Latest / The Adults In The AI Room / Paul Zhao on Democratizing AI - Episode #8

Transcript

Paul Zhao on Democratizing AI - Episode #8 · The Adults In The AI Room

• 0:22
  SPEAKER_00: Thanks so much for listening.
• 0:23
  SPEAKER_00: If this is your first time, we'd like you to subscribe wherever you listen to the podcast.
• 0:29
  SPEAKER_00: You're your host.
• 0:30
  SPEAKER_00: Seth Early and Chris Featherstone.
• 0:36
  SPEAKER_03: Good afternoon, good morning, good evening, depending upon your time zone.
• 0:40
  SPEAKER_03: Uh welcome to today's podcast.
• 0:43
  SPEAKER_03: I am Seth Early and uh I'm Chris Featherstone.
• 0:46
  SPEAKER_02: Good to be with you.
• 0:46
  SPEAKER_03: Yeah, good to see you, Chris.
• 0:48
  SPEAKER_03: I'm sure you had a uh jammed week as usual.
• 0:51
  SPEAKER_03: How was your week?
• 0:52
  SPEAKER_02: You know, it's it's been really interesting because I get uh I'm headed to New York next week, right?
• 0:57
  SPEAKER_02: So not only the combination of all of the uh coronavirus updates up and down, like hey, today you have to be in a in a a bubble 12 feet away from each other, and then the next hour get a meet get another notification.
• 1:12
  SPEAKER_02: No, don't worry about it.
• 1:13
  SPEAKER_02: In an approved location, you don't even need to wear a mask.
• 1:16
  SPEAKER_02: I'm like, what in the hell is up, right?
• 1:18
  SPEAKER_02: Plus the eastern seaboard is frozen from up, right?
• 1:23
  SPEAKER_02: So I don't know if if flights are going to be going, so it'll it'll be uh fun.
• 1:27
  SPEAKER_02: Probably have to hitchhike.
• 1:28
  SPEAKER_02: You know what?
• 1:29
  SPEAKER_02: Intermodal uh transportation built this country, right?
• 1:31
  SPEAKER_03: So let's jump in.
• 1:32
  SPEAKER_03: Our guest today is someone who sits comfortably at the intersection of technology and business.
• 1:37
  SPEAKER_03: Most recently, he'd been at Snowflake as a product manager for machine learning, but his entrepreneurial spirit has guided him on a journey of very interesting waypoints, which we look forward to hearing more about uh in our talk.
• 1:52
  SPEAKER_03: Please welcome from uh I guess it's Chile Seattle these days, uh Paul Zhao.
• 1:57
  SPEAKER_04: Paul, welcome.
• 1:58
  SPEAKER_04: Hi, Seth.
• 1:59
  SPEAKER_04: Uh Chris.
• 1:59
  SPEAKER_04: Thanks uh for having me.
• 2:01
  SPEAKER_04: Happy to be here.
• 2:02
  SPEAKER_03: So so Chris, you and Paul had uh connected AWS.
• 2:05
  SPEAKER_03: Do you want to uh tell us how that happened and give us a little bit of history there?
• 2:09
  SPEAKER_02: Yeah, so I'm so I'm super excited to have Paul on.
• 2:12
  SPEAKER_02: He's he's uh not only is he a good, you know, just a phenomenal mentor to me, but he's also a just you know great friend and person.
• 2:22
  SPEAKER_02: And uh, you know, one of my main goals of uh of you know some companies that I chose to go work for were to literally go find great people to work with.
• 2:32
  SPEAKER_02: And uh, you know, because once you get to you know, past all the trying to you know climb the corporate ladder and do all that other fictitious BS that's out there, it's more along the lines of just working with great people.
• 2:43
  SPEAKER_02: And uh when I met Paul, super impressed by just you know his approach to things, but as we got to know each other, you know, I you know, just I love the guy, his family, his kids are I think are the cutest kids on the planet.
• 2:56
  SPEAKER_02: And um, so when he, you know, when when take him, I absolutely will.
• 3:02
  SPEAKER_02: But I'm but I'm also equally um just blown away by his background and what he's been able to accomplish, right, in his career.
• 3:10
  SPEAKER_02: So that's part of you know, just surrounding your p yourself with great people is you get, you know, you get the we'll say the the overlap or the exhaust or just some of that excellence that just you know rubs off on to you and makes you a better person.
• 3:24
  SPEAKER_02: So I'm stoked to have him on.
• 3:26
  SPEAKER_03: So tell us a little bit about uh your background.
• 3:28
  SPEAKER_03: I know you uh you have some interesting uh stories, and maybe you can just kind of tell us a little about how you got into the space and uh give us a little bit of the history according to uh Paul.
• 3:39
  SPEAKER_04: Yeah, yeah, happy to.
• 3:40
  SPEAKER_04: And you you know, you cut me short when the the tail runs long, as they say.
• 3:44
  SPEAKER_04: Um so most recently, like uh Chris and you introduced, I was at uh Snowflake.
• 3:50
  SPEAKER_04: I was a senior manager there responsible for uh the ML strategy, specifically working on ML platforms that really give accessibility, if you will, to um folks who are not always coming from an ML background to be able to use these tools.
• 4:05
  SPEAKER_04: And then preceding that, I was with AWS, where Chris and I met.
• 4:10
  SPEAKER_04: Um, and I just remember Chris, you know, maybe off of the first five minutes of the conversation, it was just entirely a firecracker.
• 4:18
  SPEAKER_04: Um, so there I worked on speech recognition uh under the ML group.
• 4:22
  SPEAKER_04: And uh Chris and I had a chance to really go tackle some pretty tough uh enterprise customers.
• 4:28
  SPEAKER_04: And uh Chris helped us to open a lot of doors, and uh we felt like uh a team, one, two punch.
• 4:34
  SPEAKER_04: Even farther back, I ran my own business uh in a company called Black Swan Tech, which is uh responsible for selling cloud insurance.
• 4:41
  SPEAKER_04: And the truth of that is you know, if you are a company that runs a public cloud service, like many enterprises do, and uh, you know, a public cloud service has you know downtime or failure that causes disruptions your business, then our company will come in the middle here and through a contract with us, we'd make you whole.
• 4:58
  SPEAKER_04: So we'd issue out a payout that makes you whole for that business disruption that you might be suffering.
• 5:03
  SPEAKER_04: Um, I went to business school before that, and then preceding that, I ran my own business for about five years in China in uh IoT.
• 5:11
  SPEAKER_04: So that's the short, short version of the of the journey.
• 5:15
  SPEAKER_03: Yeah, I I loved hearing the longer version about the IoT initiative.
• 5:19
  SPEAKER_03: And this is back in what year was this?
• 5:21
  SPEAKER_03: So this is when it was fairly new stuff, right?
• 5:24
  SPEAKER_04: Yeah, this was in uh through 2013.
• 5:29
  SPEAKER_04: So right at the height of the economic recession.
• 5:34
  SPEAKER_04: Yeah, my my co-founder, my buddy, and I had the probably stupidity, although I'll package it as wisdom to go to naivete.
• 5:44
  SPEAKER_03: It's not knowing what's not possible.
• 5:47
  SPEAKER_04: That's right.
• 5:47
  SPEAKER_04: It's it's ignorance, actually.
• 5:49
  SPEAKER_04: Yeah, um, so we flew over to China and said, you know, uh we'd like to basically retrofit large-scale commercial facilities with sensors and then hook it up to uh software that allowed these operators to essentially make real-time decisions based on real-time data.
• 6:07
  SPEAKER_04: Um, so that was kind of how it all began.
• 6:10
  SPEAKER_03: And one of the ones you you chose were the parking uh facilities, correct?
• 6:14
  SPEAKER_03: That's right.
• 6:15
  SPEAKER_03: That's right.
• 6:16
  SPEAKER_03: And you started, you tried in the US, but couldn't get much uptake.
• 6:19
  SPEAKER_03: I mean, people didn't really understand it at the time.
• 6:22
  SPEAKER_03: And you know, you didn't you didn't have a reference point, people really were.
• 6:27
  SPEAKER_03: I mean, it wasn't called even probably called IoT at the time.
• 6:30
  SPEAKER_03: It was it was just sensors, and so you kind of had to explain what this whole paradigm was.
• 6:37
  SPEAKER_03: And that's a that's a tough sell sometimes.
• 6:39
  SPEAKER_03: That's you can be early to the market, uh, a little too early to the market, where people don't have that understanding.
• 6:45
  SPEAKER_03: You want to talk a little bit more about that?
• 6:47
  SPEAKER_03: And because I think that bears a lot on you know how we're looking at paradigm changes today, right?
• 6:54
  SPEAKER_03: And how people need to look at things very differently.
• 6:56
  SPEAKER_03: So back then, you were trying to get people to think about wait a minute, you have this inventory, you know, and you're not optimizing this inventory.
• 7:05
  SPEAKER_03: So, how can you help it?
• 7:06
  SPEAKER_03: That and so that started.
• 7:07
  SPEAKER_03: Uh, did you think of other areas to add sensors or just just dig into that story a little bit more, if you would?
• 7:14
  SPEAKER_04: Yeah, happy to.
• 7:14
  SPEAKER_04: And I think you're you're touching on really two important challenges.
• 7:18
  SPEAKER_04: One is more of a macro challenge that we face, and I'll talk about that.
• 7:21
  SPEAKER_04: The other one is more of a micro challenge that we face as you know, founders and operators of our own business.
• 7:27
  SPEAKER_04: Um, you know, like you alluded to, the macro challenge was that 08 was a time uh when we first started in the US, like you said, we didn't have a lot of traction, but it was a time in which we had the hurdle of being in the middle of digitization, right?
• 7:42
  SPEAKER_04: Um, today we take digitization for granted.
• 7:45
  SPEAKER_04: Everybody's like a few people are on paper.
• 7:47
  SPEAKER_04: A lot of processes and operations are digitized, automated.
• 7:50
  SPEAKER_04: Um, there's there's a lot of tooling.
• 7:52
  SPEAKER_04: But at that time in 08, a lot of businesses were actually still pen and paper.
• 7:55
  SPEAKER_04: And it didn't seem like a long time ago, 14 years ago, but actually that was what the world was you know composed of.
• 8:01
  SPEAKER_04: It was pen and paper.
• 8:02
  SPEAKER_03: Uh, you know, you could forget how quickly all of this has changed.
• 8:07
  SPEAKER_03: Absolutely.
• 8:08
  SPEAKER_03: Forget how fast this has been.
• 8:11
  SPEAKER_03: Absolutely.
• 8:12
  SPEAKER_02: We get we we got the uh notification what last week that finally SAP tests are gonna be digital.
• 8:20
  SPEAKER_02: Not till 2024.
• 8:22
  SPEAKER_02: So good job.
• 8:23
  SPEAKER_02: Yeah, yeah.
• 8:25
  SPEAKER_04: Anyway, uh I'll miss those number two pencils, though.
• 8:28
  SPEAKER_04: I'll tell you anyway, keep going.
• 8:33
  SPEAKER_02: Very good job.
• 8:34
  SPEAKER_04: Yeah, so that was the first challenge is that you know, uh, we started in the US uh reaching out to garage operators or commercial garage operators, and there was this big question mark like, why do we need to care about optimizing our pricing in real time?
• 8:51
  SPEAKER_04: People are gonna park if they want to park, if they don't want to park, they're not gonna park, even you know, even if it's free, right?
• 8:56
  SPEAKER_04: So there was a lot of uh you know non-data driven type of uh consensus around the lack of need for digitizing the data, right?
• 9:05
  SPEAKER_04: Whether there are cars or not, how many spots there are surrounding their operations.
• 9:09
  SPEAKER_03: Um and they didn't know what they didn't know, they didn't know, you know, they ask people what they need, they they put it in the context of what they have.
• 9:18
  SPEAKER_03: They they can't envision this shift that is foreign to them because they don't even under they can't conceptualize what the problem really is and how that problem can be solved.
• 9:31
  SPEAKER_03: Like if you ask people 20 years ago if they needed you know one of these magical devices, right?
• 9:37
  SPEAKER_03: They wouldn't say they did.
• 9:39
  SPEAKER_03: You know, they didn't know, they didn't know what they could see, they didn't know what they could do.
• 9:43
  SPEAKER_02: But you didn't you Seth, you gotta tell me you didn't you didn't have an Apple Newton at the time.
• 9:49
  SPEAKER_03: I it was very funny.
• 9:51
  SPEAKER_03: I had I went from uh pager to Blackberry, uh, and for a long time a flip phone, and then I wasn't brought into the Apple ecosystem until relatively late in life.
• 10:05
  SPEAKER_02: So so I'm not I'm not gonna date myself, but you know, VHS tapes, right?
• 10:10
  SPEAKER_02: Remember that big old black, you know, tape that.
• 10:12
  SPEAKER_02: But the Apple Newton came out and it had a stylus total monochrome in '92 or '93, right?
• 10:20
  SPEAKER_02: Yeah.
• 10:21
  SPEAKER_02: Yeah.
• 10:21
  unknown: Yeah.
• 10:21
  SPEAKER_02: And I I mean it wasn't connected.
• 10:23
  SPEAKER_03: And people envisioned this day that where everything would be available on this device, but it was so far into the future, it was ridiculous.
• 10:31
  SPEAKER_03: So, so Paul, you know, go back to your thinking or your story at that time, you know, what you were facing and how you tried to overcome that.
• 10:39
  SPEAKER_03: And then, you know, what were the arguments back from people?
• 10:43
  SPEAKER_03: And and like you say, they didn't understand that they needed this, they didn't understand what was possible.
• 10:48
  SPEAKER_03: We face a very similar situation today with any of the machine learning and AI algorithms that we're starting to build.
• 10:56
  SPEAKER_03: There's a lot of uncertainty and unknown about what this can bring to organizations.
• 11:01
  SPEAKER_03: And the at the one hand, there's a big picture arm wavy, yes, it's gonna help all these years, but it's at that granular level of what is it gonna do for a process.
• 11:11
  SPEAKER_03: So, talk a little more about what you had to go through to kind of bridge that gap.
• 11:15
  SPEAKER_04: Sure.
• 11:16
  SPEAKER_04: Uh, and I I love the you know, the Apple Newton and the cell phone, you know, analogies you guys are bringing up.
• 11:21
  SPEAKER_04: Because at the time I actually had a palm pilot, if you guys remember that before POM just went bust.
• 11:26
  SPEAKER_04: Um, and actually a lot of the business owners we went to actually used to give us that example.
• 11:31
  SPEAKER_04: It's like, you know, we've adopted technology before early, and it's kind of blown up or presented as low value.
• 11:37
  SPEAKER_04: And one of the examples was the palm.
• 11:39
  SPEAKER_04: Um, like I remember one of the COs I was talking to said, you know, you see this thing and he whips it out.
• 11:43
  SPEAKER_04: This this palm that's obviously collected, thus never been charged.
• 11:46
  SPEAKER_04: And he says, you know, here's what I fear is that this is a personal item, but if I apply this technology and buy it at the commercial level, that's a lot of costs and a lot of stuff to do that won't pay off.
• 11:56
  SPEAKER_04: So there was a lot of doubt.
• 11:58
  SPEAKER_04: Also, 08 was a time when there was, you know, uh it was like an economic crunch, right?
• 12:03
  SPEAKER_04: So liquidity wasn't always in place.
• 12:04
  SPEAKER_04: So we were talking about retrofitting very large-scale commercial garages.
• 12:09
  SPEAKER_03: So so even going back, what you're looking at, are people are taking examples of failed technology and extrapolating that to the current uh objective or possibilities.
• 12:23
  SPEAKER_03: And and as you can see, that technology obviously evolved and changed.
• 12:28
  SPEAKER_03: We have to look at technology in the same way.
• 12:30
  SPEAKER_03: We know, yes, there's going to be a lot of failures, a lot of errors, a lot of mistakes, a lot of things that aren't going to provide value.
• 12:38
  SPEAKER_03: But just as the Palm pilot led to our, you know, our our handheld device and mobile world, which is everything, you got to think of the same ideas, the same analogies of other technologies that may look like they're not producing value.
• 12:54
  SPEAKER_03: But if you extrapolate into the future, they will produce value.
• 12:58
  SPEAKER_03: It's inevitable.
• 12:59
  SPEAKER_03: It's inevitable.
• 13:00
  SPEAKER_03: So keep going, keep going.
• 13:01
  SPEAKER_04: Yeah, no, no, you're you're spot on.
• 13:03
  SPEAKER_04: So with I think any form of innovation, uh, there's going to be these early adopters who essentially do the world a favor, if you will, of running that experiment, consuming whatever that new innovation is.
• 13:16
  SPEAKER_04: And they're kind of footing the bill, if you will.
• 13:18
  SPEAKER_04: And um, in some sense, sometimes it works, sometimes you know the technology doesn't fly.
• 13:22
  SPEAKER_04: But either way, there is that inadvertent experimentation, and they're kind of you know doing this, like I said, doing the world a favor, and we we wouldn't be anywhere without these early adopters, right?
• 13:32
  SPEAKER_04: Um the pioneers get the best land, but they also get arrows in the back.
• 13:38
  SPEAKER_04: That's right.
• 13:38
  SPEAKER_04: Sometimes you get a lot more than they bargain for, uh, especially depending on the scale of what they're looking at, right?
• 13:44
  SPEAKER_04: Um, so you're you're spot on.
• 13:46
  SPEAKER_04: But you know, going back to the the other challenge that we had was, you know, at the end of the day, we were actually presenting two theses to uh prospective customers.
• 13:57
  SPEAKER_04: We were saying you have to digitize, that's the first step.
• 14:00
  SPEAKER_04: And underlying that, there's this whole notion of data and the utility of data that supplements, supposed to supplement, but many saw as supplanting, right?
• 14:09
  SPEAKER_04: Substituting away experience and judgment.
• 14:12
  SPEAKER_04: And so there is that concern.
• 14:14
  SPEAKER_04: We've been operating our businesses for a long time.
• 14:16
  SPEAKER_04: It's worked, we've seen revenues grow.
• 14:19
  SPEAKER_04: Why would we need you guys?
• 14:21
  SPEAKER_04: You guys just look like a cost item, a line, you know, a line of expense.
• 14:25
  SPEAKER_04: Um, and then of course, when you couple that notion with, well, at the end of the day, we actually have to build intelligence, which is software, processing the data on top of this digitization process of collecting the data.
• 14:38
  SPEAKER_04: And then we've now made this very, very undigestible or at least not palatable um for a lot of customers at that junction uh in 08.
• 14:48
  SPEAKER_04: So, you know, we were kind of like at our you know wit's end.
• 14:51
  SPEAKER_04: We tried we tried to sell our tech into Chicago, into New York, basically large metropolitan that has you know parking, uh parking facilities.
• 15:01
  SPEAKER_04: We basically got like no, no, no, no, no, everywhere we went.
• 15:05
  SPEAKER_03: How many did you try?
• 15:06
  SPEAKER_03: Just out of curiosity, do you remember how many?
• 15:08
  SPEAKER_04: Oh god, uh formal conversations over 50 informal ones.
• 15:14
  SPEAKER_03: Imagine that.
• 15:14
  SPEAKER_03: Think of that.
• 15:15
  SPEAKER_03: Think of when people get discouraged after two or three or five.
• 15:19
  SPEAKER_03: No, and you went to 50 and were still believing in your vision.
• 15:25
  SPEAKER_03: That is that is a testament to number one, the resiliency, number two, the belief, and maybe it's number one that's the belief, number two, it's the resiliency, but the but the the um you know the the the the the the conviction and the the on on that outcome.
• 15:42
  SPEAKER_03: And I'm trying to I'm struggling for the word, but it's the dog at the bottom, right?
• 15:47
  SPEAKER_03: Ignorance well you know the thing about but the thing about being an entrepreneur, there's two things that's interesting.
• 15:53
  SPEAKER_03: One is uh that inability to give up, right?
• 15:56
  SPEAKER_03: They keep pushing, they keep pushing.
• 15:58
  SPEAKER_03: I'm gonna keep trying.
• 15:59
  SPEAKER_03: I don't want to fail.
• 16:00
  SPEAKER_03: I'm gonna go, I'm gonna go.
• 16:01
  SPEAKER_03: And other times it's like I should give up, right?
• 16:04
  SPEAKER_03: Like going too far.
• 16:06
  SPEAKER_03: And so you continue to believe in this after 50 formal meetings and those, uh tell us what you did then.
• 16:14
  SPEAKER_03: But that is a testament to the the uh uh the belief in what you're doing and the resilience around pushing that belief.
• 16:23
  SPEAKER_03: Even you're generous, you're you're generous.
• 16:26
  SPEAKER_04: I I think it's uh you know, looking back, I think it it looks that way.
• 16:31
  SPEAKER_04: And I think from the outside in, it's like this guy and his co-founder had a lot of uh you know, tough skin or grit.
• 16:37
  SPEAKER_04: And and I you know, I appreciate all that.
• 16:40
  SPEAKER_04: Um, you know, from the inside out, it's more around the lines of a combination of perhaps ignorance, not knowing what we're up against, right?
• 16:48
  SPEAKER_04: To some degree of stupidity because we're not learning going up that learning curve quickly enough, or like, oh, 50 failed conversations, guys.
• 16:54
  SPEAKER_04: 50 first will make do, right?
• 16:56
  SPEAKER_04: Um, then there's like some degree of arrogance, you know.
• 16:59
  SPEAKER_04: Hey, look, we're introducing something novel, or at least we'll solve a problem, but the old guard, if you will, just can't see it.
• 17:06
  SPEAKER_04: Um, and that all culminates into this uh wonderful self-packaging of you know being able to push through.
• 17:12
  SPEAKER_04: But I think at the end of the day, it's really those three things that you know helped me build that tough skin.
• 17:17
  SPEAKER_04: I just didn't know better.
• 17:18
  SPEAKER_03: You're you're calling it you're calling it ignorance and arrogance.
• 17:21
  SPEAKER_03: Yeah, combo.
• 17:24
  SPEAKER_03: So ignorance, arrogance, resilience, and uh grit.
• 17:28
  SPEAKER_03: You know what?
• 17:30
  SPEAKER_03: Yeah, but you know what though?
• 17:31
  SPEAKER_02: If you need that, I think I think if if if if we want to tease out what the theme of this could and should be, it's it's centered around that those topics, thoughts, and ideas of when you know you have something that is innovative that will provide efficiency or revenue generation or a better experience in all, you know that it's it's ahead of the curve in how you actually help you know mold somebody's thinking into the that belief system because there's are there are early adopters, there's you know, there's um late adopters, but there's those folks that you can hopefully take as the beginning and get the ground swell to actually adhere to the idea.
• 18:12
  SPEAKER_02: We do this.
• 18:13
  SPEAKER_02: I mean, this is what maybe demystifying AI and ML is all about is this idea of hey, we've got these very early innovative thoughts and ideas that will help you in some way of your business.
• 18:26
  SPEAKER_02: How are we gonna get you across?
• 18:28
  SPEAKER_02: I think Paul, from from what I know of you and your experience and your background and everything else, you have to do this over and over and over again because you've been on the on the tip of the spear for all these different experiences.
• 18:40
  SPEAKER_02: And I think that for me is is um not even those types of things.
• 18:46
  SPEAKER_02: I think ignorance is a choice, and naivete is like you said, not knowing what you don't know.
• 18:51
  SPEAKER_02: But you know that you have a skin too.
• 18:54
  SPEAKER_02: You're not I mean you're not stupid.
• 18:55
  SPEAKER_02: There's no way.
• 18:56
  SPEAKER_02: I mean, look at the little uh look at the the um framed pictures behind you, right?
• 19:01
  SPEAKER_02: It just shows that you know you're a smart guy.
• 19:04
  SPEAKER_02: So I think the key thing, especially for this, is that theme of how to help convince others and what's that process look like in order to actually help them, you know, get past the early doctors across the chasm and into the ground swallow people, you know, broadly adopting these technologies because you had to do it your entire career.
• 19:22
  SPEAKER_02: Is that fair?
• 19:23
  SPEAKER_04: And what is you know what you know again, you're you guys are so generous.
• 19:28
  SPEAKER_04: Um, it's definitely a I think a combination of things that you mentioned.
• 19:32
  SPEAKER_04: Um, you know, it's a good and a bad you know, entrepreneurship is just like a very hard thing to dissect.
• 19:38
  SPEAKER_04: I know there's like a lot of books and a lot of people who have much more success than I do, who have given like the breakdown of how do you do zero, one, two, three, and more importantly, how do you do that in the age of AI?
• 19:48
  SPEAKER_04: Because AI is all the rage when entrepreneurship in AI is all the rage.
• 19:53
  SPEAKER_04: When I look at this, I don't personally, and I could be mistaken, I don't personally see an easy recipe, whether it's AI or non-AI startups, um, whatever arena.
• 20:03
  SPEAKER_04: I think I think if there if there were a recipe recipe, we'd have way more successful startups and we wouldn't be dealing with um a pretty stable statistic of less than you know 1% businesses succeeding off the ground.
• 20:17
  SPEAKER_04: Um, and you know, we've had a lot of data to support this.
• 20:20
  SPEAKER_04: So, you know, Chris to your point, it's probably a lot of things going on all at once.
• 20:23
  SPEAKER_04: Um, you know, just this torrent of experimentation, uh, not giving up, some degree of ignorance sprinkled with arrogance, you name it, it's probably in there.
• 20:34
  SPEAKER_04: Um, but failure is the one common thing that I feel like uh has to be part of it, uh, whether it's like a forcing function or whether it's just um something that forces us to grow individually.
• 20:46
  SPEAKER_04: It that's definitely part of it.
• 20:48
  SPEAKER_04: I've had my for sure.
• 20:51
  SPEAKER_03: Well go ahead.
• 20:52
  SPEAKER_03: Uh I was just gonna ask it to to how did you finally get traction?
• 20:57
  SPEAKER_03: You had to go to China to get traction.
• 21:00
  SPEAKER_03: What was different about um that per you know, that environment versus the US environment?
• 21:07
  SPEAKER_04: Were they more open to or did you how well the short of it was that we had kind of run out of runway?
• 21:14
  SPEAKER_04: Um, we were coming down to the wire on what little savings and funds that we had.
• 21:19
  SPEAKER_04: We had boot trapped everything financially on our own uh you know savings.
• 21:24
  SPEAKER_04: Uh we we had formerly both worked at Goldman, so that helped a little bit as you know, dumb college kids coming out of you know from a university, but that's never enough to keep you going for multiple years.
• 21:34
  SPEAKER_04: Um, so that was one reason.
• 21:36
  SPEAKER_04: But on the business side, it was also that we felt China was a highly competitive place.
• 21:43
  SPEAKER_04: It was also during a time when uh basically businesses were willing to try anything new to get an edge on their competition.
• 21:52
  SPEAKER_04: And so the appetite and the culture of experimentation felt extremely open.
• 21:57
  SPEAKER_04: Um, and the other thing was everything in China is big.
• 22:00
  SPEAKER_04: Right.
• 22:00
  SPEAKER_04: Like just because of the population, everything there is big.
• 22:03
  SPEAKER_04: So even a small percentile success would eke out larger absolute value for us.
• 22:08
  SPEAKER_04: Right.
• 22:09
  SPEAKER_04: That was the supposed wisdom at the time.
• 22:12
  SPEAKER_04: You know, we'll never know if we just got lucky or not, you know, because we don't have a control group.
• 22:18
  SPEAKER_02: Well, I mean, what they you know, I I was uh I was watching my son from basketball last night, and um the opposing coach was um, you know, his son was on the team and he made a bad move or bad pass or something, and his dad just unloaded on him, right?
• 22:35
  SPEAKER_02: And I I was thinking about it.
• 22:37
  SPEAKER_02: I'm like, listen, if you're always winning, you never learn.
• 22:40
  SPEAKER_02: If you're failing, then you're learning, right?
• 22:42
  SPEAKER_02: Every every step of the way.
• 22:43
  SPEAKER_02: And I was like, it was an opportunity to just I've learned a lot.
• 22:46
  SPEAKER_02: Yeah, yeah, exactly.
• 22:47
  SPEAKER_02: But just to take a minute and say, listen, let me let me teach you something, right?
• 22:52
  SPEAKER_02: Let me coach you through something.
• 22:54
  SPEAKER_02: And I think that's probably, you know, and what's you know, define irony.
• 22:58
  SPEAKER_02: Every time we do something in in AI and in ML, we have to utilize some type of hypothesis in order to get clarity of what the outcomes that we're driving are.
• 23:09
  SPEAKER_02: Set an initial hypothesis, okay.
• 23:10
  SPEAKER_02: That's what we want to think that this we're gonna do.
• 23:13
  SPEAKER_02: And then yeah, but the problem with that is now if you add any more data points in the data model, then it completely blows up what that end goal is gonna be anyway.
• 23:23
  SPEAKER_02: So I think part of that too, you know, Paul, is is obviously you learned a ridiculous amount of of you know, humans are humans, we have all these patterns that we you know associate with psychologically and stuff, but you know, just in terms of how much you learn over failure, you know.
• 23:41
  SPEAKER_03: Absolutely.
• 23:42
  SPEAKER_03: So I yeah, I think go ahead, Paul.
• 23:44
  SPEAKER_04: I'll let you finish that and then uh I I was I'll be brief.
• 23:48
  SPEAKER_04: I was just gonna say that you know, Chris touches on an interesting thing because um uh you know, ML AI is all about reducing the problem scope for whatever problem or decision you have to make today.
• 23:59
  SPEAKER_04: Um and at the time our business hadn't actually begun using ARML because this is 08 going on in 2013, AI wasn't hot yet, right?
• 24:11
  SPEAKER_04: It was still like niche.
• 24:12
  SPEAKER_04: Um, but nonetheless, the application of technology to crunch data was still in play.
• 24:19
  SPEAKER_04: Like our goal was how do we take data that you weren't using about your vacancies, about the cars you know, that were you know lining up around the block to get in and time, right?
• 24:29
  SPEAKER_04: How do we take all that data, put it into an algorithm that wasn't AI driven, but it was still an algorithm nonetheless that can eke out value.
• 24:38
  SPEAKER_04: It isn't until much later, like in my second startup, that we actually started thinking about how do we use AI, how do we use more sophisticated self-learning models and whatever distinction we call it between the dumb algorithms, as I call them, or the smarter ones about self-learning, the general goal is always the same.
• 24:56
  SPEAKER_04: The goal is always how do you capture value?
• 24:58
  SPEAKER_04: First, how do you identify value, and then how do you capture that value for whoever the business stakeholder is?
• 25:03
  SPEAKER_04: Um, so I think Chris, you touched upon a really good thing, which is you said the word judgment.
• 25:07
  SPEAKER_04: And and I when I think about AI and ML, without getting into like any of the weeds, it's always how do you supplement human judgment?
• 25:15
  SPEAKER_04: Um, but not substitute it, because unless we're talking about you know, uh Terminator 2 or any of those fantastical things, we're we're not at that junction yet.
• 25:25
  SPEAKER_04: It's going to be a complementary um effect between human judgment and experience and execution for data analysis and machine learning and all that jazz.
• 25:34
  SPEAKER_03: So I totally agree.
• 25:37
  SPEAKER_03: I mean, you know, we always talk about how uh AI isn't the support of a human.
• 25:43
  SPEAKER_03: It's another tool in the toolkit.
• 25:45
  SPEAKER_03: Uh it doesn't supply an A or replace uh an entire job, but uh it provides uh the ability to accelerate uh uh certain tasks and the ability to make decisions and find insights and find patterns, uh uh, but it's not something that uh you know in the near term is going to uh replace uh a human and human judgment.
• 26:07
  SPEAKER_03: So that's a really good observation.
• 26:09
  SPEAKER_03: Just to just to sort of finish the story.
• 26:11
  SPEAKER_03: So how did you uh uh what it just gives us the end of the the uh story arc around the first startup?
• 26:18
  SPEAKER_03: Um you you found a customer, uh they got some traction, then it did it.
• 26:23
  SPEAKER_03: Was it one customer that expanded to lots of places?
• 26:26
  SPEAKER_03: Were you able to then grow tell tell us about the the end or uh your first adventure?
• 26:33
  SPEAKER_04: Yeah, the the short version of it was we still had to shop around.
• 26:37
  SPEAKER_04: You know, I used whatever you know, business relations that I had, I used family, you know, you kind of pull all the stops, right?
• 26:43
  SPEAKER_04: When you're desperate.
• 26:44
  SPEAKER_03: Um desperation is a good motivator, by the way.
• 26:49
  SPEAKER_04: It is, it is the the initial mother of some success, I think.
• 26:54
  SPEAKER_04: Yeah.
• 26:56
  SPEAKER_04: Um, but yeah, but this time around it was a little bit easier just because of the sheer scale of things.
• 27:00
  SPEAKER_04: So we talked to maybe like 15, 16 customers, and by that point we had some good um indicators like MOUs that we had signed that uh these garage facilities where we started were going to let us do experimental setups.
• 27:16
  SPEAKER_04: We would flip the bill for everything and essentially do a proof of concept in the field for them, and we would have all the costs.
• 27:23
  SPEAKER_04: So you know, the short of it was that that that lower that do a revenue share on the delta.
• 27:30
  SPEAKER_04: Did you do a revenue share?
• 27:31
  SPEAKER_04: That's right.
• 27:32
  SPEAKER_04: That's right.
• 27:32
  SPEAKER_04: We we would do then a revenue share on the incremental uh improvement, uh not on the base, because our whole point was we would drive incremental value, right?
• 27:43
  SPEAKER_04: Um, and that just de-risked everything on the economics.
• 27:46
  SPEAKER_04: Basically, we're saying you got nothing to lose, we'll put the bill for everything.
• 27:50
  SPEAKER_04: Um, right, right.
• 27:51
  SPEAKER_04: Yeah, so that breeded a degree of confidence too, right?
• 27:54
  SPEAKER_04: You know, if these guys are that confident, you know, of course, we had to go solve back end problems like vendors and getting all our sensored manufactured and the contractors to get installation done, but all of that is what I call like no supporting some hard work if people are willing to give you a shot.
• 28:09
  SPEAKER_04: Um and then to answer your other questions, you know, how did this eventually take off?
• 28:13
  SPEAKER_04: It was um customer success breeds more customer success.
• 28:17
  SPEAKER_04: Uh you the competition, oddly, talk about they get together, you know, competition they get together to talk about, you know, what are they going to do to work together or not work together?
• 28:26
  SPEAKER_04: And you know, sometimes they get together to to criticize each other and the process news about your business kind of just grows.
• 28:34
  SPEAKER_04: And nothing's better than word of mouth.
• 28:36
  SPEAKER_04: And I guess any publicity is is better than no publicity.
• 28:39
  SPEAKER_04: So that's that's how it kind of spread.
• 28:42
  SPEAKER_03: And then when did you what was the so did it uh take it took off in China?
• 28:47
  SPEAKER_03: Did you bring it back to the US as well, or just sell it, uh finished with it in China?
• 28:52
  SPEAKER_03: Were were you still involved with it?
• 28:54
  SPEAKER_03: I don't know.
• 28:54
  SPEAKER_03: I didn't know if that was the case.
• 28:56
  SPEAKER_04: Yeah, the um so we ultimately sold the business.
• 29:00
  SPEAKER_04: Um, so the the business grew to cover you know commercial parking facilities, we did airports eventually, um, sporting arenas, uh, even train stations.
• 29:11
  SPEAKER_04: So basically anything you can measure about a physical commercial environment, we we essentially um deployed our sensors and measured.
• 29:20
  SPEAKER_04: And with that, we began to saturate most of uh Guangzhou, which is one of the tier one cities in China.
• 29:27
  SPEAKER_04: Uh, we spread to other cities like Shenzhen and whatnot.
• 29:30
  SPEAKER_04: Um, at some point, our uh largest customer, Grandview, decided that it was just more profitable for them to buy us outright, and then as they do new construction, integrate our technology into any sort of new development that they were doing.
• 29:45
  SPEAKER_04: So we wrapped it up, loaded to those guys, and I came back home.
• 29:49
  SPEAKER_03: And how many people did you what was the whatever you can share about the size, either deployments or people, or I don't know what you want, what you're comfortable sharing.
• 29:57
  SPEAKER_03: What what scale did you get it to?
• 29:59
  SPEAKER_04: Yeah, we had over seven million sensors physically deployed across Southeast China, um, probably around 2012 at that point.
• 30:09
  SPEAKER_04: Um, that gave us significant uh what I call like regional scale and data monopolization, if you will, um, in the sense that data, you know, data kind of piggybacks off of itself, right?
• 30:22
  SPEAKER_04: Like the more you have, the better.
• 30:24
  SPEAKER_04: Right.
• 30:24
  SPEAKER_04: But more importantly, it's the nature of the data.
• 30:26
  SPEAKER_04: So if we had data about parking garages, not just for say one customer, but actual parking garages for other customers, meaning our customers, right?
• 30:36
  SPEAKER_04: Then what we could essentially do is create a data map of all the vacancies, right?
• 30:40
  SPEAKER_04: Just parking garages as a segment, like this data map of where there are vacancies, where it's easy to park.
• 30:48
  SPEAKER_04: And so all of a sudden, our intelligence is more than just talking about one incident of a price point for a garage in one place, but we can do that much better by looking at across the board data in a full region.
• 31:02
  SPEAKER_04: We can also do other things, which is um, well, uh, we can now change the flexibility of the times at which you're changing your prices.
• 31:13
  SPEAKER_04: So not just adjusting, you know, adjusting the prices, but when you adjust the prices matters a lot, right?
• 31:18
  SPEAKER_04: The supply and demand will vary.
• 31:20
  SPEAKER_04: And then from there, more incremental value uh could crop up because we're looking at the interactive effects of data.
• 31:26
  SPEAKER_04: And then we can essentially sell more data-driven services or intelligence to our customers, right?
• 31:32
  SPEAKER_04: Um, you know, we might you know notify other garages that there's an overflow, right?
• 31:38
  SPEAKER_04: That's something that couldn't have been done before.
• 31:40
  SPEAKER_03: Right, right.
• 31:41
  SPEAKER_03: And at that point, I imagine you needed to use more of the machine learning algorithms to be able to make sense of that large amount of data.
• 31:49
  SPEAKER_04: We we began to experiment, um, you know, but uh to be quite frank, at that time my team hadn't been, we we were not equipped to really engage in ML just yet.
• 32:01
  SPEAKER_04: Understood.
• 32:01
  SPEAKER_04: Yeah, but um we had started tinkering with it, you know.
• 32:04
  SPEAKER_04: But then Randview came in and made a nice offer, and we said, Hey, not a bad exit.
• 32:09
  SPEAKER_03: Yeah, right, exactly.
• 32:10
  SPEAKER_03: That's wonderful.
• 32:11
  SPEAKER_03: How many employees were you up to at that time?
• 32:13
  SPEAKER_03: Do you remember?
• 32:14
  SPEAKER_03: I had over a hundred employees at that point.
• 32:15
  SPEAKER_03: Nice, nice point.
• 32:16
  SPEAKER_03: Yeah, we leveraged a lot of contractors, yeah.
• 32:19
  SPEAKER_03: Yeah, yeah, that's fantastic.
• 32:21
  SPEAKER_03: So then what did you do?
• 32:22
  SPEAKER_03: Then you had uh did you go back to then you went back to a regular job for a while?
• 32:29
  SPEAKER_04: No, no.
• 32:29
  SPEAKER_04: So up to that point, I basically told myself I have no idea how to do a normal job.
• 32:34
  SPEAKER_04: Um and uh, you know, the the title of CEO founder of a you know uh small startup or a small medium business is really confusing to a lot of companies that hire.
• 32:46
  SPEAKER_04: Totally get it.
• 32:47
  SPEAKER_03: Oh, by the way, before you go on, I just I for I forgot to mention our sponsors, CMS Wire, uh Simpler Media, uh the Marketing AI Institute, and early information science.
• 32:57
  SPEAKER_03: And I have to start putting that in my intro script because I forgot to start with that.
• 33:01
  SPEAKER_03: I wanted to throw that in there.
• 33:02
  SPEAKER_03: So I'm sorry, Paul.
• 33:03
  SPEAKER_03: Keep keep going here.
• 33:04
  SPEAKER_02: We'll we'll pre-record it so we don't have to worry about it.
• 33:07
  SPEAKER_02: Yeah, right.
• 33:08
  SPEAKER_02: Yeah, awesome.
• 33:09
  SPEAKER_04: It's awesome.
• 33:10
  SPEAKER_04: Your sponsors must be must be you know very privileged to get that plug.
• 33:16
  SPEAKER_04: Um I'm sorry, I I don't know.
• 33:19
  SPEAKER_03: So you were saying, so you were saying that uh that being a founder and a CEO is a hard um credential to kind of bring back into other organizations.
• 33:27
  SPEAKER_03: So what did you do then?
• 33:29
  SPEAKER_04: Yeah, yeah, it's it's strange because I think I was out of college for five years.
• 33:33
  SPEAKER_04: The exit was relatively large because China scale is large, and people just had a hard time understanding where this guy fits.
• 33:40
  SPEAKER_04: Is he like a manager?
• 33:41
  SPEAKER_04: Is he uh individual contributor?
• 33:44
  SPEAKER_04: Is he a director?
• 33:45
  SPEAKER_04: Is he a manager?
• 33:47
  SPEAKER_04: You know, it was it was very confused.
• 33:48
  SPEAKER_04: So I was like, you know what?
• 33:50
  SPEAKER_04: Um I'm okay taking a break.
• 33:52
  SPEAKER_04: So I went to go to business school because I saw these brochures.
• 33:57
  SPEAKER_04: Um, you'll find that Yavite is kind of like a thread of mine.
• 34:00
  SPEAKER_04: I saw the brochures, I believe what I see in brochures, and I they're all happy, all these students, right?
• 34:06
  SPEAKER_04: They're either holding glasses of wine or you know, they have their nice uh graduation gowns.
• 34:11
  SPEAKER_04: Everybody's got like a pearlescent smile.
• 34:13
  SPEAKER_04: So I said, Hey, that looks good.
• 34:14
  SPEAKER_04: I'd like to go do that.
• 34:16
  SPEAKER_04: Um, so I applied to business school.
• 34:18
  SPEAKER_04: Um I want to be in a brochure.
• 34:21
  SPEAKER_03: Yeah, what's that?
• 34:22
  SPEAKER_03: I want to be in a brochure.
• 34:24
  SPEAKER_03: Well, you know, you you've you've got the life feeling.
• 34:29
  SPEAKER_04: Um yeah, so I I I saw the brochures and I was, you know, my father is in education.
• 34:34
  SPEAKER_04: He actually teaches uh business school at uh St.
• 34:37
  SPEAKER_04: Louis University.
• 34:39
  SPEAKER_04: So I said, Hey, you know, you you teach you know MBAs.
• 34:43
  SPEAKER_04: What's your take on this?
• 34:44
  SPEAKER_04: And what he said to me was really simple.
• 34:46
  SPEAKER_04: He's like, If you if you really want to go learn stuff, go get a PhD.
• 34:50
  SPEAKER_04: You know, and uh, you know, you learn an undergrad, you learn in a master's program, but an MBA is probably not where you go to learn like deep knowledge about anything.
• 34:59
  SPEAKER_04: I was like, Well, it's obviously a good, you know, good business.
• 35:03
  SPEAKER_04: There's students paying for this stuff.
• 35:05
  SPEAKER_04: And he said, uh, well, you know, what you go there is to pay for the privilege of knowing really great people and building that network.
• 35:13
  SPEAKER_04: And I was like, all right, that sounds fun.
• 35:15
  SPEAKER_04: So we did that for a couple of years, and um, you know, then you go in through the whole notion of like MBA to you know job workforce recruiting process, and there's a very standardized protocol for how that works, right?
• 35:27
  SPEAKER_04: HR just goes to these schools and they're like, hey, interview and whatnot.
• 35:31
  SPEAKER_04: So there's a standard practice there.
• 35:34
  SPEAKER_02: I think the one thing too, Paul, is I want to fast forward to what you're doing today because it translates from what we were talking about before.
• 35:41
  SPEAKER_02: But before we do that, um I think it's also interesting, you know, given your parents, what they did to get you here, and then how many places you lived growing up, you don't have to go into all the details because those stories are there are eat they owe they own each one of them deserved their own time, you know, just to sit there, um, you know, in terms of just listening to the stories.
• 36:05
  SPEAKER_02: But I think it's interesting though, you were born on a cruise ship, and then you know, but your parents were trying to to figure out a better life in general, very educated people.
• 36:16
  SPEAKER_02: So, and then how many places you lived, you know, after that, before you actually formally you know went to school and stuff.
• 36:22
  SPEAKER_02: I think it's just it's it's fascinating.
• 36:24
  SPEAKER_01: Yeah, go ahead.
• 36:25
  SPEAKER_04: Yeah, I've I've lived in a lot of different places.
• 36:28
  SPEAKER_04: Um, the short version is that you know, my folks came to the US as immigrants, they gave up their careers and their life relatively like late in their life, not you know, not really super young guns, right?
• 36:42
  SPEAKER_04: Like in their teens or anything.
• 36:43
  SPEAKER_02: Well, but double click on that because your mom, um, what her occupation was and what your dad's occupation were fascinating.
• 36:49
  SPEAKER_02: Because it's not just a for me, it's it's kind of mind blown because it's not just that they gave up their careers, but their careers were significant.
• 36:56
  SPEAKER_02: Go ahead.
• 36:56
  SPEAKER_02: Sorry, yeah.
• 36:58
  SPEAKER_04: No, that's a that's a fair point.
• 36:59
  SPEAKER_04: They so my mother was a surgeon uh in China, um, and my father was a translator um at the embassy.
• 37:09
  SPEAKER_04: Uh I don't remember which one, which embassy, but um yeah, they they I think they had really happy life in in China, but ultimately what they found was there were certain cultural constraints, right?
• 37:23
  SPEAKER_04: Um, China wasn't fully open at the time.
• 37:25
  SPEAKER_04: So, you know, eventually they they uh my father actually came here first for his PhD because that's what he wanted to do.
• 37:32
  SPEAKER_04: He wanted to learn he's a he's an intellectual.
• 37:34
  SPEAKER_04: You'll you'll find that you won't get a lot of that from me.
• 37:36
  SPEAKER_04: Uh my mother is also an intellectual, she wanted to get her PhD, and so they were like, Well, we'll we'll just come to the US because this is a land of you know, the free.
• 37:46
  SPEAKER_04: This is a land of opportunity, this is a land where people are encouraged to change.
• 37:50
  SPEAKER_04: Um, and so you know, they're they're they came here.
• 37:54
  SPEAKER_04: Uh, you know, I think they're some of the most staunch patriots in the US, if I may, because I think they understood the contrast of what it's like not to be in the US for so long.
• 38:04
  SPEAKER_04: And when they came here, the level of appreciation they had for our culture, our attitude, I think is really important, um, and our spirit for innovation and change.
• 38:14
  SPEAKER_04: Um, I think a lot of those things kind of translated into uh me subconsciously.
• 38:18
  SPEAKER_04: Um, which is why I told them I said I'll do entrepreneurship.
• 38:21
  SPEAKER_04: They're like, okay, like they didn't bat an eye at that.
• 38:23
  SPEAKER_04: They're like, what do you need?
• 38:24
  SPEAKER_04: Right.
• 38:25
  SPEAKER_04: Like if we didn't know the language, then we can come here and you know do what we do.
• 38:29
  SPEAKER_04: Uh, you know, then you schools trust us to teach ha ha jokes on them, then you know, you can't be too far off.
• 38:36
  SPEAKER_04: So that was it just eventually probably drumming a lot too long here, but uh to answer your question, Chris.
• 38:42
  SPEAKER_04: I I lived in in various places like Japan, New Zealand, and Australia because my mother first came here.
• 38:48
  SPEAKER_04: Um, she gave up career as a surgeon, she didn't know English, and she basically served as an au pair to a very kind and wealthy family who had essentially taken us in.
• 38:59
  SPEAKER_04: And they had a private network whereby they, you know, did student exchange or familial exchanges for their students or or or children across the world, and kind enough to basically subsidize me to be part of that, which is why I ended up living in all these different parts of the world, um, which turned out to be this incredible education.
• 39:20
  SPEAKER_01: Yeah.
• 39:21
  SPEAKER_02: So let's let's fast forward to to today, because you know, you're in this spot where it's a common theme.
• 39:28
  SPEAKER_02: We're now, and you you you you said something, and I think that is the the broad theme of this this whole thing that's centered around now with uh with uh these concepts of AI and data, you know, as prolific as it is, how do you make it easier for, let's say, the common folks that don't know AI to take all their data and now apply it to actually utilize these models to actually generate the outcomes and output, right?
• 39:56
  SPEAKER_02: Because that's essentially what we're doing is it either if it's a really hard concept for them to understand, you know, going forward, but now you know taking through your thinking around how how do you make it easier for somebody that's not in this space to to utilize these these types of approaches?
• 40:12
  SPEAKER_04: Yeah, I think um that's a really good question.
• 40:15
  SPEAKER_04: So we're I still think we're like at the you know the beginning of this entire journey.
• 40:22
  SPEAKER_04: Um, yeah, I'm sure my view is not unique, but at the same time it's it's debatable.
• 40:28
  SPEAKER_04: I think we're on a journey where we're still heavily reliant on folks who have relatively personal expertise to apply uh AI, right?
• 40:38
  SPEAKER_04: Um, and that's that's natural, but it's also in some senses like a tragedy because it means that some of the most powerful tools that are at our disposal are really not being put into uh the hands of you know, you know, even a person like myself, because I don't come from an ML background.
• 40:56
  SPEAKER_04: But but that type of deprivation is severely limiting because there's all this unknown value that could be excavated.
• 41:04
  SPEAKER_04: Um, that's part one.
• 41:05
  SPEAKER_04: Part two is like, well, you're tortured if you know there's value, but you can't apply the proper tools to excavate it.
• 41:11
  SPEAKER_04: So that's a different type of torture.
• 41:13
  SPEAKER_04: Um, and then the third type of torture is like, well, you're you try to be a sharp guy or gal, you try to use those tools, and guess what?
• 41:20
  SPEAKER_04: They're not designed for folks like say you or me, or in this case, probably myself.
• 41:24
  SPEAKER_04: We have a misapplication and we derive the wrong value out of that, you know, analytical process.
• 41:30
  SPEAKER_04: And so that's why I think it's so important that we, you know, I don't know if I love the term democratize because it can be misconstrued, but I like it for some reasons, which is truly making it really easy to use, right?
• 41:42
  SPEAKER_04: Um, also like taking the PhD-ness, if you will, out of that process.
• 41:47
  SPEAKER_04: Um I always go back to like, you know, when I have a conversation with my parents about a certain subject matter and they they're they go into PhD mode, I say, well, that's wonderful if you're along a circle of PhD, say an ML, whatever domain.
• 42:01
  SPEAKER_04: But as soon as you want to connect with a layman like myself, as soon as you want to connect with someone who's more practically oriented around business value, you lose me.
• 42:09
  SPEAKER_04: And if you can imagine that same analogy happening um, you know, when in the application of AI or ML, then that's a tremendous blocker, right?
• 42:18
  SPEAKER_04: Tremendous blocker to value, accessibility, and transparency.
• 42:22
  SPEAKER_04: So my most of my work recently has surrounded around, okay, if you're interested in speech recognition, can it just be plug and play?
• 42:29
  SPEAKER_04: Right, Chris, you know this when we worked at AWS.
• 42:32
  SPEAKER_04: Um, you know, when I was at Snowflake, the question is, okay, if you're a business analyst, you work for, I don't know, name your name your big insurance company or something like that, you wanted to understand some trends about insurance uh, you know, uh behaviors or patterns.
• 42:48
  SPEAKER_04: Today you commonly rely on an internal expert who is an ML specialist to go crunch data, do all this fancy modeling with the algorithms, you're relying on them.
• 42:58
  SPEAKER_04: But as an analyst, you have a lot more domain knowledge about actual cars, actual human behavior.
• 43:05
  SPEAKER_04: And what this means is that your job is now slowed down because you have a dependency on someone else.
• 43:11
  SPEAKER_04: What if we can put those ML tools in your hands so you didn't have this interdependency?
• 43:15
  SPEAKER_04: Then your ML scientists can be freed up to do really cutting edge RD, which they love, by the way, because they don't want to do a lot of mundane data crunching.
• 43:24
  SPEAKER_04: And then for you as an analyst, you can get to immediate value realization because you're just treating this as a tool.
• 43:30
  SPEAKER_04: Like we're we're taking the mythos and the mystery out of ML.
• 43:35
  SPEAKER_04: And I think that's what's really exciting.
• 43:37
  SPEAKER_02: I think the is it's leveling the playing field so that you've got your below the budget line and above the budget line folks who are thinking about the value of this driving the business further.
• 43:50
  SPEAKER_02: PhDs don't care about that necessarily.
• 43:52
  SPEAKER_02: What they're what they care about is how do I actually create a more efficient model?
• 43:56
  SPEAKER_02: How do I get more accuracy?
• 43:58
  SPEAKER_02: How do I get more you know production?
• 44:00
  SPEAKER_02: Predictable outcomes, things like that, which do not, you know, which somebody then has to take and translate into here's the business value that that's going to derive in order for me to accomplish X, Y, and Z.
• 44:10
  SPEAKER_02: And I think that's the key too is not only have you, you know, if you want to say democratize, but level the playing field so that we're taking this concept idea and applying it to really provide now the application of business value that people can actually bet a business on, make this you know, insightful decisions on, and actually drive a business further, as opposed to, well, it's super kick ass that you've got me or more accuracy.
• 44:36
  SPEAKER_02: But at the end of the day, I'm not asking the right question of what does that get me value-wise in the company?
• 44:40
  SPEAKER_02: Because otherwise, I'm just spending a lot of investment into this thing to do what, you know, to actually drive me further, whatever.
• 44:48
  SPEAKER_02: And I think that's what you're what the core part of IC is how to take it from this great conceptual idea, and now it's gonna make a ton of sense in shortening the value, you know, the you know, the time to value for a good organization.
• 45:01
  SPEAKER_04: Yeah, yeah.
• 45:02
  SPEAKER_04: So one thing I'd add to that, which is you bring up a lot of interesting points.
• 45:07
  SPEAKER_04: Um, it's one reason why I like hang out with the over margaritas all the time.
• 45:11
  SPEAKER_02: Um we've we've created the best business models that we'll never go anywhere over over margaritas.
• 45:18
  SPEAKER_04: Yeah, anyway, um, but but one thing you know that you jog my brain on is there's if you if you if we use a calculator as an analogy, right?
• 45:33
  SPEAKER_04: Um, I I think in the future what's gonna happen is your ML scientists, if you will, are the guys who design that calculator to make it better, right?
• 45:43
  SPEAKER_04: Better and better and better.
• 45:44
  SPEAKER_04: You've heard of the TI, you know, Texas Instrument 83s and the 83 plus and 89s and so forth and so forth.
• 45:50
  SPEAKER_04: So using an analogy, they're in the future, they're gonna be focused and more happy doing that because that's what they're trained to do.
• 45:55
  SPEAKER_04: They're experts at building that type of stuff.
• 45:57
  SPEAKER_04: But then there's like the accountants, the mathematicians who don't want to be bogged down in the mechanics of how a calculator works, but they know how to apply the calculator really well, right?
• 46:07
  SPEAKER_04: Like they'll know when to use a square root versus a quadratic versus something else, even if they don't think about the mechanics of how the quadratic is solved by the calculator.
• 46:16
  SPEAKER_04: And I think that's really the world that we will we should aspire to because that's what specialization does.
• 46:22
  SPEAKER_04: Specialization allows us to be really good at what we do for the maximum ROI based on our expertise, our training, our education, you name it.
• 46:31
  SPEAKER_04: Um, and now what we're facing, Seth, and I think we we kind of touched on this earlier, is facing this strange cost function.
• 46:38
  SPEAKER_04: And that cost function is if you want to be relevant in the workforce, a lot of times what you find is, hey, do you have any AI, ML, or data analytics experience?
• 46:47
  SPEAKER_04: The problem with that statement is well, how do we qualify what experience is in that domain?
• 46:52
  SPEAKER_04: Is it the application or is it in the design of algorithms and research of that?
• 46:56
  SPEAKER_04: Right.
• 46:57
  SPEAKER_04: We haven't answered that question well, which is why it brings a lot of nervousness to the extended conversation of what's going to happen to the workforce?
• 47:06
  SPEAKER_04: Are we all just going to be replaced by robots?
• 47:08
  SPEAKER_04: I think the answer is no, because we need more people in application rather than design.
• 47:13
  SPEAKER_02: Yeah.
• 47:13
  SPEAKER_03: I think the long answer to that is also no.
• 47:16
  SPEAKER_03: You know, that we do need that.
• 47:19
  SPEAKER_03: And it and but you bring up some really interesting points because um when you think about the layers of complexity and the layers of uh of abstraction, you know, at the deepest level are the you know, the uh uh the uh engineering and and mathematics and the deep uh algorithmic uh understanding of of fine-tuning the hundreds and hundreds of machine learning algorithms that are out there and using hyperparameters and really refining those things.
• 47:51
  SPEAKER_03: And that's that uh that that uh computer science engineering, deep uh data and uh expertise and knowledge uh and programming knowledge and so on.
• 48:04
  SPEAKER_03: And then there's another layer above that, right, which is taking those algorithms and maybe you know abstracting certain hyperparameters so that they can be better tuned to a particular application.
• 48:15
  SPEAKER_03: Then there's a layer above that, which is getting into the uh primitives and the components that can then be assembled into an application.
• 48:23
  SPEAKER_03: Then there's a layer above that, you know.
• 48:25
  SPEAKER_03: Again, maybe I'm I'm being too much of a splitter versus a lumper here, but say now we can take those building blocks of business functionality and assemble them into an application of some sort.
• 48:37
  SPEAKER_03: And then there's really the fine-tuning of those business processes and the integration of multiple algorithms and multiple applications across that ecosystem.
• 48:46
  SPEAKER_03: Do you see those levels of uh abstraction or what's your sense of that?
• 48:52
  SPEAKER_03: I mean, I've looked at some of the uh AWS uh suites of tools and platforms, and it seems like you can come in at any of those levels of granularity and detail.
• 49:03
  SPEAKER_03: And the question is, where should you be playing and where should the organization be playing?
• 49:08
  SPEAKER_03: If your job, if your goal is to is to competitively differentiate yourself by having a bit better speech recognition capability, well, you know, now you're playing at that very deep level, but is that really your competitive advantage?
• 49:21
  SPEAKER_03: Is that where you're what core to your business, or you just say, you know what, that's fine, that's commoditized.
• 49:26
  SPEAKER_03: I'm gonna use that service, I'm gonna pay for it, I'm gonna build my value at the next level above that.
• 49:32
  SPEAKER_04: Yeah, yeah.
• 49:33
  SPEAKER_04: These are uh these are really sophisticated questions.
• 49:37
  SPEAKER_04: Um I uh I think the one answer would be um it's a build versus buy conversation.
• 49:46
  SPEAKER_04: Um it seems like a simple answer, but here's a here's a short rationale for why.
• 49:51
  SPEAKER_04: It comes down to whether you have differentiating data.
• 49:56
  SPEAKER_04: If you're a data monster, like the Googles of the world, right?
• 50:01
  SPEAKER_04: The Amazons of the world, you name it, then you're going to want to build because whatever you build will be differentiated because the data gives you that competitive edge.
• 50:11
  SPEAKER_04: If you're a company who really doesn't have access to very differentiated data in your domain or in your competitive space, you might be better off buying and then tweaking.
• 50:22
  SPEAKER_04: Um, I think if you have to make it simple, that's probably the approach that I would look at the world.
• 50:29
  SPEAKER_04: Um ultimately, I think there's a lot of confusion across the value chain, across this tech stack.
• 50:36
  SPEAKER_04: And that's going to be disastrous.
• 50:39
  SPEAKER_04: If you're not an ML-driven or data-driven company and you pretend to be and not know your identity, you're likely to drive a very, very tough roadmap, and vice versa.
• 50:50
  SPEAKER_03: And you know, it's funny because um uh, you know, we were uh uh dealing with uh, you know, thinking we're dealing with an organization that is trying to build out some capabilities that that were to better serve the customer.
• 51:06
  SPEAKER_03: They do have differentiated data.
• 51:08
  SPEAKER_03: It's their first party data about their knowledge of their customer, right?
• 51:11
  SPEAKER_03: That's what they are competing on because they understand that customer and they understand their needs and so on.
• 51:17
  SPEAKER_03: But you know, now the question is at what level of that tech stack do they try to leverage that?
• 51:23
  SPEAKER_03: You know, what what building blocks do they use?
• 51:26
  SPEAKER_03: Because, you know, when I think of this, I think of in the cognitive AI space, you know, that last mile of cognitive AI is the organization's knowledge, right?
• 51:35
  SPEAKER_03: It's how they compete on their knowledge.
• 51:38
  SPEAKER_03: And so you can have uh lots of off-the-shelf tools or platforms that you can rent from you know the big providers, but your differentiator and your competitive advantage is really that knowledge layer, right?
• 51:51
  SPEAKER_03: Because that's what you're competing about competing on.
• 51:54
  SPEAKER_03: So I think there's a certain level of this that every organization is going to need to own because it is their competitive advantage.
• 52:01
  SPEAKER_03: And you can't outsource your competitive advantage, right?
• 52:04
  SPEAKER_03: You have to be able to have that capability.
• 52:06
  SPEAKER_03: But then the question is at what level of granularity, right?
• 52:10
  SPEAKER_03: Where do they need to, you know, hire what what skill set do they need to hire in their organization versus what skill set can they rent, you know, and and outsource?
• 52:21
  SPEAKER_03: And I think that's really one of those questions of where are you trying to uh get the greatest value from your data and where does that competitive advantage come from?
• 52:32
  SPEAKER_03: So I I think it's it's it's really uh it's a build versus buy, but it's like build versus buy at what level, right?
• 52:39
  SPEAKER_03: Yeah, at what level?
• 52:41
  SPEAKER_03: Because you have to build something at one level, but you have to buy the components, you're not gonna reinvent everything, right?
• 52:47
  SPEAKER_03: Right.
• 52:47
  SPEAKER_04: Right.
• 52:48
  SPEAKER_04: Yeah, I I think you, you know, to your point, said you my thought is you would buy the commodities, right?
• 52:54
  SPEAKER_04: Yeah, exactly.
• 52:56
  SPEAKER_04: You buy the commodities, like general compute, uh, you know, network and storage are commodities, they're like infrastructural stuff.
• 53:05
  SPEAKER_04: Um, you might do the mix and match a little bit higher up the stack, right?
• 53:09
  SPEAKER_04: When you get to the analytics layer, and then like you might have very, very niche differentiated capabilities and knowledge and data and experience.
• 53:18
  SPEAKER_04: That's where you might have to build.
• 53:20
  SPEAKER_04: Like we we as we climb this conversational chain, you might have to configure or configure, yeah.
• 53:26
  SPEAKER_04: Or configure, right?
• 53:28
  SPEAKER_04: Well, no, maybe somebody else just suddenly doesn't have all the recipes you need because as we go higher up that curve, the more specified knowledge and experience you're required to have, and the commodity sellers will not have that knowledge, exactly.
• 53:40
  SPEAKER_03: So you're compelled to build or configure, like you said, right?
• 53:43
  SPEAKER_03: A generalized language model is not going to apply to a life sciences technology firm or you know, industrial manufacturer, right?
• 53:52
  SPEAKER_03: It has terminology.
• 53:54
  SPEAKER_02: I think I think the uh the key here, Seth, is that if there's big organizations that are looking for questions, they should just talk to early information sciences and uh just plug in, right?
• 54:06
  SPEAKER_02: Because you can help them in a big way.
• 54:07
  SPEAKER_02: Listen, um as always, thanks so much, man.
• 54:12
  SPEAKER_02: Much love, brother.
• 54:13
  SPEAKER_02: Appreciate everything that you do.
• 54:15
  SPEAKER_02: And so I, you know, part of this is I love having my friends on here.
• 54:18
  SPEAKER_02: So thanks for uh podcast with us.
• 54:21
  SPEAKER_02: Well, thank you.
• 54:22
  SPEAKER_03: I appreciate it, Paul.
• 54:24
  SPEAKER_03: It it's been great.
• 54:25
  SPEAKER_03: I love hearing the stories.
• 54:26
  SPEAKER_03: I wish we had more time, and hopefully we can have you back and uh dive into some other areas because this is a great conversation.
• 54:34
  SPEAKER_03: Uh, I do want to thank our sponsors again, uh, Marketing AI Institute.
• 54:38
  SPEAKER_03: Uh, they have an upcoming conference that's really terrific.
• 54:41
  SPEAKER_03: We have uh Simpler Media, which is CMS Wire, and of course, Early Information Science, which is our company.
• 54:47
  SPEAKER_03: But thank you very much, Chris.
• 54:50
  SPEAKER_03: Thank you very much, Paul, and our producer, uh Sharon Foley, for all the uh work she does to keep the trains running and uh keep things on track.
• 54:59
  SPEAKER_03: Uh with that, we'll uh we'll let you all go and uh look forward to talking next time.
• 55:08
  SPEAKER_00: Thanks for listening to the Early AI podcast with your host, Seth Early and Chris Featherstone.
• 55:14
  SPEAKER_00: We hope you enjoyed the program and took away some ideas that you can start to implement in your organization right now.
• 55:20
  SPEAKER_00: If you need help, check out www.early.com for information about early information science and services.
• 55:29
  SPEAKER_00: And of course, check the show notes for links to any resources we mentioned today.
• 55:33
  SPEAKER_00: That's all for now.
• 55:34
  SPEAKER_00: See you next time.