This past July, we drew 300 attendees from 12 countries for the inaugural Marketing AI Conference (MAICON) in Cleveland, Ohio.
If you missed it: the Institute's goal in launching MAICON was to help marketing leaders understand, pilot and scale AI in their organizations. More than 50 speakers joined to help us do just that, including presenters from Facebook, Grant Thornton, HubSpot, IBM, MIT Technology Review, Publicis Sapient, SoftBank Robotics, The Natori Company and Yext.
While onsite at MAICON, Paul Roetzer (@paulroetzer), founder of the Institute and creator of MAICON, sat down with eight leading AI experts to ask a few burning questions. You can tune into these conversations by downloading the four-episode podcast series on the MAICON Speaker Series or you can simply start with episode one below.
For ongoing marketing AI know-how, subscribe to our Marketing AI Institute newsletter, and join us for MAICON 2020, July 14 - 16 in Cleveland.
In this episode, Paul interviews Karen Hao (@KarenHao), senior AI reporter for MIT Technology Review, and Cal Al-Dhubaib (@caldhubaib), chief data scientist at Pandata, regarding the boundaries of AI and how marketers can get started now.
Below, get to know more about our featured guests—plus a full transcription following the recording. Please note that the transcription was compiled using AI with Otter.ai, so blame any typos on the machine :)
Karen also writes a semi-weekly newsletter called The Algorithm and cover the sticky ethics and social impacts of the technology. Prior to that, she spanned the editorial and product teams at Quartz as a tech reporter and its first data scientist, covering the future of cities, experimenting with chatbots, and building machine learning models. Say hello to her at karendhao.com.
Cal has received both national and international recognition for his work in predictive modeling and entrepreneurship. As the first data science graduate from Case Western Reserve University, he works tirelessly to advocate for careers and educational pathways in data science and contributes to workforce development initiatives throughout Northeast Ohio.
Session Transcription
Hi, I’m Paul Roetzer, founder of Marketing AI Institute, and creator of the Marketing AI Conference (MAICON). MAICON is designed to help marketing leaders understand, pilot and scale AI in their organizations.
The inaugural event was held in Cleveland, Ohio, July 16 - 18, 2019 and drew 300 attendees from 12 countries.
There were more than 50 speakers, including presenters from Facebook, Grant Thornton, HubSpot, IBM, MIT Technology Review, Publicis Sapient, SoftBank Robotics, The Natori Company and Yext.
This (four episode) podcast series features insights from (eight of) our speakers who we interviewed onsite. We cover an array of topics including: what AI is, how to get started with AI in your organization, AI applications for voice, how to humanize your brand, and how AI will transform marketing moving forward.
AI is forecasted to have trillions of dollars in annual impact on businesses, and, yet, most marketers are still struggling to understand what it is and how to apply it to their businesses and careers.
You have a choice. You can sit back and wait for the marketing world to get smarter and change around you, or you can embrace AI now and be proactive in creating a competitive advantage for yourself and your company.
If you choose to take action, I hope you’ll subscribe to our Marketing AI Institute newsletter, and join us for MAICON 2020, July 14 - 16 in Cleveland.
Now, onto the podcast. Let’s hear what our speakers have to say...
Our first guest is Karen Hao, the artificial intelligence reporter for MIT Technology Review. She covers the advancements in, ethics of, and social impact of the technology. To begin our conversation, Karen told us about her role:
Karen: My day to day is talking with the leading researchers that are developing technology, and then sometimes I'll talk with executives or other people, other data scientists at companies that are then applying the research to products commercial products. But I always feel like it's too broad because there's just so much going on in this space and I get overwhelmed because I want to cover everything. There's so much to learn and it moves so incredibly fast. So it's constantly like running a very long marathon.
Essentially. When I cover it, I have a thesis a guiding thesis and that is that I really think it's important for technology people and humanities people/social scientists to be talking to each other because a lot of the work that comes out of A.I. is going to affect society. It's going to affect humanity and so there needs to be a dialogue between all of those different disciplines. And so when I look for stories I'm looking for things that are really going to start affecting people soon.
Paul: Our second guest is Cal Al-Dubaib (Al-Due-Babe), a data science thought leader in Northeast Ohio. He’s led teams who help organizations jump start their data science initiatives and begin using advanced technologies, like artificial intelligence, to impact their bottom line. Cal told us about his role with Pandata.
Cal: So it has really morphed a lot over the last three and a half years. And this is because the discipline of data science is evolving. So today when I think about my role as chief data scientist I'm doing four things. I spend a lot of time communicating a lot of time thinking about solution design how to bring different pieces together. I do quite a bit of thought leadership and research. What's the latest. The speed at which data science is evolving is so fast that you have to stay on top of it all the time. And then finally best practices and compliance. So when you talk about all the regulations happening today you really stay ahead of that. So tying those four pieces together and understanding that the business pain points that our customers experience translating those into solutions and making sure that they're compliant.
Paul: It’s funny, we’re at this conference exploring Marketing AI, but it begs the question, what exactly is AI? To begin our conversation, we asked Cal who loves this topic.
Cal: I love this question because I don't know, what is artificial intelligence? That's a great question! In fact we do lunch meetings every month, and we're a group of data scientists, and even we couldn't agree on a consistent definition of artificial intelligence. And I think it's a moving target. One of my favorite definitions I heard today you know read this quote here was "the science of making machines smarter." And so to me what that means is you're taking a repetitive tasks that might have some element of human decision making and you are automating that. And so that's one way of making machines smarter. Or.
Paul: Like Cal says, it seems the definition of Artificial Intelligence is a moving target, and the term often gets mangled to include any kind of algorithm or computer program. How can you differentiate whether something is using AI or not? In conversation with Karen, we explored exactly why Artificial Intelligence as a term is so hard to pin down, and she gave us a few examples of what AI looks like in our daily lives.
Karen: AI actually refers to two things that's why it's confusing. The first thing that it refers to is the field and the goal that the field has which is to build. These like really sophisticated super smart super helpful embodied human like AI systems kind of like what you see in sci-fi. Not really but like sure we'll take that metaphor and that's like one thing and then the other thing that it refers to is like the actual state of the technology now. And so that's why it can be confusing because sometimes people will conflate the two and think that what we have now is much more advanced but it really is not. It's that they're confusing it with the aspiration for the field. So what we have now is. Most of what you hear about is a category of A.I. called machine learning and machine learning is a process of using statistics to find patterns in data and then taking those patterns to make decisions.
And so. The systems that you see like face recognition that which tags your photos on Facebook or your Netflix recommendation algorithm all those things are machine learning that is taking what like your in your images of your face and then finding like the pixel patterns that make up your face and then using that to identify your face and other photos or like that that shows that you're watching and the things that you like as the data taking the patterns of oh it seems like she really likes sci fi and recommending more things like that. That's that's essentially A.I. in a nutshell right.
Paul: So AI is here today. We use it in our daily lives and marketers are looking for new ways to use it in their work. However, many marketers are having a tough time getting started with AI, so we asked Cal to explain why so many of their projects are doomed to fail.
Cal: To me artificial intelligence, machine learning, big data… it all comes out of the discipline of data science. To me data science is the process from which you're creating value from data. The term science and it is really important. It's experimental in nature. And this is a little bit new to the business world. When you say Alright we're gonna have to experiment a little bit to really prove out the value and this is not going to work sometimes and that's OK. That part is lost in translation, particularly when you're using words like A.I. and machine learning, uh you forget about the science aspect of it.
And so organizations invest a lot of resources in trying to get these initiatives off the ground. And when they fail it shuts them down. Like oh that doesn't work. And there's no value in that. So you have to have two processes that work really really well when it comes to succeeding with data science. One is experimentation and then the other is operationalization. So finding out what works when scaling it and getting into the hands of users.
Paul: So we see that companies need to be more open to experimentation with AI, but Karen also raised some of the ethical obstacles marketers face. AI isn’t just about data and science, we need to remember the end user – the real people this technology is going to affect.
Karen: Yeah. Well I mean the thing that machine learning is really good at is prediction because it's finding the patterns applying the patterns. And so. When you want to do you like it like Amazon for example uses it to recommend products to users and that's sort of like a way to increase the conversion rates for consumers or for casual browsers into buyers so that that's kind of like when you think about the marketing tasks that you're doing. Think about which tasks ultimately boil down to prediction.
And those can very likely benefit from machine learning. I will say that like the flip side of that is you have to make sure that you're also being ethical about it because. You don't necessarily want to personalize things to no end. Because like we said there's there's like positive and negative consequences to that. And if you have hyper targeted ads that are following people around all the time. They could be helpful to people that have lots of money to spend. That could be a little bit predatory to people that don't.
And also, there was that lawsuit that came out like a few months ago where the department of housing the U.S. Department of Housing sued Facebook because they were. The way that Facebook has built its advertising and marketing tools. uses machine learning to a point of discrimination. So I think the most widely reported. So that HUD sued Facebook on two accounts. And the most widely reported one is that. Facebook was allowing advertisers to explicitly narrow down the audience target audience of their ads. And obviously that causes like housing discrimination because then you can say look I only want white people to see the ads or whatever. But the second thing the second account that Facebook was sued for which was less reported on is even if you get rid of the explicit filtering the machine learning is still doing implicit filtering.
So there was research that shows that. Yeah. If you don't limit your ads at all and you are selling housing. Houses for sale will be shown to more white users than minority users, houses for rent will be shown more or minority users than white users, and it also applies to job ads. So job ads for nurse and secretary positions will end up being shown to more women than men. Job ads to janitors and taxi driver positions are shown to more minority users than white. And so. When you're thinking about like when when you think about implementing machine learning in your marketing tasks like just be very very cognizant of thinking about OK. But are there potential negative consequences. How can I mitigate it? And I think that those two things will make you a very skillful marketer.
Paul: We can see that there are many obstacles facing marketers today both practical and ethical. This leads us to the question: in this early phase of experimentation, how can companies set up their marketing AI projects for success? Here’s what Cal had to say.
Cal: So there's these four questions we ask anybody especially at the start of a project. And we have this convenient acronym RISE that we use. Then I'll define it a little bit but: Results, Impact, Measuring success and Ethical considerations. So unpacking that a little bit starting with results you want to think about all right. Who is. When we talk about this notion of automating decisions right. Who are the decision makers and what are the types of decisions that they would be making with it. 2 impact right. If you can act on that information is there measurable potential for value.
Measuring success a little bit related but how do you know when you've arrived. How do you set those benchmarks to say you know what we're getting this a solution to improve conversions or click rates on our marketing campaigns right. What's that boost that you're expecting to see by a certain date and if you don't get there you say well you've gotta try something else. And then finally ethical considerations you know are you making recommendations where gender or demographic bias might be a consideration. Do you have regulations like GDPR are that impact what you can and cannot do with customer data? So asking those four questions is really really really important. And then as far as aligning the organization goes there's three things you want to do buy in, trust and consistent performance.
Paul: So we can use Cal’s RISE acronym, focusing our projects on: Results, Impact, Measuring Success, and Ethical Consideration in order to set our projects up for success.
We also asked our guests for their advice to beginners. What advice can you offer to marketers who are just getting started exploring AI?
Karen: Read my newsletter. It's called the algorithm. I write it for Tech Review and you can go to Bit.Ly/thealgo Or you can Google it. And it is – I like to think – the go-to source for demystifying artificial intelligence. It is. It is more research focused. But I actually encourage people who are using commercialized products to dive more into research. I think it it really helps us people build a better foundational understanding of what they're using and why it works so that when issues like bias and other things come up. There is you kind of how you can draw from your foundational knowledge to understand how to fix it or how you might approach things differently or or look out for other issues that haven't yet been reported on.
Paul: This is so key. Markets need to read up on the latest news in AI and research where it could help in their efforts. We posed the same question to Cal, asking advice for marketers who are just getting started exploring AI.
Cal: I can't stress enough the importance of starting small. And proving out success. Oftentimes you hear from the top down all you've got to do a guy everybody's doing a lie and that's the wrong reason to be doing artificial intelligence. You really want to be using it to solve a practical problem right. And sometimes the problems that you can start to solve with relatively small resources they're not that exciting. But you need to prove that it works. Right. Going back to this idea of buying trust and consistent performance. If you can align around a shared win within an organization.
And you can demonstrate value you build that trust. But if you keep putting a lot of resources in and there's no measurable impact or you're solving the wrong problem you've destroyed that trust and all of sudden it's really hard to maintain buy in.
Paul: Any last thoughts?
Cal: It's an exciting time. And oftentimes when people hear artificial intelligence and machine learning these are buzzwords and you just want to shut down like there's no way this applies to me. Be OK with the fact that everybody's learning together. Experiment figure out what works for you and really double down on asking the question of where can it solve that one small problem where you can demonstrate value.
Paul: It is an exciting time, and we hope you’ll take some time to explore how AI fits into your marketing strategy.
I’m Paul Roetzer, creator MAICON. Thanks for listening to the Marketing Artificial Intelligence Podcast. If you enjoyed today’s discussions with Cal Al-Dubaib and Karen Hao from this year’s MAICON event, I’d encourage you to check out our 2020 event at MAICON.ai. This annual conference is held in Cleveland, and brings together the leading experts of the marketing and artificial intelligence communities. We hope to see you there.
This podcast is a production of Evergreen Podcasts. A special thank you to:
Producer: Brigid Coyne
Audio Engineers: Dave Douglas and Eric Koltnow
Thanks for listening – we’ll see you next time!