The Humankind of AI and Data-Driven Machine Learning Ft. Stephen Sklarew and Tim Oates

Mark Richardson:

Hey, everyone, welcome back to another hang in the data basement. I am your host, Mark Richardson.

Stephen Sklarew:
I’m Stephen Sklarew.

Tim Oates:
And I’m Tim Oates.

Mark Richardson:
And we are on another episode of The Data-Driven Marketer. Welcome back.

Mark Richardson:
Today, we have Stephen and Tim from Synaptiq, a Portland based AI startup. Really excited to have you fellows here with us today, as we are kind of exploring this merging of data, marketing, creativity, and tech. You guys are the co-founders of Synaptiq, correct? Why don’t you give us a rundown of your bio?

Stephen Sklarew:
Sure, sure. Yeah. So I started my career in the mid ’90s. Actually dropped out of a master’s degree in biology. I was planning on spending most of my life on rivers, sampling fish, and insects, and things like that, and got hurt and switched careers into technology. Worked at Ernst & Young, lived in New York city for a while, jumped a couple of startups and big companies. And in 2015, as a technology leader, started Synaptiq with Tim, focused on AI.

Mark Richardson:
And Tim, what led you to that point?

Tim Oates:
Well, yeah, so I’m actually an academic by training. I got undergraduate degrees in electrical engineering and computer science. I added computer science because I found in every double E lab I would produce smoke in the circuits, which not good.

Tim Oates:
So went to work in the industry for a little while as a programmer. Got a PhD in 2001 in computer science, doing AI and machine learning. I was actually looking at how robots can learn grounded language, if you’re sort of talking to them, and they’re seeing things in the environment. Did a postdoc at MIT for a year in the AI lab. Then, got a job as a professor and I’ve been there ever since.

Tim Oates:
A long time ago, I connected up with Stephen at another company that he was at. Was their chief data scientist for a while, and then, at some point some number of years ago, we just started chatting again. Enjoyed working together back in the day, and thought it would be fun to start doing services together.

Mark Richardson:
And maybe if you could, for listeners who aren’t fully aware, could you talk a little bit about… I’m wondering if there’s a way we can, if you can do it briefly, but as best can, tell people what AI and machine learning is. And maybe then extrapolate on how Synaptiq started, and the way you started helping brands tackle that space.

Tim Oates:
So how about I’ll go the AI ML route, and then I’ll hand it over to Stephen?

Tim Oates:
Yeah, so artificial intelligence has changed, right? So when I was younger, artificial intelligence was the study of how you can get machines to do things that we think are typically associated with human cognition. So think about planning a route from point A to point B, engaging in a conversation like this one, and solving optimization problems. Right? So how do I get my trucks to deliver all the goods in the most efficient way?

Tim Oates:
And machine learning was a subfield of that for quite a long time. And machine learning is the study of how machines can get better at some tasks with experience. And I actually stole that quote from a textbook by Tom Mitchell. So you and I learn lots of stuff. We learned to walk, talk, solve math problems, and swing a golf club. And all of that is done through some combination of just trying the activity, but then also people telling us how to do that.

Tim Oates:
So machine learning was a subfield of AI for quite a long time. But what I would say is that machine learning has eaten AI. So these days, if you have a person who says they’re doing AI, what they really mean is that they’re doing machine learning.

Mark Richardson:
That’s what I wanted to kind of get at. It feels like the two have become very conflated, especially in the marketing and ad space. So I wanted to kind of set this as a foundation for what we’re talking about, to kind of decouple the two topics as best we can.

Tim Oates:
But again, in the sort of data-driven world, it’s all about, “Give me data,” which you can think of like the experience, “And I will get a machine to solve some problem.”

Tim Oates:
So when should I send this email? How should I formulate this subject line for a marketing message? Sort of any number of things. But yeah, it really is the case, I think most companies when they say they’ve got AI or machine learning for a marketing tool, for example, mean they’re doing data-driven machine learning.

Stephen Sklarew:
And so I think a follow-up was how we got into this space as a company.

Stephen Sklarew:
So I didn’t go into all the details of my background, but since 2002, I was involved in running sales and marketing systems and building sales and marketing products. And so when Tim and I first started the company it was a side gig for both of us. And ultimately, we started getting work that substantiated becoming a full-time job.

Stephen Sklarew:
And very quickly I was focused on, how can I use this technology to solve problems in sales and marketing? We had a number of projects, and even some product development efforts we did, around customer segmentation. I had experience as a product manager for about 10 years. And one of the challenges there was, how do we figure out where to direct our messages? Right?

Mark Richardson:
Yeah, exactly. That’s a big existential question, isn’t it?

Stephen Sklarew:
Exactly, exactly.

Stephen Sklarew:
And so my experience there as a product manager in a big company was to look at revenue, and industry, and maybe number of employees. And that’s a segment. And we knew that was way too coarse to really have a targeted message.

Stephen Sklarew:
And so Tim and I, and actually our CTO at the time came up with this idea of, maybe we can use all these interactions that we have with customers through email, and maybe derive customer segments by looking at the engagement of those interactions. And so Tim built some models. Our CTO and I kind of came up with the approach, and we started pitching it as a potential solution to companies.

Stephen Sklarew:
And along the way, realized that the same sort of tool could be used to kind of derive optimal subject lines. And so we kind of modified our prototype and started pitching that. And we got a couple of customers to pilot it, I think one or two that actually bought it. But that was kind of our first foray into the marketing world with this.

Mark Richardson:
So your pitch was basically, it was, “Run this list of inputs, run this list of subject lines. And our program will intuit, will gather input data, or gather engagement data on, and return to you the most optimal subject line.” Is that sort of how you would position it?

Stephen Sklarew:
Yeah, that’s kind of where it ended up at the end. Initially, it was, “Give us all the interactions you’ve had with your customers in, let’s say, Marketo or HubSpot, and we’ll identify customer segments.” Like, “Here are different types of customers that are distinct based on their interactions with those emails.”

Stephen Sklarew:
But then yeah, maybe Tim, you can talk a little bit about the subject line piece that came out of that.

Tim Oates:
Yeah. So as part of that, what we did was, we were looking at emails that were sent. So we’ve got the subject lines of the emails, who they were sent to, and whether they opened it or not. So we had open information. Also information on whether they clicked on links that were inside the email.

Tim Oates:
And so what I did was built some fairly simple natural language processing or NLP models, that would use the subject line to try and predict what the open rates were going to be. Right? And so if you can do that, you can do a couple of things.

Tim Oates:
One is, if I have a model that given a subject line will predict an open rate. Then you’ve got a tool that will allow me to write a subject line, hand it to the model, and it’ll say, “Well, your open rate is going to be,” I don’t know, “4%.” Right?

Tim Oates:
But then, you can start tweaking the subject line. And you’re trying to make changes to the subject line, such that the model tells you it’s going to have a higher and higher open rate. Right?

Mark Richardson:
Right.

Tim Oates:
So it’s this sort of, again, like very data-driven method for doing almost like A/B testing, but in silica, right? So I’m just doing computer experiments, as opposed to sending them out to people that are alive.

Tim Oates:
The other thing that came out that, which I thought was interesting, was we would do an analysis. Because basically it’s just using words and combinations of words, to try and understand, how will those lead to differential open rates?

Tim Oates:
So what we could do is, we could examine the model and say, “Well, what do you think about these individual words?” For example. Or you could look and say, “Does the structure of the subject line matter?” So for some clients, we had one in particular that had these unbelievably long subject lines for their emails. And we found that the shorter it got, the higher the open rate was. I mentioned-

Mark Richardson:
Yeah, imagine that.

Tim Oates:
Exactly. Right. So things like that. And being able to say, “Ah, well, if you use this phrasing, as opposed to that phrasing, it seems like it works better.” So that’s where we wound up. Which is, again, was sort of a super interesting exercise for me, because I got to… I don’t know a whole lot about marketing, but I know how to deal with data about marketing.

Stephen Sklarew:
Yeah. I would say another interesting thing was, a lot of the guidelines around how to write subject lines and emails in general are broad, and may not apply to a certain customer’s audience.

Mark Richardson:
And if you’re able to back a recommendation with data, I mean, help, you can just go, “Look, we’ve run 40,000 different permutations of this subject line. And we can tell, hey, this is… That shorter, with no ellipses or whatever it is, within… Keep it condensed to the visual space, et cetera.”

Mark Richardson:
This is the sort of thing… I mean, I come at this less from a data science side, I’m more from sort of an audience segmentation, activation, and clustering side. So a lot of my sort of overlap with machine learning comes from my managing of the ad platforms. I think a lot of the logic that you’re describing, I’m thinking about in terms of, in terms of what Google is doing in their option systems. Have you worked with any of the major ad platforms like Amazon, Google, Facebook?

Stephen Sklarew:
No. No. I mean, ironically, so we spent, I don’t know, maybe six months to a year in this space, and learned a lot.

Stephen Sklarew:
And one of the things we learned, from a product perspective, is there’s 5,000 other marketing SAS software companies. And while we were excited about what we were doing, we decided that it didn’t make sense to invest further in this, unless there were clients that would hire us in more of a service relationship. So we moved on to other industries and did lots of different other type of work.

Stephen Sklarew:
But it was interesting, a lot of the agencies we spoke to, we did a kind of product research with 60 different individuals in marketing roles. A lot of the agencies saw value here as a way to help their clients.

Mark Richardson:
That’s awesome. I was reading a bit of big news that you guys are partnering with Microsoft to solve central line infections. Could you talk a little bit more about that? That sounds really cool.

Stephen Sklarew:
Yeah. Yeah.

Stephen Sklarew:
So through this journey, we’re almost seven years in, Tim and I realized over time, that what we really love is to build things that make the world a better place, that help people.

Stephen Sklarew:
And so therefore, like we’ve been doing a lot of work in the healthcare space with lots of different, primarily healthcare technology companies, that build solutions to help healthcare providers and their patients. And one such opportunity opened up with Microsoft. They have many hospitals that use their technology, and they spend a lot of time engaging with chief medical officers and CIOs.

Stephen Sklarew:
And one of the problems they heard was that there are many people that get these infections when they have central lines attached after surgery. And while there are guidelines to help improve the compliance of the dressings around these central lines, it’s very difficult for them to stay on top of every patient. Right? So particularly with COVID, there were a lot of patients that had to get respirators, and then ended up having to get these direct central lines.

Stephen Sklarew:
And therefore, there seemed to be an opportunity to build software to help improve the workflow around checking compliance of these dressings. And we’ve been working with Microsoft on that, building a solution, and we’re getting close to piloting it in a hospital coming up here soon.

Mark Richardson:
That’s really exciting. That’s very cool. Are you able to share what hospital that is that’s going to be taking place?

Stephen Sklarew:
Unfortunately, not yet. I’ll just say it’s a big system in the Southeast.

Mark Richardson:
Very cool. Well, we’ll keep under NDA here. But congratulations. We definitely want to keep abreast of that. We’ll be following you guys on Twitter, and keep our listeners aware of those developments. That sounds really exciting.

Stephen Sklarew:
Yeah. Thank you very much. Yeah. Obviously, saving people from infections, reducing hospital costs, really important.

Mark Richardson:
Were you connected with the hospitals through Microsoft themselves, or was this something where you had relationships?

Stephen Sklarew:
Yeah, we take a two prong approach in our marketing and sales. Obviously, if we have partners like Microsoft, they add a lot of credibility, and they have great contacts. This particular opportunity came through them, but we also market and sell directly. So we’re working on some things like that in the Northeast with our own clients.

Mark Richardson:
Got it. I wanted to poke in, there was one bit of your background, Tim. One of the fields you… I [inaudible 00:14:01] it was, machine vision. Can you explain? Can you delineate from machine vision from machine learning? Or if that’s, for our listeners who may not understand, if that’s a component of machine learning? I don’t fully understand myself what that is, and how that contributes to your work.

Tim Oates:
Right. So, yep, and I’ll connect it back to the project that Stephen was just talking about.

Tim Oates:
So machine vision, as it’s practiced today, is a subfield of machine learning, which is concerned with processing images and videos. Right? So a lot of the work that we do these days in Synaptiq is around machine vision.

Tim Oates:
So like here’s an example. Imagine you’re a construction company and you want to make sure that all the people who come onto a site are being safe. Right? So everybody should be wearing a helmet. In certain areas, they should be wearing face coverings. Vehicles should be driven safely, and people should be paying attention to where they are relative to vehicles. So people need to stay away from them when they’re moving.

Tim Oates:
So it’s a big issue because it’s liability for the companies. What if you had a camera that was just watching your site? Or multiple cameras? Could you go into the video, find people? And then find the people, and look at their head, and say, “Are they wearing a helmet or not?” Right?

Tim Oates:
Or can I find people and vehicles and say, “How far apart are the people and the vehicles? And are people getting too close to moving vehicles?” Those kinds of things.

Tim Oates:
So machine vision is all about processing images, still images and video. And all of that work these days, or the vast majority of that work these days is done using machine learning methods. For example, there are these things called neural networks, or deep neural networks, which everybody uses these days, where you can take a ton of images. Suppose I’ve got 50,000 images, and I just put a little box around every person in every image. You can hand that to a neural network, and it will learn to recognize people. So you can give it a new image, and it’ll put a box around every person that’s in the image.

Mark Richardson:
I see.

Stephen Sklarew:
So as it relates to the central line stuff we were just talking about. We’ll train up models that recognize dressings around these central lines, and what a good dressing looks like, and a bad dressing looks like. Train the model, and then the model can inform the care team, “Hey, it looks like you have a bad dressing. Go over there and check it out, make sure there’s no issues.”

Mark Richardson:
That’s awesome. And did anything of this sort… What were they using before? Was there any sort of solution to this problem prior to your script?

Stephen Sklarew:
Not technology. Yeah.

Stephen Sklarew:
Yeah, you could think of this as almost like an early detection system. So the goal in healthcare in particular, is to make sure that humans are in the loop. The machine is not going to replace the human. The idea is to help the humans on the ground proactively diagnose potential issues. Right?

Stephen Sklarew:
And so in this case, the solution is helping them know that, “Hey, there’s a patient that may have an issue. Break out of your current workflow, and go look at it, because it could cause a problem.”

Stephen Sklarew:
So prior to this they would send nurses, as an example, out on a regular basis to check the dressings. But given the pandemic, as an example, a lot of these nurses were completely overwhelmed. Right?

Mark Richardson:
Right.

Stephen Sklarew:
So instead of having to look at everybody every hour, now they can kind of self-select which ones are high priority to look at now.

Mark Richardson:
You feel like this needs state existed before the pandemic, and the pandemic sort of accelerated an existing concern?

Stephen Sklarew:
Yeah. Yeah. Yeah.

Stephen Sklarew:
So there are guidelines in hospitals to kind of manage the compliance around these central lines, they’re just difficult to handle when there’s a surge. So certainly, labs… Sorry. Certainly the pandemic has made it a bigger issue, but it exists after operations in all hospitals. It’s something that they need to watch very closely, because unfortunately, they can cause death. When you get a blood infection from a central line, you’re in big trouble.

Mark Richardson:
You talk about using data and technology to make the world a better place. I think that kind of hits. It really speaks to, man, just the ability to do something altruistic with your expertise.

Mark Richardson:
It kind of speaks to one of the frustrations we hit here is like, “Man, there’s all this creativity, and it’s just kind of going to power these big tech bottom lines.” So it’s really refreshing to interact with folks who are doing something that has such a positive benefit.

Stephen Sklarew:
Yeah, thank you. It definitely gives us meaning and purpose for what we’re doing every day.

Mark Richardson:
I believe it.

Mark Richardson:
Have you had any other applications regarding, whether it’s government contractors, anything else related to COVID since the pandemic hit? Have you found any weird requests from potential clients around applications for your services?

Stephen Sklarew:
So, yes. I’m not sure if I can talk about all of them. I’ll say that when COVID hit, we weren’t sure what it was going to do to our business. We had a really great end of 2019. And then coming into 2020, when March hit, we weren’t sure what was going to happen.

Stephen Sklarew:
So we decided to start a volunteer set of projects. And we got some really amazing people outside the company to collaborate with us on a variety of things that could potentially help with COVID. And so we felt like that was the best way to do our part, is to kind of use our capabilities, and do something voluntarily, and talk about it. And so we had our own kind of webinars and things around it, the summer of 2020.

Stephen Sklarew:
Unfortunately, nothing that we built ended up going to production and we can talk about it. But I will say it was very meaningful, because we met all these people around the world that wanted to work with us on this stuff. And one of those people became one of our team members long-term, who is fantastic.

Stephen Sklarew:
One other thing I’ll mention is, going back to Tim’s point about machine vision. Before COVID, we were working with a company that provides ultrasound services. And so if you’ve ever had a friend or family member that’s had a stroke, you know one of the things that’s really important is to look at your carotid artery on a regular basis using an ultrasound. And check the vascular age and the plaques in that artery. And so we had an opportunity to build models, machine vision models, to expedite that process. And it turns out that given COVID, there were a lot more people having cardiovascular issues.

Mark Richardson:
Mm-hmm. That’s-

Stephen Sklarew:
And so while we didn’t focus on the pandemic initially, because we were working on it prior, it was something that certainly has an application to the pandemic.

Mark Richardson:
Yeah, as a byproduct of people’s conditions they were… I’m sure you had more, a larger dataset, probably, if nothing else.

Stephen Sklarew:
Yep. Yeah.

Mark Richardson:
Where do you see the next, in terms of your own marketing, where do you see kind of the next, the challenges existing? Whether it’s communication, or sort of lead generation, or education and awareness. I guess, where do you see the biggest challenges and opportunities in 2022? And then maybe like five years out, where do you see the AI and machine learning going?

Stephen Sklarew:
Right, so I’ll take the first one, and then Tim, maybe Tim could take the second one.

Stephen Sklarew:
So on the marketing side, I mean, as an AI company, and you’re probably aware, like every company is becoming an AI company. Whether it’s a startup or a big company, they’re all saying, “Hey, we’re data first. We’re AI.” So we have a challenge in differentiating ourselves.

Stephen Sklarew:
And that’s why we really focused on this humankind of AI message, because that’s what we want to do. And that’s how we believe we are different. I tell folks all the time, “The challenge with marketing and selling AI is that everybody is marketing and selling AI. And everybody wants to talk about marketing and selling AI.”

Stephen Sklarew:
So our challenge is finding partners that actually want to do something with AI. And so we’ve built, by our journeys, we’ve been building campaigns that we believe are more differentiating. And, candidly, we’re still figuring it out. One of the things that we’ve learned over the last seven years is that more and more companies want to do AI, but don’t have a strategy for it. And they may end up implementing AI projects, but many of them fail. And so we are spending quite a bit of time on our side kind of productizing AI strategy, so that we can help companies in a more scalable fashion. Yeah.

Mark Richardson:
So if I just jump in there, so you’re, in a way, it’s sort of the thing companies know they need the service, but they don’t actually know why or exactly what of the service they need. So there’s a bit of education. You’re almost consulting them, kind of demystifying, and educating companies around what they already know they need?

Stephen Sklarew:
Yeah, pretty much. I mean the early days, maybe Tim could talk to this, but we used to go on site and just kind of do these free workshops and, say, teach them about machine learning and AI. And then talk about some of the pains they have in their company. And then we would discuss how AI may or may not be a solution to solve that pain.

Stephen Sklarew:
But now we’ve structured it into more of a productized offering, and we’re about to heavily market it. And we’re curious to see how that goes, because the stats seem to say that a lot of companies are struggling to figure out how to apply it. We’ve had quite a few companies that come to us and just say, “Hey, we want to do AI. Can you help us?” To us, that’s the wrong-

Mark Richardson:
That’s the way I was thinking. It was like-

Stephen Sklarew:
… The wrong way to start.

Mark Richardson:
I’m working lead gen and audience segmentation. And my background, I got to this point through SEO. Which is the classic kind of, “We’ve got a website. We need SEO. Can you come do the SEO?”

Mark Richardson:
And it’s like, “Well, what part of SEO do you need? Do you need the technical? Do you need code cleanup? Do we need web? We need vitals cleaned up? Site speed? Et cetera, et cetera. Or do you need content marketing? Do you need link?”

Mark Richardson:
When someone says, “Hey just fix my SEO. I know I need it. Just do it.” It seems like a similar, “Hey, can you help me do AI?” It’s a similar.

Tim Oates:
Yeah.

Stephen Sklarew:
Exactly. Exactly.

Tim Oates:
We’ve actually had some people come and just say, “We just need AI, because we need to be able to say on our website that we’ve got AI and not be lying about it.”

Mark Richardson:
Mm-hmm.

Tim Oates:
Right? And so when you think about where AI and machine learning are going to go, it’s interesting. We’ve talked about applying ML to marketing, but there’s a lot of marketing around machine learning and AI. I’m hoping that people get better at discriminating what’s real and what’s not real.

Tim Oates:
There’s so many times I’ll read about a company and, “We’ve got AI for this or ML for this.” And there’s like, no, there’s no AI in there. Right? It’s just, they’ve got a spreadsheet, and they’re computing statistics, or something like that.

Tim Oates:
So I would hope that people would get a little bit more sophisticated about the use of those terms in marketing. That they don’t always mean that there’s anything novel, or powerful, or interesting, or useful going on.

Tim Oates:
I also think, I mean to Stephen’s point about the humankind of AI. I mean, we’re really, as a society, trying to push data in lots and lots of different directions. So for example, making decisions about which resumes are going to get given to managers to make hiring decisions. Or, who should get parole in a prison system?

Tim Oates:
And I’m also hoping that in the next five years, we get smarter about what are good uses of AI, and what are uses where we need to be real careful, and make sure that humans are still firmly in control of things? But basically, if you look at the trajectory of the field, the thing that we’ve learned is that more data, faster compute, equals better results. Right?

Mark Richardson:
Right.

Tim Oates:
And so you’re going to see this steady march towards, “Let’s make sure we get all the data. We keep everything. We throw as much horsepower at it as we can.” And turns out, that you can actually do a lot of good in those cases. So there’s this old paper which is real famous, and the title of the paper was the Unreasonable Effectiveness of Data.

Mark Richardson:
Wow. I mean, that might have to be the… I might have to borrow that for the title of the podcast. The Unreasonable… I got to write that down. The Unreasonable Effectiveness of Data. Ooh.

Tim Oates:
Data, right. But what it showed was, there are lots of different kinds of learning algorithms, right? And what it showed was, is if you don’t have much data, some algorithms perform better than others. But if you have a boatload of data, the differences just kind of disappear. Right?

Tim Oates:
Like if I get a ton of data, I don’t have to be smart about the learning algorithm. I just pick one, and you’ll do really well. Right? So we’re really, I think, just kind of riding that wave of lots of data, lots of compute. And if you look at companies like Google, for example, the reason they’ve got the best language models is because they got all the data, right? They’ve got all your data and my data, and they’ve got more compute than anybody else on the planet. And so it’s not surprising that they can crank out a model that has a trillion parameters in it, that does a pretty good job of writing text. You say, “Hey, write a story about a boy going to school and falling in a mud puddle and…” Right? And they can do it.

Mark Richardson:
They can do it. And I think this is where it’s really compelling for me as a storyteller, a content creator. At what point does… And I’m wondering how would one, to automate, let’s say to automate an entire ad campaign from sort of soup to nuts, using deep fake rendering, CGI.

Mark Richardson:
Like how close are we to someone being able to go, to use a machine vision input, and say, “All right design me my digital ads, my YouTube ads, and my Super Bowl ad.” And all of this purely from computed data.

Tim Oates:
Right. So I will say that, in my opinion, we’re real close to having the pieces to do that. Right? We can use these things called GANs to generate incredibly realistic people. I’m sure you’ve seen news reports about that.

Tim Oates:
The challenge though, is that even if you look… So a language model is just a thing where you give it a little bit of text, and it sort of predicts what the next words are going to be. So if you say, “The cat.” Right? It’ll say, “Well, the next word might be sat,” and then, “On the mat.” Right? So you can give it a prompt, and it’ll just start writing stuff for you. And these language models these days are so good that they can write these very long coherent pieces of text.

Tim Oates:
I think the challenge is that they are based on what they’ve seen in the past. So they’re trained on data. And it’s real trite in AI circles to say, “Well, the thing that’s missing is creativity.” But on some level it is, right?

Mark Richardson:
Yeah.

Tim Oates:
So it’s the thing that you were saying when we were chatting before, about like, “We found this weird word. When you use that word it’s awesome in marketing.” Right? Like we throw that in the title and subject line and everybody’s clicking on us.

Tim Oates:
That’s not the kind of thing that a pre-trained system like that is going to discover easily. And so I think it’ll be hard. So if there are people who are really good at marketing and creating content, replacing them is going to be really hard. But replacing me when it comes to marketing is going to be easy, right? Because I’m not good at it. And I think [inaudible 00:29:28] do as good a job.

Mark Richardson:
I don’t think Stephen wants to replace you anytime soon.

Stephen Sklarew:
No, definitely not.

Tim Oates:
Well, he doesn’t want me to [inaudible 00:29:34].

Mark Richardson:
Right.

Stephen Sklarew:
Yeah. I would just add one more thing to what Tim said. I think data and AI literacy on the business side of companies is the next frontier. Right?

Mark Richardson:
Mm-hmm.

Stephen Sklarew:
I think we’ve seen, obviously, amazing things come out of the technical folks in this world. But I think as the business side of an organization learns how and where to apply AI, I think that’s the next frontier. And there’s just a large number of companies that are not early adopters in this space that have a vast amount of information, whether it’s in documents or databases, that really aren’t taking advantage of this technology right now. And that’s because the business people don’t understand it.

Mark Richardson:
So you’d say… Yeah, so the main, I guess, barrier to entry is the unknown unknowns kind of getting adoption, and getting understanding from the deciders, would you say?

Stephen Sklarew:
That’s right. That’s right. Like you said, you mentioned it earlier, is education. I happen to be a guest speaker at a lecture for an MBA program here. And obviously, the new business folks coming out of school have some of this education. But the folks that are leading large, traditional companies, salt and earth companies, many of them have no idea how to even get started, or how to even use this.

Mark Richardson:
Yeah. It’s so funny how you… Kind of, call it the circle of life, if you will. The people who are most educated and most hands dirty working with some of the emergent technologies are often the ones in the entry level roles, and least empowered to make the most impactful decisions. Even if a brand really needs that decision to be made.

Mark Richardson:
You just spoke to kind of a, probably, a real systemic issue in marketing and tech teams. Or even larger brands that aren’t in technology, like you were saying in the healthcare. Real estate, construction, any number of industries where this sort of predictive modeling could be eliminating wasteful processes.

Stephen Sklarew:
Absolutely. Absolutely. Yeah.

Mark Richardson:
Very cool. So who was the… You said, “The Unreasonable Effectiveness of Data.” We’ll link to that in the show notes. Who is the author of that again?

Tim Oates:
So Peter Norvig is one of the authors. He’s a big guy in AI who works at Google. And then, Fernando Pereira, who was a old school, been around doing natural language processing forever. And then, Alon Halevy. So just look for Peter Norvig, The Unreasonable Effectiveness of Data, and you’ll find it.

Mark Richardson:
Awesome. Yeah. This is, like I said, this is something for someone who’s been working in the vertex of creativity and algorithmic calculation through, for the last 10, 12 years. It’s interesting to see where everything has come from.

Mark Richardson:
Like I remember… I like to use chess kind of as a barometer. Back in ’97, it was, Kasparov could beat Deep Blue. And now we look at, you have things like Stockfish, and Leela, and these amazing engines that I assume have some, are based in some type of AI scripting. So just look at how far we’ve come. And I wonder where we’ll be in another 10 years.

Mark Richardson:
So as we wrap up, I guess, what are you most excited about, either in your pursuits with Synaptiq or just otherwise in the industry?

Tim Oates:
Yeah. I’ll give you an answer that’s probably not the one you’re looking for.

Tim Oates:
Yeah, so there are two things. One is, I’m always amazed when we talk to prospective clients about the ways in which they’re using data to make money. To solve problems that people care about, and make money from it. And when we have discussions with potential clients, I’m always like, “That’s a really amazing idea. I wish I had thought of that.” Right? And so just kind of helping them understand how the technology can apply.

Tim Oates:
And then the other thing that’s kind of interesting to me is, I started out, again, when I was younger. I was very much a, what’s called a strong AI person. Meaning that I believe that computer systems will eventually be equally intelligent as humans. Right?

Tim Oates:
There’s weak AI people who say, “Ah, AI is cool. It can do stuff, but they’ll never be as smart as humans.”

Tim Oates:
And when I was younger, I’d say, “It’ll be 500 years before that happens.” But now I’m like, “I don’t know. Things are moving really quickly.” And then I’ve also sort of started flip flopping on whether I think that’s a good thing or a bad thing.

Mark Richardson:
So you’re on team replicant, right?

Tim Oates:
Yeah. I was. I was firmly on team replicant. But I might be [inaudible 00:34:21]. So I guess I just like really sort of following the research and the news, and trying to sort out what’s real and what’s marketing hype. And just see where the field goes as we start marching towards more and more of the kinds of things that humans do, that machines don’t

Mark Richardson:
Awesome. I worry about the marketing hype as well. Every time Google tells me, “Hey, we improved our bid strategy algorithm.” I’m going, “Did you improve it? Or did you figure out a way to make more money off of your clients?”

Tim Oates:
That’s exactly right. Right.

Mark Richardson:
Stephen, what about you?

Stephen Sklarew:
Yeah, for me, there’s probably at least two dimensions to that answer. One is, I get jazzed about using this capability to help solve the world’s largest problems. So whether it’s climate change or healthcare, these things are massive. And if Synaptiq can help in a little way in that big pie, that really, really excites me. So that’s one piece.

Stephen Sklarew:
The other would be just being a place that amazing people want to join, and ultimately leaving the nest and doing amazing things on their own. So both Tim and I really enjoy working with other people, helping them grow their careers, and then letting them do their own things. And so seeing the legacy of what we do, kind of transcend our generation, and hopefully all those people will make a big impact on these world problems. That is very inspirational for me.

Mark Richardson:
That’s awesome. Well, I’ve really enjoyed speaking with both of you today. And, man, we’ve got a lot to chew on, and definitely want to recommend our listeners go check out Synaptiq. That’s S-Y-N-A-P-T-I-Q. And I believe it’s synaptiq.ai.

Stephen Sklarew:
That’s right.

Tim Oates:
Yep.

Mark Richardson:
And then, tell folks where they can find you, to follow you on the socials. Tim, why don’t you go first? You have a Twitter, right?

Tim Oates:
Oh, I don’t have a Twitter.

Mark Richardson:
You don’t have a Twitter. Stephen has a Twitter, sorry.

Tim Oates:
I am the least technically [inaudible 00:36:29] chief data scientist you’ll meet when it comes to social media. Talk to Stephen. He’ll-

Mark Richardson:
Off the grid. I like it.

Tim Oates:
Right.

Stephen Sklarew:
Yeah, so I do. I do have a Twitter handle. And I, candidly, use LinkedIn way more. But I can be found @SynaptiqAI. S-Y-N-A-P-T-I-Q-A-I. That’s the handle for the company, and I’m behind that. And so, yeah, that’s the handle that I use for Twitter.

Mark Richardson:
Awesome. Well, yeah, a lot of great stuff. I encourage folks to check out SynaptiqAI, and look into what they’re doing with Microsoft. It’s really, really cool stuff. And certainly look forward to seeing the other applications of your great work, and how you all are helping make the world a better place using technology. All right?

Stephen Sklarew:
Yeah. Thanks for having us, Mark. Appreciate it.

Tim Oates:
Yeah, thanks Mark.

Mark Richardson:
Hey, it’s my pleasure. With that, this has been another episode of The Data-Driven Marketer. I’ve been Mark.

Stephen Sklarew:
I’ve been Stephen.

Tim Oates:
And I’ve been Tim.

Mark Richardson:
Thanks so much. We’ll catch you next time on The Data-Driven Marketer.