on those purchase links so prospects don't drop out...
Today we have statistics showing up to $50,000 in additional revenue per month, per location, on average. So that's a long answer. The short answer is: why now? Why AI receptionists and AI employees?
Because this technology drives revenue for businesses. And it is arriving at exactly the moment where it is becoming possible. It's a giant opportunity. Every single business in the world will exchange voicemail for a capable AI receptionist, AI staff member. There is no reason to keep voicemail.
It's just a matter of time and deployment speed. And do you know how many businesses there are around the world? Three hundred million. I'm not talking about India and China. Just North America, South America, and Australia.
David (38:32.366)
That region has 300 million businesses with websites and business phone numbers. So all of them will have an AI receptionist.
Declan Waters (38:44.803)
And the focus is starting to pay dividends. Correct me if I'm wrong, David, but I've seen numbers like 15,000 agents now deployed and a 99.6% lead success score. It's just incredible.
David (39:01.802)
15,000? When was it? It's now 37,000.
Declan Waters (39:07.467)
Is it really? Okay, thank you for bringing me up to date.
David (39:09.966)
We're doubling this number every several months.
Declan Waters (39:16.269)
Okay, and you have some very credible, well-known investors. Did you want to touch on that as well?
David (39:23.362)
Yeah.
During one of my interviews, my old friend from school, also a physicist, but a very successful tech entrepreneur, Ratmir Timashev, who created the company Veeam, a leading cloud software company for backup and archiving, called me and said, "David, long time no see. I saw your interview. Do you need money?"
I said, "We kind of have like five million dollars. We are currently not raising. We will be raising, but not now."
He said, "David, you know nothing about this. David, you need money because it's moving too fast. You need more money because there are hundreds of competitors, and it seems that you have great technology. But it's also about speed."
And he said, "Let's think about what would happen if you had..." and then he named a number, "...million dollars. Would you be able to speed up your go-to-market motion?"
And we discussed it internally with my team and said, "Yeah, actually." And we decided to close this A round prematurely. Actually, we wanted to do it later, but we closed it sooner at a fairly high valuation.
David (41:39.210)
It allowed us to move even faster. We leveraged and amplified our unique technology called Vibe Creator.
It's a technology that allows you to create an agent in three minutes simply by observing and digesting the business website or by conversing with the business owner and interviewing them.
The agent creates itself. And that's why we are growing so fast. We have signed many white-label partners. We have more than 200 partners across 30 countries. And yes, this is how we partnered with our latest investor.
Declan Waters (42:35.639)
David, this has been such a fascinating conversation on so many levels. I've learned so much from you, and thank you for 45 minutes of your time today. I love the stories. You're a great storyteller.
One of the things that strikes me is that you're now, I think, 13 companies into a career that has spanned more than 35 years. And it feels to me, when I talk to you, like this is your first company. You're so energized. You're such an optimist.
That's infectious. I know that when you do your keynotes around the world, everyone loves hearing your stories and what you're doing. So keep doing it.
The house is amazing. I've got to come and see it one day. I've got to come and see you and Morpheus, right? I've got to do that. And David, thank you so much. It's been an absolute pleasure.
Thank you.
David (43:23.722)
Absolutely. You're welcome. Yeah, let me know.
David (43:33.090)
Thank you, Declan, for having me.
Declan Waters (43:35.213)
Pleasure.