Michael (00:02.838)
Welcome back to another episode of Adventures in Machine Learning. I'm one of your hosts, Michael Burke, and I do machine learning and data engineering at Databricks and I'm joined by my cohost.
Ben (00:12.607)
Ben Wilson, I debug integration tests at Databricks.
Michael (00:16.422)
And he's so good at it. It's, it's just mind blowing. So today we have a guest that I'm very excited to speak with. His name is Paul Allen and he is the co-founder of ancestry.com and ran the company from 1997 to 2002. He then founded familylink.com an organization focused on sort of family social networking. And more recently he moved into the personal productivity space, working with Gallup's strength finder product.
And if you're familiar with sort of Myers Briggs or any of those personality tests, you can think of strength finder as a personality test for your strengths. And now currently he's CEO of soar.com, which leverages machine learning to provide feedback that boosts organizational productivity. So Paul jumping right in about a decade back, you shifted from genealogy to productivity, and so I was wondering, was there a catalyst for this transition? Or have you always had an interest in.
Paul Allen (01:15.874)
It's a great question. I think that the thing that led me into Gallup and into what you call productivity was my background as a pioneer in the internet on creating viral products, engineering viral loops, designing products that would spread from user to user to user. Prior to Gallup, my teams had three successes. The first one was MyFamily.com, which we launched.
in December 26th, 1998. And it was designed to go viral. It was a free social network for families. And we advertised it to all of our ancestry users. It turns out that 78% of our ancestry researchers wanted a place to share their genealogical discoveries with other family members, their children, their relatives. So we set up myfamily.com. We invited the genealogist of the family to start inviting.
all their relatives, their extended family to the site. And it grew by a million users in the first 140 days and then started growing by 30,000 users a day. And this was pre-Google, pre LinkedIn, pre Facebook. So we had a organic, we called it genetic marketing as opposed to viral marketing, but it really was viral marketing. And then ancestry.com had some viral elements as well. The biggest viral experience that I had was in 2007.
when I attended the Facebook F8 conference and met Mark Zuckerberg the day that they announced Facebook platform. And FamilyLink, as you said, Michael, was building social networks for families. We kind of, I left Ancestry and my family and decided to try it again as a separate company. And I got on the phone with my head of product and head of engineering that day that I met Mark Zuckerberg and heard his announcement about the Facebook platform.
And I said, we're not going to build websites for families. We're going to build apps for families on top of the Facebook platform. So by October, 2007, we launched an app called we're related. The product manager was Jason McGowan, who's the founder and CEO of crumble cookies, which is the fastest growing food cookie business in the world. I think they did over a billion dollars in revenue like two years ago. And he'll be a multi-billionaire, a fantastic guy. And we're related got.
Paul Allen (03:37.598)
a million users in 29 days using the Facebook viral channels and then started growing by a million a week. So by 2010, we had over a hundred million users and Disney came to us and started selling all of our ad inventory because we were the most family friendly app on Facebook. It was insane how viral Facebook's platform allowed you to be until they kicked off all the apps and shut down all the APIs and basically.
destroyed our business along with a hundred thousand other app developers. So we lost all of our revenue, hundreds, 750,000 a month in revenue. But the ancestry, my family, and we're related viral loop creation experience is what caught Gallup's attention because as you said, StrengthsFinder is a, it's not a personality test. It's a test, a psychometric analysis to determine your top talents, your potential for strength.
It's a brilliant, positive psychological assessment. And they wanted it to go from, you know, millions of users to a much bigger number. And so my friend, Clint Carlos, who worked at Gallup said, I know the guy who can help Strengthfinder go viral. And so that was the thread. That was the, you know, track record of three viral applications led me to join Gallup for five years. I loved every minute at Gallup.
Strengths finders now grown to 30 million Assessment takers and their goal is a billion in John Clifton's lifetime So that was it and in the five years at Gallup. I felt like I got a postgraduate degree in organizational psychology and Productivity leadership culture Gallup is the world's best I think at discovering what makes people tick what makes organizations tick
They kind of created the idea of employee engagement surveys decades ago. They are a brilliant, amazing, pioneering company. And so I was so blessed to be there for five years. And now my goal in life is to help every human being reach their full potential. Number one, by discovering their talents and strengths. And number two, by organizing their life and implementing systems and tools, including AI, that will help them reach maximum fulfillment.
Paul Allen (06:00.062)
at work and an optimal flourishing life outside of work. All the things that matter in Gallup's global wellbeing study can now be brought about with the assistance of technology, AI. So I'm having a blast. I'm a lot older than I was 26 years ago when I started Ancestry, but I feel like the AI wave is so much bigger than the dot com web.
you know, wave that I caught in 95-96. So it's thrilling to be here. And I love the title of your podcast, Adventures in Machine Learning, because my goodness, what an adventure it is right now. Every day, the story is changing.
Michael (06:45.022)
Yeah. And speaking of machine learning, uh, before we get into that topic, I was wondering if you could elaborate a bit on virality, sort of how you define it and how you think about creating viral content or viral products in general.
Paul Allen (06:59.618)
Great question. I see virality as inherently designing a product and a user experience using that product where getting other people to use the product is an innate part of the need and opportunity of the software. So myfamily.com would be a very lonely place if there was one person posting photos and updating their calendar. It wouldn't be fulfilling at all. But when there's 20 people sharing photos and calendar and...
Anniversaries and birth dates and notifications are going out almost every day about some special thing that's coming up that becomes a little mini Community and engaged social site now what we found at my family comm is that you would start your nuclear Family site and then your extended family site and then your spouse would say well I want one of these for my family too because each of us belongs to more than one quote family You've got your nuclear family your extended family
Maybe you're part of a surname genealogy group of, you know, all the, all the Willsons out there, by the way, I have a lot of relatives that are Willsons. And so we might, we might be related. Who knows? So, but yeah, virality to me is the product works better as you share it with other users, Slack is a great example because anyone in an organization can become the first user of Slack and then pretty soon they share it with a few colleagues. And pretty soon.
Ben (08:07.123)
Me too.
Paul Allen (08:26.722)
The CIO is alarmed because so much company data is going through the free version of Slack and now you have to upgrade to the premium version. So for me, virality requires a free version to be really effective and then different tiers of premium on top of it, but you know, clubhouse was a great example of they, they allowed you to sign up, import your address book. And pretty soon you've invited, you know, 50 other people to join in and that, you know, each user.
There's a mathematical formula in virology called the K factor. You can call that the viral coefficient. And when each user spreads the product to one other user per month, you have a viral coefficient of 1.0, which means your user base doubles every month. And that exponential growth leads to really big numbers really fast. Now at Facebook, we actually had a coefficient that was really high.
until they started disconnecting all the channels. That's why we got 120 million users in two and a half years. It's much harder to do that organically, but a lot of viral apps will allow for phone address book importing or other social graphs imported so that the ease of spreading is built into the product design.
Ben (09:47.263)
So when you talked about earlier about productivity and the lessons that you've learned throughout creating all of these products that have been wildly successful, does your knowledge of human behavior and particularly workplace levels of
Ben (10:10.663)
of being able to get things done without overly stressing the team. Is that a by-product of viral success? I imagine from a team perspective, whether it being product or sales or engineering in particular, when something takes off and becomes wildly successful, everybody's really excited. They're like, well, we built this thing that everybody really cares about. This is awesome. But then you realize the next thought that you have is.
Oh no, we had to fix all that stuff that's broken. Were you kind of the person in the room, but also being a wallflower observing people's reaction to that and controlling the stress that brings?
Paul Allen (10:54.998)
Boy, that is the best stress you can have. It is really exhilarating to see something grow. The Bible says charity covers a multitude of sins. And I think in the workplace, revenue covers a multitude of sins and technical debt, accounting debt, legal, all the shortcuts you took to get there. The revenue, which is often led by the viral growth engine, it just takes away. Now there's definitely stress and the need to read.
refactor code and build things that are more scalable. I remember when ancestry would go down for a few hours and I would calculate the lost revenue per hour. I mean, it was, it was very expensive to have a site outage, like tens of thousands or hundreds of thousands of dollars in a very short period of time. And so, yeah, there is the stress, but the joy and exhilaration of catching that wave supersedes it. Now I will admit that pre Gallup, I was oblivious to human nature and psychology.
And I assumed so much about people. I just assumed I projected my own interests and values and strengths on everyone on my team. I used to do silly things like my number one strength is learner. And I remember trying to motivate my team in the late nineties. And I thought, okay, what's the best incentive I can give if we hit our goals or if an individual hits their personal goals, I'm like, okay.
We'll let them go to any conference they want to all, all expenses paid, like a $2,000 Jupiter communications conference or whatever, and, or subscribe to any database or subscription, you know, very premium subscription. And guess what? Nobody cared. I'm, I'm wired to love learning. I love going to conferences. I love sitting at the feet of the experts and I love reading premium content that is just experts created it and it's exclusive and, and
because I'm a learner, but my team wasn't that way. And so at Gallup, I realized how badly I had misunderstood, misperceived, mismanaged most of my employees. In fact, my lifelong best engineer is still with me at SOAR. He's been with me for 32 years in eight companies. His name's Richard Stauffer, and he was the original Ancestry engineer. He built 34 applications as the only engineer at Ancestry from 96, 97, 98.
Paul Allen (13:13.206)
Within a year, we grew to 60 engineers, but two years later, we were still using 14 of his applications to run the company. He's a brilliant 0.9 version of anything in a week or two. He's just so fast. So about three months ago, I was working with him on GPT prompt engineering. You know, we got access to the GPT four APIs in March and GPT 3.5 blew my mind. But four was a hundred times better at so many things. It just completely changes the.
the everything about machine learning in the future. We used to do a geo and BERT, human training data, thousands of sentences, thousands of transcripts. We did a multi-month building of custom AI and ML models for sales and training companies. And now we can do that in a hundred to a thousand times faster with some really good prompt engineering. It's a game changer, but I found out from Richard who has Achiever as one of his top five strengths.
that he gets no satisfaction from tinkering. He only gets satisfaction from shipping code that day that will actually be used by people. He's been spoiled because for 30 plus years he's been shipping code iteratively. That's how we work before Eric Reese came out with lean startup. That's how all the dot coms that I know of were iterating, listening to customers. Like we all adopted that mindset. Eric Reese described it and
created a vocabulary and assistant for it and popularized the idea of Lean Startup. But all the successful.coms were doing it that way. And so Richard has been able to ship code almost every day for 30 years. He told me, Paul, I get zero satisfaction from tinkering or doing prompt engineering that doesn't affect the user experience. So now I've got other data scientists that I let do all the tinkering because Richard needs the satisfaction of shipping code every day.
And I only know that and Richard only articulated that because we both took the Clifton strengths, finder assessment, and we talk about it all the time inside the company. I am, I'm almost, I'm legally blind without corrective lenses. I just have, you know, really bad eyesight. My, my corrective lens is like minus 12. Uh, at one point it was minus 14 and a half, very, very thick glasses. So when I was the entrepreneur of the year in 2000 at ancestry.com
Paul Allen (15:36.234)
I gave a speech at an event and I thanked God for whoever invented glasses and contact lenses because without them, I wouldn't be able to navigate this modern world at all. I would have not survived really. And now I get to thank Don Clifton for a strength assessment that allows me to see the differences between people. If I know your top five strengths, I will work with you totally differently. If I don't know your, I will no longer mentor anyone.
if I don't know their top five strengths, because my advice will be wrong. It will be what I would do. And now with this lens of clearly understanding out of the 34 top strengths that Don Clifton identified, if I know your top five and if we can talk about it, it completely changes our culture. I'm so much more excited about source culture than I ever was at Ancestry or FamilyLink or any of my other previous seven companies, because I was unaware of even...
how to work with people or manage them or inspire them. I just didn't understand at all. And now I do. And it will be fun to see SOAR scale to thousands of employees and see if that culture of productivity and vision and focus can be accompanied by thriving wellbeing and not stress and despair and depression and anxiety. And one of the AI systems that we're launching soon is really exciting to me.
then I think it will be our effort to do what you're saying. How do you grow and scale without causing too much human damage along the way? Well, Chuck Conrad, the author of Game of Work, he's called the grandfather of gamification by Forbes Magazine. So he published his book in the 70s. The Game of Work teaches about human nature and how to motivate people in an intrinsic way.
so that their work becomes play. They get to keep score. They know if they're winning. Work can be more fun if you borrow some of the techniques of sports and even recreation. So bowling wouldn't be fun. Golf wouldn't be fun if you didn't keep score. And if you pretend there's no winners, everyone's keeping score in their head anyway. So Game of Work, we've partnered with Chuck and SOAR will soon be incorporating the Game of Work AI.
Paul Allen (18:01.306)
into the co-pilots that are being built for every employee in every role in every company. So there will be a retrieval augmented generative AI that is like a chat GPT that is tuned to your area of expertise. Let's say you're an internet marketer. We'll have thousands of the best articles, books and podcasts and YouTubes about internet marketing. Your chat GPT co-pilot, your co-pilot built by SOAR.
will be super, super knowledgeable of all the experts in the space. It'll be tracking what's being said every day, what's working, what's not. And as you engage with your co-pilot on a daily basis, it will be teaching you more and more about your field. But the game of AI concepts will be woven into that so that you can have it, help you keep score and remind you of your goals and experience the exhilaration of winning on a daily basis, a weekly basis, or on a quarterly basis.
So we're taking some of the knowledge that Chuck had about human nature and adding that gamification element to the future AI co-pilots that are gonna be rolled out in the next few months for any role in any company.
Ben (19:14.443)
That sounds so fascinating to me. From the perspective of somebody who at my core, I think when I was younger, I really loved the learning journey. Like I would get incredibly excited about, it's like, oh, there's this new cool technology. I just want to know how it works. I want to take it apart metaphorically or physically and understand how it's built and read about it. Like read about who invented it and why they did this way.
And over the last, I'd say decade of my working life, I still do that in my personal time. I love learning new things and geeking out and doing that type of stuff. But professionally, I've learned the hard way that due to my direction of my career, I've become like your coworker of 30 something years, which is I actually, I used to not care if what I built ever saw the light of day. I just liked building it.
And now at work, I don't care if I build something or learn something new. I know that I'm going to do that, but my only motivation is how many people are using this and what do they hate about it so that I can fix what they hate. And that's more like that achiever type thing. But I've noticed that particularly coming into a highly, you know, high velocity engineering team.
is that everybody around me has that same perspective, but their personality isn't the same. Like everybody's diverse and has different senses of humor. They come from different cultures, different places in the world. And when it comes to work, we're all aligned on what we actually need to do and what motivates us. You can tell the joy that the team has when we ship something new and it actually works and everybody's using it.
is the intention of some of these, you know, work co-pilots that you're creating and the, the direction of where you want to go with the company. Is it to guide people on how to best function within that job role that they're doing, regardless of how they are core as a personality.
Paul Allen (21:32.362)
100%. Yeah, that's you've said it really well. And by the way, your mindset that you've adopted over the years, it makes you so much more valuable for your employers. If you were if your company were filled with people who love learning and geeking out and tinkering, but don't care about shipping, the company wouldn't be shipping very well. And and the joy that you get from accomplishment would be missing. So it's cool that you have that now the way you described.
Ben (21:35.187)
Awesome.
Paul Allen (22:00.29)
your personality. At first, I thought you should lie down on the couch. We could psychoanalyze you a little bit more because I really liked what you were saying. But, but I, I heard a hint that you love to fix things. There's a Clifton strength called restorative. My wife has it as her number two strength and there's so much satisfaction from finding problems and flaws and fixing them. And I don't have that. I actually don't like problems. I actually like, you know, to just come up with new ideas and strategies and
And I have analytical in my top six, but I got so many themes that come to life every day for me in my current job. What we'd like to do is have AI remind you on a regular basis of the strengths that you have. Now, if the goal is shipping and revenue and success inside the company, you need to figure out which of my strengths are aligned with the task at hand. And can I?
consciously use those in a way that's satisfying. Now, there's a strengths book called Selling with Your Strengths. And it's designed to help millions of salespeople discover how you can sell. The goal is selling, but based on which of the 34 strengths you have, it will give you advice that will make you enjoy sales more. And if you do it someone else's way, you might not enjoy it. You might be successful, but you might be burning out.
And so I really think there's a beautiful potential for AI to understand all of the differences in personalities and talents, but also the objective of the company. As, as Peter Drucker said, the goal of any corporation is customers. You have to have revenue in order to even exist. And so if we're not attuned to that, we can become a cost center, not a value creator for our employer, but each of us has a different set of patterns of thinking, feeling, and behaving.
And when you put language around those and become conscious and aware of what makes you tick, you can actually get into flow state by doing the work you have to do in the way that best plays to your strengths. So when you can experience personal peak performance and flow and get the job done that your employer is requiring. Now, there are some jobs where you are not going to be a fit for. There might be a role inside of Databricks where if they put you in it.
Paul Allen (24:23.486)
It wouldn't play to any of your strengths and you would start experiencing burnout. And that's where conversations with managers and leaders around your strengths themes can be really helpful. But I also heard you say something that the book selling with your strengths also says, which is if your job only allows you to use, let's say two of your strengths at work, find a way to use the other three outside of work so that in your life, you're still flexing those muscles that you were born with or that you developed at an early age.
So I do think human flourishing in the context of modern capitalism is possible, but it's only possible if we go in with a lot more self-awareness and team awareness than we currently have.
Ben (24:59.775)
Hehe.
Ben (25:06.183)
Yeah, I couldn't agree more. I think tools like this are invaluable. What I, when I was just listening to your explanation of how to take your strengths and apply them as best to your work that you need to get done. It made me think of what if there was a meta sort of model that understood all of the strengths of a particular dynamic team and also has access to a work backlog so that you can prioritize critical things, but
When you're planning a sprint, you're like, all right, I've got these six people and they have this many hours that we can do in this two week period. And here's this list of all these tickets that we need to get done. Sometimes you kind of just guess at, oh, well, this person's worked on this part of the code before and they know this really well. You don't know though, did this person hate working on this? Was this the worst thing that they worked on six weeks ago and they wish that they never have to touch this again?
Paul Allen (25:59.426)
Hehehe
Ben (26:05.363)
because of the process that was involved in figuring this out or fixing this. Whereas somebody else who's never done it, they did something adjacently similar to this and they really enjoyed it. They loved getting it across the finish line and it gave them job satisfaction. Just the thought of having a meta model that has access to interaction history with these other agents to say, actually,
Michael really likes to do this. You should assign these three tickets to Michael. They're not all the exact same ticket, but they're, they're similar. There are things that we, it is believed that he will enjoy. Um, that sounds like, please, please get that in Jira.
Paul Allen (26:48.662)
Ben, that is absolutely brilliant, Ben. So what I said about Richard wanting to ship code and Frank loving to tinker and do prompt engineering along with Clint and several others on my team. Now I know what tasks in that case to not give Richard. But what you just said could be woven into the copilot. So the copilot is designed for you to chat with it at the beginning of the day, during the day, at the end of the day. And maybe what you're saying is...
Each time I finish a task and check in some code, the co-pilot asks me how I liked that project or, you know, was that fulfilling or it gives me a smiley or a, you know, frown or whatever, a little bit of a feedback that the AI then captures and then goes back into JIRA to whoever the manager is that's assigning the tickets. That's not that hard to do, but your idea is absolutely brilliant. It takes what we're already planning to a whole new level. So I'd love to send you a, an advisory agreement of Ben, cause, cause we need
We need more ideas like that. And to weave it into every sprint for every company that's, you know, practicing, uh, agile and lean. Uh, I mean, so many millions of tasks are being assigned every single day. And probably many of them to the wrong people, just because of the lack of insights about what makes you tick. Like maximizers hate problems. They only want to work on a project that is good and perfected. They like good to great maximizers. And that's a very strong.
Maybe 10% of the population has maximizer and maybe 15 or 20% has restorative. Restoratives get energy when they find flaws and bugs and problems and then fix them. They want to take something that's suboptimal and make it restored or good again. But the maximizers it's like kryptonite. You give a maximizer a problem and it's like, ah, just shoot me already. I don't ever want to do this. I don't want to come back to work again. And so the themes of CliftonStrengths.
Ben (28:39.475)
Mm-hmm.
Paul Allen (28:44.79)
combined with a task management type system like you're describing is very achievable. I can't wait now to tinker with it, with my team and start using it internally. All 35 people at SOAR are gonna be starting using our co-pilots in the next couple of weeks for our internal experimentation. And we're looking at the kind of content, the retrieval augmented generation that's required.
Ben (28:54.676)
Hahaha
Paul Allen (29:11.586)
for each role, we won't roll out all roles at once, but you take sales, what 10 or 15% of the public are in a sales type business development role. And a lot of them are retail. So AI might not be at a Starbucks, maybe that's not as effective. But for phone based remote knowledge workers that are in sales, getting a co-pilot for sales that can chat with you during the day and watch every call that you make and tell you how well you did, did you follow the training?
Did you ask the right questions? Uh, and that co-pilot will be, you know, chatting with you, but to your point in an engineering type thing where products are being built and debugged and shipped and supported, um, I think you're exactly right then that will be a fantastic addition to the, uh, the meta knowledge that you described that at the team level, we already have team grids where every strength of every person on the team is visible to everybody. And there's already an email composer that's like Grammarly.
Ben (30:08.168)
Mm-hmm.
Paul Allen (30:10.882)
that if I'm sending an email to one of my direct reports and they have restorative and achiever, I can write the email and then have it rewrite it to tune it to their strengths so that it will be received better by that person. Same with disc profiles. If you're in a sales context and you've got someone with a high D, low S and you're sending them an email, your email may be completely off base to their style of wanting to get information.
Ben (30:22.598)
Interesting.
Paul Allen (30:38.686)
And the AI can rewrite it to any disk profile so that it's attuning your normal, your, your overall message to resonate and land well with the recipient of that message.
Ben (30:52.219)
So if you, sorry, this is a tangent, but if we were to send a group message to 50 people at a company, you're saying 50 different copies of the same message, but it still contains the same. That's awesome.
Paul Allen (31:00.69)
You could personalize it to each one. Yeah. You could, yeah, you could. It could consain this, yeah, exactly. That is possible right now.
Ben (31:14.27)
interesting.
Paul Allen (31:14.634)
We haven't tested it in real in the real world to see if it has a backlash. Like I don't, we actually built something six months ago that could just draft an email to anybody in your company based on their strengths. And we tried it internally. And I'm like, I'll never use this because if it's not authentic and genuinely written by me, then it's going to seem robotic and any direct report of mine who gets an AI generated message, that just shows that I don't even care enough to write a message to them.
That's a really bad signal, but to prompt me to remember their strength and fine tune the message. Okay. That's doable. We can do that right now. But then the question is on a group message. Do you really want to do that? Or maybe you want to send the same message to all 50, but highlight certain things that you think will resonate with each person. I don't know. We're going to have to do a lot of experimentation and the goal of SOAR.
And we're doing this in partnership with SHRM, the Society for Human Resources Management. We're in a very deep partnership with them. And in October, some really fun products will be shipping into the world of work through the SHRM-SOAR partnership. But, you know, responsible AI and ethical AI are buzzwords. We would like to propose a new acronym for the kind of AI that ought to be embraced by workplaces everywhere. We call it Pure AI.
P stands for personalized. It knows your strengths, your values, your goals. U stands for uplifting. You actually get energy and excitement and fun out of using it. It's not depressing, coercive or manipulative. It's not psychologically fear inducing. It's a joy to use, okay? So personalized, uplifting, responsible, ethical AI. If there is a standard like a G2 or a Captera,
that can start grading and scoring all of the tens of thousands of AI systems that are shipping around how pure they are, how personalized, how uplifting. Then you end up having workplaces and HR leaders and C-level executives embracing the very tools that will increase productivity, but at the same time, increase the joy of going to work and the fun of working and being successful together. I think we're on the cusp of the greatest productivity leap in human history.
Paul Allen (33:35.246)
And there's the risk that the bad AI, the manipulative, coercive, extractive AI that is going to be launched by lots of companies that want to make lots of money, it will spread at the detriment of humans. And I think if the HR world and the C level executives realize that there can be personalized uplifting AI that's also responsible and ethical and that gets the job done, but it doesn't harm humans along the way. I think that is.
a really exciting cusp of history that we're sitting on right now.
Ben (34:10.747)
Yeah, I'm gonna need...
Michael (34:10.83)
Wait, I have a question really quick. How are you defining uplifting as a metric? That's a fascinating concept that everybody's scared of the lack of uplift, but how would you measure that?
Paul Allen (34:22.358)
Well, Gallup is the world's best organization for measuring how humans feel about things, whether that's workplace, whether that's politics, that they invented modern statistical sampling and polling. They have dozens of brilliant people that figure out how to ask the right questions. So I would say that the right answer is, did the individuals using it feel uplifted by it? So I don't know the right words or how many questions will be involved. But I think what we'll do is pre and post surveys.
of users of the SOAR SHRM AI systems to find out how uplifting was it. And uplifting is a word that SOAR has been using for five years. And so, and it fits into the pure acronym. So it might not be the perfect word and it might not lead us to the exact right questions, but I think when you use software, you know, whether you liked it and had fun with it and it was exhilarating and inspiring, or whether it was like.
a grind. Like, do you know anyone that likes their CRM? Do you know anyone that likes their Oracle accounting system? Like, why is it that the world has standardized on systems that $100 billion companies are deploying in almost all major corporations and everyone hates using that software? That is like, detrimental to human well being. I mean, Jeffrey Pfeffer at Stanford wrote a book called Dying for a Paycheck. And he studies
Michael (35:24.791)
I do not.
Paul Allen (35:46.486)
that what happens when corporations lay off large numbers of people in order to hit their financial goals and you know, stock goes way up, but he actually says there's a death toll every time there's a mass layoff because people die. Some people take themselves out, but many people have heart attacks or other fatal things that happened because they were unemployed for so many months. The shock of losing your job. And then the longterm.
detrimental effects on your emotional well-being. It can actually create a death toll. And he's measured that. So you could argue that every board in the world that lays off a thousand, 5,000, 10,000 people ought to be aware of the death toll of that decision. Is it worth a five cent increase in our stock price next quarter to kill 17 people? Like that's how Jeffrey Pfeffer mentioned it. Well, you could say the same thing about the slow toll
on human wellbeing that large systems used by millions of workers that all hate it.
That's a very different thing than a sudden layoff. But I think the human toll is staggering when you add it up. How many moments of agony and pain and distress are imposed upon millions of salespeople and HR professionals and attorneys that like to do this horrible, use this horrible system to accomplish the work that you have to do in order to get your paycheck.
Michael (37:23.974)
on it. That's, that's fascinating. And my initial sort of reaction was pre and post and sort of subjective measurements of uplift might not be super reliable, but I actually, as I'm thinking about it more, I like that we go and we actually ask the person and give them the power to basically choose whether they like something or not. Cause going behind the scenes and saying, you actually like this, even if you don't know it seems a bit big brother. Ask.
Paul Allen (37:51.19)
Very, very, very much so.
Ben (37:52.027)
Yeah, you would have to use explicit feedback, like implicit measurement for, does somebody like this API? It's like, what is the alternative? Like, yeah, of course they have, but if it's part of their core job, it's like, you have to use this. So when we implicitly measure that it's bogus data, but that would be interesting to get feed like explicit feedback from users of all sorts of things. And
Michael (38:03.254)
Usage.
Ben (38:22.239)
To go along with that, Paul, the idea that you brought up with regarding this utilitarianism associated with AI and products that are being built with these tools now is very interesting to me. I think there's a corollary to it with being able to go viral. Basically you launch something, people start using it.
it's so much better than the alternative or it brings so much joy or it, the measure that we've always had in the ML practitioners over the past several decades, when you're launching something that's backed by ML, our main goal is to reduce friction by automating away things that people hate or would be putting too much of a burden on a user. So we use ML to automate some of that stuff or make decisions on behalf of things that
people don't really want to do. And it's interesting to hear how you're talking about building products with advanced LLMs and things that can help people enjoy interacting with it. And it's interesting to me how the tooling that we're building in support of these things now also has that same goal of that's a great motivating factor for the uptake in usage of anything that you're building.
saying.
without using this, is that a worse experience than using this? I don't know. I'm getting all these ideas of like things that we should be measuring about. Like that's such an important aspect of API design or product design that a lot of times we do it by canvassing and asking people, what do you like about this? How much does it suck? What do you think we should change? But to bake that into an algorithm itself, the powers of product is, is a very interesting idea. And I think.
Paul Allen (39:58.338)
But, what?
Ben (40:21.02)
It should catch on.
Paul Allen (40:23.654)
So to your, well, yeah, there's just quite a few thoughts that are that are going through my head right now. One one thing based on what Michael was saying was that how do you measure if it's uplifting or not? There's a concept in organizational psychology called psychological safety, and it's become quite popular. In fact, many people say that you can't even use your strengths at work and be yourself unless you have a level of psychological safety where you can actually
take risks, say things that you believe without fear of reprisal. Well, one of the software design features that we're planning for all of our future AI is what we call incognito mode or psychological safety mode where you get to use your system, where your leaders and managers aren't seeing your grade or your score on the feedback after your first 50 calls. You get to practice, it's a training wheels mode where you get to start using the software.
privately without risk that someone's going to shame you or embarrass you until you've gotten to the point where you're Mastering the system or the implementing the training that your employer has invested in and now you're pretty good Or even really good and now you proudly share your zoom calls and your email Scores and all the all the grades and things like that with your managers So that's one thing that could be built into lots of software
psychological safety, incognito mode, training wheel mode, whatever you call it. The other thing that's possible in a world, let's say where a billion people know what their CliftonStrengths top five are, is that user interface designers could actually customize the same application. And when you log in, it knows that you're a restorative and not a maximizer. And so it emphasizes the features that will appeal to you and de-emphasizes the features that will be kryptonite to you.
And you could still go get that functionality, but it could actually adapt to your Wavelength that you're on and the patterns of thinking feeling behaving. That's a big concept An easier concept that michael and I were talking about right before we started this call was AI as a curation engine So my background prior to gallup and prior to ancestry.com was electronic publishing of cd-rom collections are
Paul Allen (42:46.602)
Our goal of our first company was to identify the best books ever published in every field of human knowledge And to index them and sell them on cd-rom in libraries of searchable collections To schools libraries and works workplaces and homes So my background was I started a master's degree in library science And I really would like to know the top books and lectures on every subject In history in the entire world and today
Knowledge is advancing so quickly that you have to look to podcasts and YouTube for the latest greatest things even before papers get published and books get published so if you have this AI that becomes super smart about the field that you're studying and Then you get transcription costs plummeting so that every one of let's say there's 20 machine learning AI Podcasts that you really wish you could listen to every episode, but you can't there's no time. We don't have a neural link implant yet
so that we can, you know, matrix style, download the whole thing into our brain. We don't have that yet. So could AI transcribe those 20 podcasts on a weekly basis and based on rules that you give it, tell you the clips that you really can't afford to miss and the clips that will be so inspiring or exciting to you. So for example, I have ideation as my number three strength. If I go to a conference, whether it's an HR conference or a machine learning AI conference,
There's a lot of speakers, let's say a thousand speakers at a three day conference, and I probably get to go here five or 10 of them, and I have to pick, oh my gosh, which ones am I gonna miss? There's only, oh, there's three speakers right now. I can't go to all of them, I have to pick one. Well, AI could go to the conference, transcribe every single talk, analyze the things you're interested in, and if I've got ideation, it would look for anyone in the conference who's shared a new idea.
Well, how do you know if it's a new idea? Well, number one, they might say that they're sharing a new idea, or it's an idea that the AI has not seen before. And that could be pretty interesting. And then I get a playlist, a personalized playlist after the conference for the next three months that makes sure that all the great stuff shared by those thousand people, that I didn't miss the stuff that I can't afford to miss. And I got clips of all the stuff that would be most exciting and appealing to my ideation. Okay, I think that's doable in the next six months.
Paul Allen (45:11.55)
industry by industry by industry, when you get conference organizers saying, we want AI to come in and add value to the conference attendees by having this post-conference AI curation tool that really brings all that thousand hours of content to you as you can, you know, budget time in the day to catch up on that.
Ben (45:37.315)
So, would this be user controlled or would this be sort of AI controlled with respect to let's say I'm a hobbyist of, I don't know, model rocketry for instance, this isn't true, but I'm like really enthused with that. But I spend all of my time watching YouTube videos and reading articles and blog posts about.
amateur model rocketry. I sort of insulate myself from outside knowledge that could potentially benefit my passion. Also just you're putting all your eggs in one basket for this one thing, right? I don't know. I would argue is not the healthiest way to grow your mind and knowledge of just keeping in this one specific thing.
Would a tool like this in an ideal world understand what content is being asked of it to produce and then autonomously broaden that in order to provide additional insight and knowledge and experiences to a consumer of this that would say, hey, I know you're really into model rocketry, but there's this other thing that you could also get into that'll make you better at that.
and also diversify your knowledge of your existence on this planet.
Paul Allen (47:11.302)
Once again, you have asked a question or shared an idea that's just absolutely wonderful. I love that so much, Ben. I think it's gonna be a combination of number one, all AI in my opinion ought to be controlled by the user of it. It ought to be the rules, the guidelines, the guardrails should be, the agency ought to be in the human who's using it, not the AI itself. That said, Gallup has this.
global well-being study. They've actually been studying for decades what makes a flourishing, thriving life. And they measure 98% of the world's population every year in every country or almost every country on how thriving people are. And there's five elements of a thriving life. One is your career or your purpose. One is your financial well-being, your health, physical health, relationships, your social well-being, and then community, which is where you live, your neighbors, the church, the schools. Do you feel safe in your...
in your community. Those are the five elements. Now, our team has gone further and we think you can break those five into like 21 elements or ingredients of a well lived life. Now you don't want to impose those 21 on everybody. So for example, family and faith and hobbies might be unique. Some people are in the faith community. Some people aren't. Some people get involved in civics and politics and others don't. But with Gallup's
expanded upon wellbeing framework. A lot of life coaching will use a wheel of life that has 10 or 12 elements. We think it's deeper than that. We think that there are at least 21 ingredients or elements of a joyful, well-lived life. And what SOAR is doing is we're an AI studio, kind of a venture studio. We're incorporating 15 or more companies between now and the end of 2024.
And each of those companies will focus on one ingredient or element of life. Uh, and so at the end of the day, each company will have a specific product AI powered by our entire multimillion dollar tech stack and powered by the latest LLMs and the, the retrieval, augmented generation and all the latest cutting edge ML systems. And so the first company we started this year in January was called citizen portal.
Paul Allen (49:36.874)
Citizen Portal indexes, hundreds of thousands of public meetings in the United States transcribes all of them. So any citizen can know what your elected officials said on any topic with just a few minutes of searching. The AI will start helping you find representatives that reflect your values and beliefs so that when you vote in 2024, you go into the voting booth knowing based on AI and clips and transcriptions that I'm gonna vote for this person because they match 88% of the values that I care about at the state level.
And so anyway, that Citizen Portal is live. It's growing. It's very exciting. We think it has a chance to reach tens of millions of users. Faith Portal was formed next. And then Founder Portal, which is a co-pilot for entrepreneurs to help you make fewer mistakes, raise money in a more effective way, get AI feedback after every pitch meeting that you do. That, you know, helps you get better and better between pitches. And they even find the angels and VCs that are most likely to resonate with your story.
Founder Portal is gonna be very interesting to millions of entrepreneurs. Faith Portal, sorry, Family Portal is coming. Friend Portal is coming. That will be AI to help you deepen your friendships and destroy the notion that you have 1,500 or 5,000 friends on Facebook. You really don't have that many friends. They've destroyed the word friend and we live in a world that's got the greatest loneliness epidemic in history.
Ben (50:35.699)
Heh.
Ben (50:54.516)
Ha!
Paul Allen (51:01.954)
because people don't know what friends are and what roles they play in each other's life. So with a powerful friend assessment, which Gallup invented, if there's a partnership with them, we can then help millions or tens of millions of people figure out who their closest friends are and why, and deepen friendships with the help of AI. Well, we're gonna have a health portal, a legal portal, a finance portal, all these different portals, even gardening and animals, things that sports, music.
things that make life worth living. As those 15 to 20 companies are incorporated, raise seed funding and grow, the parent company, SOAR, will partner with each one of them so that when somebody comes to SOAR in two, three, four years to build their own bespoke AI, they can pick which of the content sets and AI co-pilots they want. And sometimes there's a values conflict. Let's say...
faith, family, friends, civics, and work are super important to you. Well, your AI needs to know what the trade-offs are. And so, you don't have time every day for all 21 ingredients. What you really need is an AI that responds to you and guides and coaches you so that if you're lonely, that maybe it suggests you spend more time on the family portal or the friend portal, or maybe it teaches you how to.
form relationships at work so that you can have a best friend at work. If you have a best friend at work, according to Gallup's decades long employee engagement survey, you're eight times more likely to be engaged at work than if you don't have a best friend at work. That's the single biggest predictor of engagement at work. Do you have a best friend at work? It doesn't have to be your very best friend in life, but if you say, I highly agree, I very strongly agree that I have a best friend at work.
You're eight times more likely to be engaged. That means you've got emotional investment and loyalty because you've got a person there. We are social creatures and workplaces have done a lot to damage our sociality. There's a lot of corporate leaders that say, don't make friends with the people who work for you. It's too dangerous because you might have to lay them off. So we check our humanity at the door and we don't bring our humanity to work where we actually can find friendship and collegiality. And that's a...
Paul Allen (53:23.406)
huge cost, human cost of capitalism and misinterpreted. So yeah, I think this family of, this portfolio of 15 to 20 AI companies, each using the same tech stack, but aimed at different content with different objectives, that will form a, I think, ecosystem that any future user of a source kind of co-pilot for life can pick and choose
Ben (53:28.991)
Mm-hmm.
Paul Allen (53:52.806)
ingredients from all of those others. And each portfolio company will get a fee, a rev share for the contribution they made to that individual. So the long-term goal, our moonshot is a Jarvis AI for everybody that they get to design or what I like to call a Janet AI, if you've ever seen Janet from the good place, the Netflix, she's my favorite fictional AI. I don't even know if they call her an AI, but I just love everything about the good Janet and I want a good Janet AI.
Ben (54:13.091)
Mm.
Paul Allen (54:22.862)
Copilot in my life that knows all the 21 elements and can help me maximize my time on earth Including deepening my most important relationships Accomplishing the purpose of my life as I see it in my faith community and also at work and it's knows my family my family History it helps me draw from the past to get through the few the present and you know Accomplish something important in the future
Ben (54:51.999)
such a fascinating proposal of almost pure unadulterated objectivity to dampen and control personal bias. Because we all have bias about everything that we do. We make, you know, we're pattern recognition machines in a biological form. Like we try to fill in the gaps of missing data. That's how our brains kind of work. Is like, hey, I don't know.
if what I'm saying right now is resonating with my audience or with the people that are on this right now. I'm assuming this by reading things from them or listening to the banter or I don't know if I'm doing the right thing at work because I got this type of feedback but not this other type. So we just are trying to constantly fill in gaps about wondering where our place is within the universe and having something that has.
objectivity instead of our own personal subjectivity is just a fascinating concept about regulation of incorrect emotional vectors that we may go down. Making a bad decision based on ill-informed data is something that I think holds a lot of people back in their life. They just assume things.
So it's just fascinating. I'm, I'm eager to try out some of this stuff.
Paul Allen (56:17.448)
The bias thing is really intense. When you look at the Wikipedia article about biases, there's 51 at least cognitive biases that we all have at some degree or another. And as I studied it, I'm like, this is impossibly complex to kind of study and eliminate those biases from your life. That's focusing on what's wrong. And the more you focus on what's wrong, it's just very...
disturbing and difficult. Now we have to overcome those biases, but Don Clifton, the father of StrengthsBinder, basically has some very simple heuristics that if you focus on those simple heuristics, it eliminates the need for all this kind of deep dive into the dozens of types of cognitive biases. One of his heuristics is find out what you do well and do more of it. Find out what you don't do well and do less of it. He also tells you to look at everybody through the lens of their strengths.
And it's really like, I think Don Clifton is like a modern way of saying, love your neighbor, you know, love others as you would, you know, treat them like you want to be treated. Don Clifton invented a beautiful metaphor in the seventies called the dipper and the bucket. And it's been popularized in grade schools everywhere. There's millions of kids that are taught about the bucket and the dipper. He basically said every human being is born with a bucket and a dipper.
And every time you encounter another human being, you're either filling their bucket because it's a positive interaction or you're depleting their bucket. And the only way for you to have a full bucket is to be a bucket filler and fill other people's buckets. And so basically he's basically saying maximize every human interaction. Look for what's good in people have a positive, you know, celebrate, recognize good work. Um, his whole book about strengths development, which he wrote in 1992 with a journalist.
is called soar with your strengths. And that's why we picked the name soar.com is that at the end of the day, if Don Clifton's vision for a strength-based world could be realized potentially with the help of AI, the world could be a very beautiful different place where everybody gets to play to their strengths, experience flow state on a regular basis. They get to see others through the best possible lens and light, which is what are your talents and capabilities that I maybe don't, maybe I take for granted, or maybe I don't even recognize them until I.
Paul Allen (58:41.47)
until I see the words and now I get to start seeing those patterns. But you're looking for positive patterns and behavior reinforcement instead of all the negative patterns that we have in abundance. It's like the DSM, the Diagnostic Statistical Manual for Psychology and Psychiatry is filled with thousands of diagnoses of all the phobias and fears and neuroses and mental illnesses that humans can have. And the field of psychology up until Don Clifton
was largely focused on figuring out what's wrong with you so that we can fix that. Don introduced, he really was the pioneer in positive psychology, which really started taking hold in early 2000s, but he's the father of strength psychology, which is looking for the talents that everybody has. And that's a complete, it's the other side of the coin. Humans need to be aware of their problems and flaws and weaknesses. But if you focus on what's right, a lot of that other stuff fades.
And I think it's possible for that to be done at scale through a strengths-based AI that's aware of relationships, well-being, work success, what that looks like, why that matters, but all areas of life. I truly think AI can lead us to a much brighter future.
Ben (59:58.515)
100%. I mean, look at anybody who is super successful in their career, where they seem to be fulfilled, happy, successful. People respect them. They respect what they can give to other people in their career.
Look at any of those people and you look far back enough in their history, there's going to be a handful or even just one person who is a mentor to them early on enough that taught them these core values about here's how humans work and here's how you can work with other humans in the best way possible. It's not somebody who's teaching you rote memorization of here's the technical skills that you need to do.
There are millions of people in technical fields who are technically excellent. They're not generally happy people at work. They, they know how to do something. They're really good at it. Probably. Maybe they don't get the recognition that they feel that they've earned or maybe they're, they think they're better than they actually are. It's just, there's some block that's, that's happened to them over the years. And.
based on my bias interaction with thousands of different people in tech roles and in engineering in general, the unhappy ones are the ones that don't know how to interact with other people properly. They may be friendly with other people, but they don't really know how to connect with others. They would never be seen as a mentor to somebody else, but the people that are super successful are mentoring.
dozens of people, they're passing on what they were taught and what they've learned through their addition of cumulative experience. And it's all positive focused, exactly as you said. If all you're focused on is fulfillment and doing the right thing for other people and doing your best in your profession and in your personal life, you actually just don't have time on any given day to do anything but that. So I...
Ben (01:02:11.215)
I 100% agree with that assessment and I can tell you that I've lived that and I know what that is where it's that whole idle time, like idle hands or the devil's play thing sort of thing. So if you don't have that idle hands because you're focused on just the things that bring you the most fulfillment and bring your community the most fulfillment and your family, yeah, it's much better than focusing on the negative.
Paul Allen (01:02:38.59)
I'm curious, Mike, Ben, who in your life played that role of a mentor or a teacher that saw potential in you or that helped you along the way? Can you think of a person or more than one person that you're grateful for that interaction because it led you to be a better person or understand the world better?
Ben (01:03:01.487)
Oh, there's been dozens. I'm a glutton for seeking those people out. I even do it nowadays. Even though I do mentor other people and other people see me as that, I still seek people out. I just always have so much to learn from other people. But some of the earliest ones were in high school. High school English teacher, Mrs. Morrill, she was fantastic.
Paul Allen (01:03:06.027)
That's awesome.
Ben (01:03:31.079)
teacher but pushed me and many times took me out of the hall during class to dress me down about you can do better and here's why. You get a paperback that is just covered with comments and it has an A minus on it. I know you can do better than this. So constantly pushing me but then providing the feedback needed to be like you're being lazy here. Use your mind.
to think about why this was being told in this story or why this was written. So that motivation was good, even though it wasn't positive all the time. It was usually like, hey, you're not doing this. Right.
Paul Allen (01:04:12.526)
But she cared. She cared for you enough to spend time and give you that feedback and that help and increase your aspirations. Don Clifton titled one of the chapters of his book, Soar With Your Strengths. Strengths only develop in relation to another human being. They only develop in relation to another human being. And if that's the formula for human development, we all need to become better at mentoring and being mentored and sharing that caring and that wisdom.
Ben (01:04:29.354)
Yes.
Paul Allen (01:04:41.846)
with other people in our circles, and especially with those that are outside of the normal circles of flourishing. Isn't it the fact that something like 50% of people in the United States make less than $15 an hour. There are so many people that are not included in the wealth creation tech industry. There are so many kind of blue collar workers. I just don't know yet how AI will reach them, although thankfully they all have smartphones and...
And if there's a way to build an AI that will help the underprivileged start their journey towards financial well-being, physical health, social well-being, career and purpose well-being. There are some very exciting scalable education companies and even universities, Western Governors University. I have several friends that work there. They have 150,000 students now.
It's Gallup has been studying Western governors for years. They have some of the best student outcomes of any university in the United States. And their tuition went down recently. They're actually a very scale and they have a mentor assigned to every single student. 150,000 students at Western governors founded by governor Michael Levitt when he was in Utah with several other Western states governors. And that was the seed of an idea 20 plus years ago. Now Scott Pulsifer runs it.
Ben (01:05:41.972)
Mm-hmm.
Paul Allen (01:06:11.85)
and they've grown to 150,000. He told me personally, they'd like to get 10 million students someday. Well, imagine every one of those students has a mentor that I think meets with them weekly. And there's really amazing, Khan Academy has teamed up with Instructure. There's so many learning platforms. And so, well, a lot of economic models in the past depended on scarcity. AI could lead us to a new era of abundance of information.
The Library of Congress, I was in Chicago at the ALA conference in June, and I got to meet with some of the folks at the Library of Congress exhibit. And I noted that Library of Congress has made a policy statement that anything generated by AI cannot be copyrighted because it wasn't created directly by a human. Now, if an artist is using mid journey and interacting and going back and forth or Photoshop with its new AI, I believe that humans can say, well, I created this with the help of AI.
But if, if chat GPT was to answer every question on Quora, let's say you scraped Quora and had every single question ever asked, and you created several personas, let's say 10 or 20 different personas, okay, GPT, you're the world's best physicist answer this question, GPT, you're the world's best political scientist, historian, teacher, nurse, doctor, answer this question from the lens of you're the smartest in the field. Okay. Imagine 20 answers.
Each from a different persona, none of that copyrightable. I mean, AI if now there will be legislation, there'll be litigation. It's going to be an ugly five or 10 years ahead of us to, you know, it's going to be a bloodbath in many ways. But if that ruling stands that generative AI cannot be copyrighted, just like pre 1928 or federal government content, there's a lot of copyright free material in the creative commons that.
is so wonderful for human development. Open source code. I mean, think about ancestry.com only exists because billions and billions of records could be put online with no permission from anybody. We didn't have to go pay royalties or license public domain content of deceased people. Well, in a world where copyright laws don't matter as much anymore because generative AI is copyright free, then think of the ability to create generative AI
Ben (01:08:14.12)
Mm-hmm.
Paul Allen (01:08:39.63)
content, it could be in two or three years when you're doing Hollywood quality video movies, all done with AI. And maybe that's, I don't know what the copyright rulings will be, but I just imagine we could enter into an era where there's no kind of extreme cost barrier and no exclusivity. There's an abundance of knowledge and an abundance of AI access to anyone on a smartphone, anywhere in the world that you can learn what you need to learn.
but it's personalized learning based on your strengths, your values and your goals in life. And the chance to have that kind of companion co-pilot, it doesn't replace the need for human friendships and mentoring by any means. I don't think Don Clifton's principle that strengths only develop in relation to other human beings will be replaceable by an AI. I think it could be enhanced by an AI. When you don't have your coach or your manager or your friend with you, the AI can...
you know, interact with you and it could be helpful, but it won't take the place of Ben, those people who saw your potential and gave time and caring to help you believe in yourself and, and stretch beyond what you were currently doing.
Ben (01:09:53.912)
Exactly. Very well put.
Michael (01:09:58.018)
So it feels like the conversation.
Paul Allen (01:09:58.758)
Michael, I'm curious about Michael's influences. Can you think of a person in your life that helped you along the way?
Michael (01:10:08.526)
Oh, I definitely can. Ben is for sure one of them, but just for diversity of conversation. I remember my first sort of tech internship, I had a manager named Elliot Starr. This man is the second most intelligent person I've ever met and he was incredibly emotionally attuned to the team. And I...
Yeah, we'll forever be grateful. I only worked with him for three months, but, um, this man really knew how to manage. And as Ben said, sort of when you go out and seek people to learn from, uh, I feel like those types of people attract each other and he was one of those types as well. And I remember a story that he told me, uh, I went into his office and I was like, so by the way, I like grinding out on problems and like,
banging my head against a wall until stuff works. And he's like, well, that's dumb. I was like, but that's how you learn, right? And he said, no, just get someone to teach you. And he went on to tell a story about how he had a buddy who was a borderline Olympic soccer coach. And he said one day, oh, I wanna learn soccer. Can you teach me? So his buddy went and basically.
once a week for like a couple hours, explained the concepts of soccer. And by the end, Elliot was, was quite proficient. And if he had gone and just played with random people, watched highlights, try to figure it out on his own. That learning path would be so much less efficient. And so sort of knowledge transfer between humans is, is a beautiful thing and it's really, really powerful.
Paul Allen (01:11:46.97)
a great story Michael. I love it that's way more than two cents worth.
Michael (01:11:47.834)
So that's my two cents. Yeah.
Michael (01:11:55.538)
Um, cool. So it feels like this conversation is starting, but, uh, it's actually ending because we have a set amount of time. Um, I personally have 100 more questions, but, uh, instead I'm going to wrap. It's very sad. Um, so in summary, we talked about a lot of really, really cool stuff and some things that stood out first and foremost, the concept of virality. So it's often sort of hard to define. And you can think of.
TikTok videos or whatever it might be. But the way Paul describes it is sharing the product inherently adds to the product's value. So the more users, the better the product is. And this typically requires a free version and then a premium version so that you can support your, your company. Um, but yeah, it's breaking it down in that way makes a ton of sense. And then also check out a viral coefficient, which is essentially the number of users that a specific user can bring on per unit time.
Sore is his company and it introduces the concept of sort of uplifting via AI. And this is absolutely fascinating when everybody's talking about how AI is going to end the world, how it's Terminator part seven, um, AI can actually improve the quality of life dramatically. And looking to quantify this in a, in a repeatable manner is really awesome. And then some applications that you can apply AI to personalized knowledge recommendations, improving communication based on personal style.
task management slash assignment. These are just a few of the things that we talked about. Um, and there are many more, but it, one thing that stood out to me is the common underlying theme is that LLMs and sort of these embedding models create. A structure or sort of a way to convert between data and sort of how the recipient wants to receive that data. So it's sort of knowledge distillation and information format conversion. And that's opened the door to.
millions of industries. So over the next 10, 15 years, it'll be really exciting to see what is built on top of these LLMs. So Paul, if people want to learn more about you or your work, where should they go?
Paul Allen (01:14:05.954)
They can go to soar.com to learn about the company and as new entities are formed, they will be announced there. But they can also reach out to me at paul at soar.com, S-O-A-R. And LinkedIn, although I do have 30,000 connections on LinkedIn, so I guess you can follow me on LinkedIn. I'm active on Twitter slash X. But yeah, I do welcome. I love hearing from people anywhere who are working on similar problems or...
or trying to find positive ways to use AI. I love doing podcasts and we actually have our own show that we'll be rebooting soon called AI to Uplift Humanity. And I think we've done 11 or 12 episodes. We're gonna reboot that later this year. But yeah, I look forward to hearing from some of your listeners and love to catch up with you all in the next few months. I would love to talk again in a year after.
All of us are using some of these new tools just to see where they're taking us.
Ben (01:15:08.563)
Definitely.
Michael (01:15:09.71)
Yeah, we'll be sure to drop Ben's advisory contract after this. Cool. Well, until next time, it's been Michael Burke and my cohost and have a good day. Everyone.
Paul Allen (01:15:12.592)
That's right.
Ben (01:15:18.911)
Ben Wilson.
We'll see you next time.