Linda Hesselberg: Hello and welcome to Lorn. My name is Linda Hesselberg. And today we are recording from the Nordic Business Forum. My guest is digital marketing evangelist for Google and Co-founder for Market Motive, Avinash Kaushik. Welcome.

Avinash Kaushik: Thank you.

Linda: So first I want to know who Avinash is and how did you get into Digital marketing?

Avinash: It’s hard I guess complicated. So, I spent some of my time at Google and some of my time just writing. I have written two best-selling books on analytics and I think that is probably my primary foray into digital because if you want to do great analytics, there is no other place than digital to have access to just very interesting datasets and very interesting amounts of data. Over the last few years I have focused on both figuring out how to get data to go farther but also how do you fundamentally rethink the kind of decision making that happens in the company and why certain things happen faster? Why do they happen slower? How can data influence strategy and how can it influence people, how can it create a sense of urgency to take action. So that’s primarily the kind of things I focus on, but I write a blog called “Occam’s razor” and I write a little newsletter and help build tools that allow people to be smarter.   

Linda: Early on stage today you talked a little about AI and I would love it if you tell our listeners how you define AI.

Avinash: Yeah, of course, one of the things I shared is that I don’t tend to use the phrase AI too much just because it is too broad and humans don’t typically understand the entirety of what it means, at some point maybe we will understand it enough. AI just stands for intelligent machine and I think both of those things are not understood enough at the moment. I think we use the phrase machine learning a lot. The definition of machine learning is just algorithms that are intelligent without being explicitly programmed.

Most software today is explicitly programmed, with standing we have to write all the rules and structures and things, and these machine learning algorithms are primarily powered by deep learning. They learned rather than we have them to teach. So I think this is a very fundamental shift to creating things that are intelligent. And not intelligent because we wrote particular sets of instructions down for them to follow but, intelligent because they learned from data. So I tend to use machine learning a lot and the technique that is most popular currently is deep learning. Almost all of the concurrent magic that we see done by machine learning is actually a technique that uses neural networks.

Linda: You also talked about another term AGI, firstly what it stands for and what it says because you said that was the thing you saw in the future.

Avinash: Yes, most of the machine learning algorithms that are being applied at the moment are what is known as Narrow AI. What that means is that I gave an example of algorithm that is trying to automate detection of a struck job that a neuro cardiologist normally does. So if you look at the algorithm its a narrow AI in the sense that it does that one job and it does that job better. Another example I shared is about an algorithm that will do smarter, profitable marketing for you but that’s all it does. Over time we will not have Silone AI, Narrow AI but we will have algorithms that will be able to do more than one job. So AGI stands for “Artificial General intelligence ”. If you think of the most powerful manifestation of AGI its us humans, because we can do multiple jobs at a time. We know how to eat, we know how to solve a math equation and we know how to tie your shoes, that’s the general intelligence .So lots of companies and lots of people are working on creating intelligence that won’t just do one job but it will do multiple jobs and there are many signs of it at the moment. When we approach the creation of the AGI of the next few years it opens up a whole entire avenue of automation, fragility of decision making, jobs that it will be able to do. And the way I think AGI will vastly accelerate the pace of change and you can see over time, if human beings still get to a workable AGI from there for it to become super intelligent  is a very short amount of time, because it will be so smart, it will observe so much. I am very excited about it. It holds a lot of potential to accelerate the revolution we are living in at the moment.

Linda: You talked about earlier that we need to work smarter and not harder and I think this is increasingly harder because where we get more and more data we get more and more possibilities and we have endless data which means we have endless possibilities. So, how can we in this digital age with a tsunami of information, actually use the information to work smarter and not be overwhelmed by it.

Avinash: I think that primarily through many of the examples I shared in my keynote today is one of the coolest things that we are doing at the moment with even narrow AI. All the machine learning algorithms are being used right now because of the expectation that a human will be able to understand 5,6,7 different variables. Does television advertising work? Now in order to answer the question I was like, okay. I don't know where we showed the ad? How long was the ad? How many people did we reach? Maybe two-three more things, but those things are not enough to understand if television advertising is to be effective. Did you have other kinds of advertising? Is it a current customer? Is it a past customer?

Linda: What happened in the moment as well.

Avinash: What are your competitors doing? What is happening in the industry? There are hundreds of variables to answer that really-simple question, does television advertising work? And what we can do now is rather than have a human be able to absorb all that complexity and use our brain to process that. You can use machine learning algorithms. Some of the coolest things about these algorithms is that some of these tasks that machines are really good at we can simply hand over to these algorithms and there are very few limitations. If you have a narrow problem to solve there are a few limitations in the complexity it can absorb. Most of these algorithms, perhaps all of them to the best of my knowledge can detect things no human can. Because they have perfect memory, they can process enormous amounts of data, they can build relationships between data structures that we won't be able to because we can't hold that much data.

Linda: And they are objective as well.

Avinash: Exactly. They're free of bias. So what I'm excited about is, as there is growth in data, so there is growth in the intelligence  that can observe the data. We don't have to do it. The role that we can play in two scenarios one is, for now, we still need human beings to take action in many cases. So once you get the collection recommendation, focus on this area. This is how television advertising works. You still have to go back and change your strategy. So we still rely on humans for now. We are working to automate that as well. And then the other place where I think we can focus on, and this is really exciting for both of us, is that we don't have to deal with all the complexity of absorbing all this data. For now, where humans are really good at, is playing a very interesting role in applying intelligence  where there are holes in data. In this way there is incomplete data or we don't really know the gap between good data and bad data. If those situations don't exist at the moment the algorithms that are currently in place in the world they're not very good in those scenarios. So okay, then we can play a role in that case for now because we can apply judgment, we can use experience, we can use these other things that humans are still good at for now. So I think our role will shift to a more value-added role and we will be doing the tasks that these algorithms are quite poor at for now.

Linda: You are dipping into my next question. What would be our role when the computers don't take over but, how should we coexist? But you talked earlier about winning the ML or machine Learning Revolution today by implementing its free key elements learn, build, and profit. Can you explain? I thought it was a great speech earlier. 

Avinash: Thank you. While my interests lie in using these algorithms to figure out how to do marketing better and how to do analytics better. I do think that the biggest promise that these algorithms hold are actually for businesses and applying them to a wide array of business functions. And so, to learn is all about making sure that you have your own hashtag AI Army, which is you have talent inside your company that is able to take advantage of these solutions. So do they have the right skill sets, do you have enough people, do they have access to the right data sets and all these things. Is that you have to build an organization that has an ability to learn and be able to take advantage of these opportunities that I see in front of us. So in the past this would have required you to invest tens of millions in machines and in software a lot of this is now available in open source, multiple open source solutions, they can take advantage of. So for a learning organization the ability to start quickly entirely rests on these skill set that you have in your team. So it is a very important sort of call to arms to our audience today here in no way to change the way that they are hiring. Build is all about figuring out how to do these things faster and rather than building or creating your own servers and hard drives and things to take advantage of cloud ML. Most of the providers out there Microsoft, Alibaba, Amazon, Google all of these people have incredible capacities inside their own cloud offerings where you can leverage machine intelligence  already. So I always say you can go from zero to 60 in 2 seconds, whereas a BMW might take 5 seconds. So you can actually start with you really fast and start for cheap. So that's all around figuring out how you build solutions with data sets that you have inside your company: smarter CRM, ERP solutions, HR systems, logistics customer support, all these core business functions. We can build solutions using cloud offerings that are substantially more intelligent  and profit is all about figuring out how to apply it in the area that I tend to live in which is marketing and analytics. Just smarter analysis and analytical systems and platforms, smarter ability to do creative targeting, bit management. All these other things that are required to do great marketing. So those are the three big clusters and together, they form a sort of an opportunity for a company to fundamentally reinvent every part of the business and not just marketing, in fact I personally believe that these algorithms can drive higher profitability outside areas from marketing analytics as much as I personally love it and my career benefits from that. I actually think it's the core systems and platforms in a company and the kind of innovation you can drive there probably has bigger profit potential than even marketing.

Linda: I think what you said in the beginning of learning to stay curious. I think that is so important and I think when you get a curious workforce, that's the start of everything, that's when you are able to build, that's when you're able to innovate, that's when you're able to use those open source solution that is out there. Because there's a lot of open source and there's a lot of podcasts such as ourselves that we can begin with. That's very easy to just begin if you're curious and if you have a work culture that accepts being curious.

Avinash:  I agree with you. In fact, you know, one of the things I had said is you know what machines are good at and what are humans good at. machines are going to be really good at frequent high-volume tasks, which is roughly 90% of what any company does today. It's all going to be taken over by machines. But that is a very important key role humans can play in tackling novel situations, but the mandatory requirement for humans to be able to tackle novel situations is that they have this incredible capacity to learn. I often say that I personally spend about 3 to 4 hours every single week learning something new not just about marketing or analytics but about multiple areas of things there are to learn out there. But if you have that commitment, you're going to do great no matter what happens to the machine because you'll be so much better at being able to tackle novel situations while the machine is doing frequent high-volume tasks.

Linda: Maybe it's also not more important but may be equally important to learn when you're in marketing, then you should learn other parts of the business to implement that in your marketing then you get the whole picture.

Avinash: Oh, a hundred percent. I've said that we live in a world where we used to reward people who have a particular core competency and they're really good at it. And so there's an accountant or an analyst or a dishwasher or a house painter. You would be really good at it. In a business capacity I would say there is a 70/30 rule now. Which is 70% for now you have to be really good at your core functions. If you are an accountant, you should be really good at accounting and you should be really good at figuring out how to apply algorithms to accounting. But 30% of your competency has to come from an adjacent area. So let's say you do accounting at SAS. You should know a lot about the airline business, you should know a lot about how planes take off and land. Is there a lot about how scheduling is done or if you're an accountant at Google you should know a lot about software is you know a lot about the search business. Because if you don't understand business strategy and adjacent business functions you as an accountant you don't have to be great at writing code that’s not what I am saying, there are engineers for that. But you should understand adjacent business functions so you are an accountant supporting the engineering team. You have to understand enough about their world that would cause you to fundamentally rethink your world. So I think we're going from people we would reward for being 100% good to now being 70/30. 70% in your core area 30% skill set in the adjacent area that surrounds you because without it the solutions you come up with, the impact you can have a company will not even be close to as good otherwise.

Linda: That's when you get the synergies. 

Avinash: That's right. 

Linda: What are your focuses in digital marketing as of now? One of the key elements you focus on.

Avinash: Two areas primarily, I think as you saw in my keynote. One of the areas that I'm really focused on in digital marketing is figuring out how to insert intelligence  in every single step. What are the kind of things that we can do in order to make things smarter. This is using, building new algorithms to analyze data or to use algorithms to understand performance, use algorithms to understand human behavior. So one is applying intelligence  at a scale that we don't understand. So if you visit our website, what can we understand from your behavior intelligently so we can create high relevance for you. So this entire space around investing in inserting intelligence into everything. Whenever there is a human involved, figure out how to have the frequent high-volume task takes things to be intelligent. So that's one big part, one of the biggest lessons I've learned which just drives the second part of my focus on how to automate taking action? Even if I supply really smart intelligent data, there is still a barrier in a company between politics and bureaucracy and layers and organizations, which actually impedes automation. If I intelligently understand how I paid search program is working is that intelligence? Why do I need to send you something so that you think it through when you have time and not having coffee and then the emails and all the other things you do in a company. So I am obsessive about closing that last mile gap where we can say intelligence  helps understand data, which in turn helps identify what to do and we have automation that takes that action automatically. That's the last part is where I think companies are less smart than they are, even when they have access to intelligence. So figuring how to automate, so here's a really simple example, you can use very sophisticated attribution models to understand how all your marketing campaigns are doing now they would still require to go change your Budgets and a little long.

But now actually the output of your machine learning attribution model is directly connected to AdWords, is directly connected to display platform. So once it understands exactly how valuable that platform is, it will automatically change your budgets in bids and target your strategy. 

So all you have to do as a human is say, here are my products and here is the metric that is valuable to me, profit, and I would like a minimum of a 100 euro profit whatever may be the numbers you set the rules of the game, but after that you get out of the way. Let the algorithms figure out how to capture data, understand, analyze it work within the rules you identify and you simply wait at the other end with your legs on your table waiting for money to fall from the sky. So that last part is where I think there is a bigger opportunity. And the reason I'm obsessing about it is that at the moment it's entirely a human problem that companies are not more profitable than they should be. So we have to figure out how to solve that, figure out how to apply intelligence  in the first case, figure how to automate decision making. 

Linda: And that's also when you eliminate kinda the problem of too much Information.

Avinash: That's right. That's exactly right.

Linda:  Because the data will automatically find the solution for you.

Avinash: For rare there is clarity we can apply intelligence and they have complete data or almost complete data no humans are required. But on the other side, there are lots of situations scenarios where there is incomplete data, you don't have this data is good or bad or we are thinking of something entirely new for which data doesn't exist at all. In all the scenarios humans can play a very important role, in fact a much larger role than the machine is playing for now. So I rather let's use people for that, is that's what people are good at it.

Linda: So we're still going to need marketers and they're just going to have a more creative job.

Avinash: That's exactly right they have more fulfilling role for sure. 

Linda: Yeah. Are there any interesting controversies in digital marketing you think?

Avinash:  Tons of them, every day there is a new one. 

Linda: If you would say the most interesting one.

Avinash: I think there is always the standard discussion in digital marketing about how do we understand people across multiple experiences. It's a very complicated problem to solve. You know, we all use multiple devices were exposed to multiple things. And so in order to really understand. Oh, what is your intent? It's a very big problem across these broken data-Sets. So that's one challenge that we have to figure out how to overcome. The other one is just the trade-off between the data that you're analyzing in the value you can create for the End customer. There always has to be an asymmetric exchange when there is an asymmetric exchange people always have negative reactions to that. So that's a very important and big problem to work through. In generally just ensuring that as we begin to understand people that ultimately the human Being at the other end which is our customer typically but your customer. The people who read your who listen to this, to your podcast that they ultimately control the data that they are sharing and they understand what data is being collected and they understand how it is being used and we have a permission based data economy. That's a really-really very important area of focus. So I think those are primarily the big challenges that are in front of us. I'm sure there will be more in the future but I think that we have to do a better job to ensure that we are creating user-centric strategies.

Linda: What do you think is relevant knowledge for the future? Of course, I would recommend everyone to take a look at your blog and also your books but is there anywhere you go for inspiration?

Avinash:  Yeah, lots of things. For machine learning and AI particularly I had written a blog post where I had outlined the videos to watch, the books to read and the people to follow and the courses to take side for things. So maybe I'll send you a link that you can add it to the show notes.

Linda: Yeah. 

Avinash: So then people know that if they want to start on this journey, here are the books, videos, people, and courses, the courses are all free so anybody in the world can take them.

Linda: Just like curiosity.

Avinash: That's exactly right.

Linda: Yeah.

Avinash: I highly recommend that even if you're a leader in the company the course I recommend has three different sort of escalating courses there. At least take the first one you will be able to do most of the first one you can give up after that if you're not technically oriented, but at least going through the first actually just stretches your mind in a very big way. Generally for me I find that the most interesting knowledge sits easily accessible to us on the website. There are a collection of people that I follow. I follow 88 people on Twitter, maybe about 40 of them fall in the professional area and the rest are you know, the Dalai Lama, who doesn't love the Dalai Lama? Everybody needs to follow them. So I think that that probably gives you a good exposure to the people that I follow. Most of it is I think to your point earlier understanding what your core love and interests are and then just being curious about seeking that knowledge because, literally any subject you want to learn about there is a video sitting on YouTube. There are 50 videos sitting on YouTube.

Linda: For free.

Avinash: Exactly.

Linda: So you just need time.

Avinash: That's it. It's a matter of priorities.

Linda: Yes agreed.

Avinash:  How you're going to use your time? People ask me. How do you do so many things here? You write newsletter and a blog and job and speak and I said sometimes a half joking me only say, I don't know what Game of Thrones is, that's how I make time.

Linda: You threw away your TV years ago. So if the listener should remember just one thing from our conversation what would you hope it will be or what would you think it should be?

Avinash: It's very important to understand that increasingly most of the jobs that we do today fall in the category of what I described in my writing as frequent high volume tasks. If there's one thing that people should take action on is try and see what would it take for them to reflect on the position they are in today. And say if my career is going to be only tackling normal situations which I predict is the future role for all human beings, then what will it take for me to get to there from where I am today, tackling frequent high volume tasks, to being able to be very comfortable and operate in a world where I am only tackling normal situations. It requires a completely different emotional aptitudes and certainly a very different intelligence aptitude. Just reflecting on that difference I think will allow you to create a plan that is unique to yourself so that you will thrive and be happy over the next couple decades. 

Linda: Thank you so much for coming to our podcast and luring us more about digital marketing, AI and everything else.

Avinash: Thank you so much. Thank you for having.

Linda: Thank you for listening.