Podcast: Play in new window | Download
Subscribe: Google Podcasts | Stitcher | RSS
In this final episode of season two, Jake focuses on change and the role of AI in our transforming world. The title “The End of the Beginning” is a reference to the increasing application of AI within our industry. Over the last several years, AI, particularly machine learning has quietly become a component of growing importance in marketing tools and applications. However, it’s really just beginning to hit its stride. The current changes in privacy, device identity and trackability provide a clear, high-profile example. Losing those core identifiers is a tectonic shift that will change nearly every aspect of ad targeting and ripple through every ad campaign, strategy and outcome. It’s a significant issue that is being addressed by machine learning – AI-powered solutions that are much more durable than device IDs and cookies.
To wrap up the season, Jake provides five key takeaways in the context of major changes we see in the marketing industry, and the role AI will play in each of them. Also, in this episode, Jake and Jeremy Lockhorn look to the automotive and healthcare industries to explore the trustworthiness of AI vs. humans.
Jake’s Five List:
- Facing Bias head on in a country and a world that is increasingly recognizing its biases and asking hard questions about how to adjust for them.
- Understanding the strengths of people versus AI as AI becomes deeper and deeper entwined into our everyday lives, both personally and professionally.
- Redefining physical retail in the age of eCommerce ubiquity.
- Embracing the changing nature of advertising, from planning and creative development to media buying, audience targeting and campaign measurement.
- Developing skill sets for the changing nature work in general and the current shifts within the organizations we lead or want to work within.
The Five podcast is presented by Ericsson Emodo and the Emodo Institute, and features original theme music by Dyaphonic and incidental music by The Small Town Symphonette. This episode was edited and mixed by Justin Newton at JaySong Studios and produced by Robert Haskitt, Liz Wynnemer, and Jake Moskowitz.
Transcript of AI E9: The End of the Beginning
Everything we’ve talked about this season will come into play because AI will play a huge role in a world in which privacy is paramount and targeting must be reimagined.
Let’s talk AI. Welcome to FIVE, the podcast that breaks down AI for marketers. This is episode nine, The End of the Beginning. I’m Jake Moskowitz.
When we began this season of FIVE, we were in the middle of the pandemic, the advertising industry, really the whole world was in an unfamiliar, uncomfortable place. But we rallied, we kept pushing forward. And some of the adjustments we had to make then kind of feel normal now.
It’s funny, throughout season one, we met at the recording studio and even conducted a few interviews in person. To bring you our second season, we scrambled to set up a makeshift recording studio here at my house, every interview was done remotely, which by the way, was much more challenging than before, because every one of our guests was suddenly having every conversation remotely.
The team that brings you the show, we haven’t seen each other in person throughout the entire season. But we rallied, we kept pushing forward, those adjustments we made them feel normal now. In fact, I don’t think any of us would want to do this any other way. Change is uncomfortable and inevitable. But it’s often essential. This show is all about that. We try to give you a glimpse of what’s coming, what it means, what will change, and how to make those big transformations work for you.
Season one focused on 5G, season two, it’s AI, two huge shifts that are becoming more and more prevalent and central to marketing a little bit more every day. And if you’ve been with us through both seasons, I hope you feel as confident as we do that you’ve been in on the ground floor for these industry changing trends, you knew before most and now you know more than most about the impact of these technologies on the work you’ll do going forward.
This is the last episode in our nine part series on AI and its impact on marketing. I’d like to wrap up this season of FIVE with a focus on change. In the previous episode, we touched on the industry shift around identity, how Apple was moving to a model that requires users to opt in to the tracking options on their phone and in their apps. How Firefox and Safari have taken a similar approach with cookies and Google Chrome is heading that way too. Losing those core identifiers, that’s not a small shift. It’s tectonic, it will change nearly every aspect of ad targeting, it will ripple through every ad campaign strategy and outcome. It will seem small and insignificant at first, then it will likely feel sudden and sizable. That’s a lot of change. And it’s already started, but we’ll rally. We’ll keep pushing forward. And everything we’ve talked about this season will come into play because AI will play a huge role in a world in which privacy is paramount and targeting must be reimagined.
So let’s talk about other key takeaways from this season. And let’s do it in the context of major changes we see in the marketing industry and the role AI will play in each of them. In many cases, AI is the change.
So number one facing bias head on. We’re living in a country in a world that is increasingly recognizing its biases and asking hard questions about how to adjust for them. AI plays right into the middle of that because AI runs the risk of codifying into a black box, not just intentional biases, but subconscious and even accidental ones. According to Pew Research, 58% of Americans feel that computer programs will always reflect some level of human bias. If we’re not careful when creating algorithms, we can systematize bias in a way that’s hard to detect and hard to undo. Rishad Tabaka Wallah gave a good example in episode two.
Every algorithm was written by a human being, every human being has built in biases. And so the best face recognition software in the United States still believes that most African American people look the same, which they don’t. And we know that that’s an issue, because in many ways, they cannot also work in Asia, because they believe all the Asians are the same. However, they work really well in a white community in the United States. Now, that is primarily because people write algorithms and they feed the data by the people and the data that they have around them. So telling the story to people why it sometimes doesn’t work or what they need to do and why they need to do is also important. These two things that I just told you, you can’t do without storytelling, which is how humans behave on nudge and how we can make mistakes because of biases. No machine will come in and say, machine, you have made a problem because this useless human carbon based person does not know how to compute.
We’re likely to see AI play a more and more central role in the societal change as the effort to stamp out cultural bias demands that algorithms become more transparent, and the makers of those algorithms take the steps necessary to prove what they’ve done to minimize bias.
Number two, AI is becoming deeper and deeper entwined into our everyday lives, both personally and professionally. And while that’s extremely empowering in many ways, it also carries with it a certain responsibility we have as humans in ensuring we don’t outsource all thought to the algorithm. To achieve that, it’s essential to understand the strengths of people versus AI. In our personal lives, we’re increasingly outsourcing our memories and our curiosities to audio assistants like Google Home and Alexa. E marketer says 128 million people in the US use voice assistants at least monthly.
On episode two, Jeremy Lockhorn and I shared some examples of how voice responses to our questions can’t be taken at face value. Here’s my favorite example, when we asked a question and got the answer, according to Scientology.
Hey Google, should immigrants learn English?
On the website governing.com, they say, immigrants come to our country for better lives. Those who learned to speak English are propelled toward the American dream. Those who don’t learn the language are destined to lurch on the periphery of society, subject to the whims of political pandering and government dependence.
Ouch! What website was that again? You and I like, we were specifically looking to poke holes in the answer and yet neither of us even picked up on what website Google was referencing. And it turns out, it’s a scientology website. Kind of surprising, huh?
You almost ignore the source and just tune into the answer.
Shelley Palmer put it well in episode one that in our professional lives, the smartest executives know that ultimately, it’s us the humans that are accountable, not to machines.
There are two kinds of executives in the world, like, you know, when you like to bifurcate the world, there are two kinds of executives, those who think AI is this magical black box that they will do their work for them. And then those who know that that’s just completely stupid. And unfortunately, you read an awful lot of science fiction and the papers and in the trades and online. People are ascribing all kinds of ridiculous attributes to artificial intelligence. And it’s getting blamed for a lot of stuff. Oh, you know, the computer. Okay, that’s sort of like just the laziest explanation ever. The problem with AI right now is that people literally do not know what it can and cannot do. And I think that’s the most important thing.
Number three, the ease and ubiquity of E-commerce is forcing physical retail to redefine itself, that was already happening long before COVID. According to Statista, the percentage of overall retail sales that occurs digitally has more than doubled since 2016. Though COVID sped up some changes to physical retail and created some new ones. Jason retail Geek Goldberg gave two great examples in episode four of how AI enabled retailers to adjust to a COVID environment.
If you think about like experiential retail is a common thing we talk about like this is the notion that people want to have a visceral experience inside of a retail store. Nobody’s really interested in dramatically increasing the dwell time in a store like we’re not, you know, putting coffee shops in stores to get you to hang out in the store and more we need you to get out of the store to make room for the next customer. I might only be allowed 25% of my former maximum capacity. I need to install a very reliable system to measure how many people I have in the store at any one time. And the best technology for monitoring store traffic happens to rely on artificial intelligence and computer vision. So you know, these are camera based systems. And at some point, I have to close the store to new visitors until someone leaves and so I need to interface those traffic systems with stoplight systems at my front door. So a lot of retailers are investing in that.
How I fulfill inventory is wildly different. I would now rather you do curbside pickup than come into my store because I have a limited number of slots in my store but all my inventory is still in my store. So I would love it if you would shop online, buy things from me and drive by and pick them up in my curbside so that you deplete my store inventory but you don’t use one of my precious slots to go in the store. And that creates a lot more supply chain complication. How many widgets should I have in the fulfillment center that I use to ship to your house versus how many widgets should I have in the store for people to take off the shelf versus how many widgets should I have available for curbside pickup. And so you know, again, artificial intelligence and machine learning is one of the best tools we use for inventory allocation and optimizing the efficiency of where I put all that inventory and predicting consumer demand.
Number four, AI is changing the nature of so many parts of what marketers and agencies do every day, from planning and creative development to media buying, audience targeting campaign measurement. A recent survey by Salesforce found that 84% of marketers are using AI up from 29% in 2018.
Josh .E. Hart gave a good summary in episode five of the mindset an agency needs to have to thrive with AI regardless of use case, it all comes down to being open minded and disconnecting ourselves from what we always thought we knew, to let the data lead us to answers we may not have thought of ourselves.
The agencies that will succeed using AI are the ones that, you know, in their core DNA, are willing to allow the solutions that they develop to be agnostic. There’s a lot of bias in the agency world around this is how we believe we can build and grow brands using this finite set of tools. And you can see it at all different points in the spectrum. You know, whether it’s a creative agency that believes that that rich video storytelling is the best vehicle to build that connection, versus a digital agency that is going to think of an opposite solution, a more digitally driven experience. So the agencies that can be most successful are the ones that are willing to follow that data and that AI to its logical conclusion and to what those solutions are.
Now, of course, as we’ve learned, we can’t just turn off our brains. It’s a delicate balance. And we have to put a human perspective on what comes out of AI and ensure what makes it out into the real world is an amalgam of the best of human and machine. But we can’t start the process thinking we know what the answer should be.
Number five, because of AI, the nature of work is changing around us. A 2020 report from the World Economic Forum said 94% of business leaders expect employees to pick up new skills on the job. And that has significant implications for the skill sets we need to foster in ourselves, and also the characteristics for the organizations we lead or want to work within. Ben Harrell, the CMO of priceline.com, put it this way in episode six.
But if you stop learning, other people will pass you by, your career pool will stagnate. Because no one’s ever learned enough than coast for the rest of their career. I do think that analytics data, whether that’s in the form of AI, or other forms of analytics is going to become increasingly important. So if you are that AI marketer, that can be really a data scientist that can build up these systems, that can monitor them, that can help return value, great, that’s fantastic. There’s going to be great opportunities for you. But if you’re not the data scientist, the AI whiz, that’s fine too. I do think you’ll have to understand what AI is, when it can be used and it’s helpful to know who you can go to to leverage that skill set. Ideally, there’s people on your team or close to you that you can say, hey, I was thinking about this. What do you think? So being able, again, to think critically to step back and say, am I okay with this? Are we in the right direction? Where are things going? Don’t be afraid to question things.
And again, as much as that’s important for senior leaders at companies, it’s expected of senior leaders of companies. It’s even more important for junior people and companies, we have to be able to take a step back and question it. Some companies don’t like that. Priceline is a company where we encourage people that are junior, really everybody, to question things, to raise issues.
When you look out into the future a year, five years, sometimes you can see the shifts coming out on the horizon. And sometimes change just feels like an unexpected punch. I think we’ve all felt a little or a lot of both over the last year. Sometimes technology is the change. Sometimes it’s the remedy. AI certainly plays both roles, and it will be both for many years to come. As it does, I hope this season of FIVE has armed you with some insights and perspectives that will help you rally and keep pushing forward and even thrive.
Before we go, I asked my colleague Jeremy Lockhorn to join me again today to talk about the subject of trust. Can we trust AI? Is AI worthy of being trusted? One way to start this is to question the question. There’s a lot written about whether AI is trustworthy. And the first step is what do we mean by trust, like how do we define trustworthiness?
I think the way that I would define trust is, do I have faith in a machine the same way that I have faith in a human?
That sounds correct to me, which makes me think that the question is not the right question, because I agree with your definition of the question. So to me, the real question is, which is more trustworthy, humans or AI? Because I don’t think you can assume that humans are trustworthy. I think today, we have that assumption. And that’s a bias. And that’s exactly how AI bias occurs is that the question you asked at the very beginning, is bias to begin with, because it assumes an answer.
And I’ll give one example. And that’s in driving. I recently got a Tesla, I’m very excited about it. And I get to play around with autopilot. And I love autopilot. But let me tell you, my wife hates autopilot. And she inherently just doesn’t trust that the computer is going to turn when the freeway turns or stop when the car in front of us is going slow. And I think that all assumes that the computer is going to do a worse job than humans at those steps. So I decided to look into it. It turns out that Tesla does a quarterly vehicle inspection report, for each quarter, they report the number of miles driven per accident while users are on autopilot, versus the number of miles driven per accident when users are not on autopilot. And they’ve done it for 10 quarters.
So I took the most recent five quarters and compared it to the five quarters previous to that. And the human driving, the non autopilot driving is basically exactly the same, that’s within a half a percentage point between the first five quarters and the next five quarters. But the computer performance of the most recent five quarters is 30% better than the previous five quarters. And in the most recent five quarters, the computer performance is almost exactly twice as good as human performance. So about twice as many miles per accident when the car is on autopilot as compared to when the car is not on autopilot.
Yeah, it’s fascinating. So the machine is getting better while humans are maintaining their level of accuracy.
Exactly. And the computer was already better to begin with. And it’s only getting better.
Yeah, fascinating. You know, I found something that was kind of similar, sticking with the autonomous vehicles thread here. What’s really interesting to me is that self-driving cars require like multiple levels of trust, right? Like everybody that’s in the car has to trust the system, people in other vehicles nearby have to trust the system. And then pedestrians have to trust that system as well. I don’t know about you but before I step off the curb in a dense urban environment, I usually try to make eye contact with a driver to make sure that he sees me he or she sees me and sort of acknowledges me, that’s not possible when there’s not a driver in the vehicle.
And so I found this really interesting study that Ford did. Do we need to replicate that sort of acknowledgement that happens between a human driver and a pedestrian? And they found that the answer to that question basically is yes. And the way that they did that is they created this thing that they call the seat suit, that they put a human driver in that basically, it looks like the driver’s seat is empty. But in fact, there is a human who is driving the vehicle. And they drove through dense urban areas for like 180 hours over 2300 miles, they recorded all the different interactions with people.
And the conclusion ultimately, that they came to is that the industry needs to develop a system of communication between a self-driving vehicle and pedestrians that will make the pedestrians feel safer. It’s kind of like the same way that we have brake lights and turn signals today to signal to pedestrians and other drivers on the road. It’s just an evolved version of that between a self-driving vehicle and a pedestrian.
So it sounds like we’re answering two different versions of the question. I was answering the question, can AI be trusted as compared to humans? And you’re asking the question, regardless of performance and whether the AI should be trusted, is it trusted? Which I actually prefer your version of the question.
Yeah, for sure. So you know, another really interesting example of trust in AI is in the medical field. And there have been a bunch of studies published, suggesting that AI can be as good or better than human doctors at diagnosing a wide variety of conditions. To be fair, there is some ongoing debate around the experimental design and thoroughness of peer review or lack thereof in some cases of those studies. So I don’t know that I want to get into being an advocate for AI powered medicine yet.
But I did find this really interesting article in Harvard Business Review, where the authors had done a series of studies to gauge patient willingness to accept an AI provider over a human doctor. And the short story here is that there’s a really strong reluctance across a variety of procedures ranging from simple things like diagnosing stress level to more advanced things like skin cancer screening and pacemaker implants surgery. Even when they’re told that the AI provider is more accurate and has a lower error rate, patients were less likely to choose that provider to utilize that AI service and they wanted to pay less for it. They believe that it’s inferior in some way.
The author’s tried to suss out their reasoning behind it. And it doesn’t appear to be a belief that AI provides inferior care, but rather, that the resistance seems to be coming from a belief that a specific patient’s circumstances are unique, right? And it’s like AI can’t possibly understand the different combinations, you know, that are unique to me, the different circumstances that are unique to me, and therefore, I’m less likely to trust it.
What was really interesting is I found a study where a group of researchers looked at 14 different academic studies that each compared human performance to AI performance on real world data that was outside the training dataset on medical diagnosis. And I do think you have a point there, because the takeaway from this study was that the computers barely outperformed the humans at diagnosis on an apples to apples basis. So when the human doctor got the same information as the computer, the diagnosis from the computer was slightly better. However, the human in that case, did not get the other information that you’re referring to there.
So what’s unique to that patient? What other information does an actual human doctor take into consideration when making a diagnosis? So perhaps one takeaway for me is the ultimate way to build trust in AI, is to combine the best of the human with the best of the computer. And maybe the question of which is more trustworthy? A human or a computer is actually the wrong question. Because really, the answer is both. What do you think?
Yeah, I think that’s a fantastic point. And the question really is, how do you take what each contributor AI versus human is best at and maximize it against whatever solution you’re trying to develop? I think at least in the near term, the most successful cases are going to be where we find ways to collaborate with AI and leverage sort of the best of both worlds. So I think you make a great point there.
And, you know, I would also say, it’s funny, like, we’re talking about, you know, all the examples we pulled from are basically life and death examples, right? Whether it’s a self driving vehicle, or medical diagnosis, like these are serious considerations. And we’re fairly fortunate to be talking about this from a marketing perspective, you know, we live in the marketing world, and these are not life and death decisions that we’re dealing with here. And brands and agencies have, in fact, you know, been embracing algorithm fueled marketing for a long time now. And I think we as an industry are getting way more sophisticated at understanding the importance of things like truth sets, and data provenance and data quality, and all of that stuff that factors into, you know, how successful and ultimately trustworthy an AI solution can be. And so I think we’ve got it way easier than these other examples that we’ve looked at.
Also, I was thinking that it’s possible that we have to adjust computers to have a human face on them to tap into human instincts about how we react to things in order to build that trust.
Yeah. Is it about putting a human face on it? Or is it about developing ways for them to communicate in standardized ways that people can understand?
Jeremy, I really appreciate you joining us not just for this episode, but the whole season. It’s been a real pleasure.
Yeah, thanks so much for having me. It’s been a lot of fun to participate. And it’s been a great season. And looking forward to the next one.
I’d like to thank all of our great guests, the CEOs, authors, strategists, technologists, and analysts who joined the show this season. And I’d like to thank you, sincerely. Thank you for listening. Oh, and one other thing we’re grateful for, you know, a lot goes into the show, so it’s especially rewarding to have the work recognized in the industry.
In 2019, our first year, FIVE won the Mark Com Gold Award for outstanding podcast. It won the AdExchanger Award for Best Industry Commentary and Analysis in 2020. And just a couple weeks ago, the FIVE podcast was awarded the Hermes Creative Award for Best Podcast Series. We’re so honored.
If you like the show, please write us a comment or give us a rating on your favorite podcast listening platform. We’d be super grateful. It definitely helps more people discover the show. And speaking of change, we have some of our own changes in mind for season three, new focus, new insights, lots of new adjustments. We’ll bring you a wider range of expert interviews and explore innovation across a range of topics. We can’t wait to kick it off in just a couple of months.
But for now, that’s season two, thanks for joining us.
The FIVE podcast was presented by Ericsson Emodo and the Emodo Institute and features original theme music by Dyaphonic and incidental themes by the Small Town Symphonette. Original episode art is by Chris Kosek. This episode was edited and mixed by Justin Newton and produced by Robert Haskitt, Liz Wynnemer and me. I’m Jake Moskowitz.
Okay, one last thing before I turn off the mic, I’d like to specifically thank my guests who were incredibly generous with their time and thought leadership, Rishad Tabaccowala, Shelley Palmer, Michael Stick, Ella Chindits, Rave Allez, Charlie Archibald, Kyra Sundance, Jason Goldberg, Guru Harry Haran, Grant McDougal, Josh E. Hart, Ben Harrell, Perry Malm, Liz Miller, Don Fluckinger and Todd Touesnard.
And a special shout out to a few people who don’t get enough recognition for their great work on the show. Justin Newton at Jay Song Studios, Lyon Solntsev at Ericsson, Liz Wynnemer, my colleague who organizes this whole thing, Robert Haskitt who does all the writing. And of course my friend and colleague Jeremy Lockhorn. See you next season. I’ve got to go get some water.