The Actionable Futurist Podcast

S3 Episode 4: Bret Greenstein from Cognizant on Evolutionary Artificial Intelligence

May 10, 2021 Andrew Grill Season 3 Episode 4
The Actionable Futurist Podcast
S3 Episode 4: Bret Greenstein from Cognizant on Evolutionary Artificial Intelligence
Chapters
The Actionable Futurist Podcast
S3 Episode 4: Bret Greenstein from Cognizant on Evolutionary Artificial Intelligence
May 10, 2021 Season 3 Episode 4
Andrew Grill

In this episode, we spoke to Bret Greenstein who is Global Head of AI & Analytics for leading consulting firm Cognizant.

Brett leads Cognizant’s Global Data Practice, focused on helping  Chief Data Officers transform their businesses through Data Modernization. 

Prior to this, Bret lead IBM Watson’s Internet of Things Offerings, establishing new IoT products and services. He holds patents in the areas of collaboration systems, a bachelor’s degree in Electrical Engineering and a master’s degree in Manufacturing Systems Engineering.

In this wide-ranging discussion, we uncovered new uses for AI, around the notion of "Evolutionary AI".

We also discussed how AI can help governments and countries prepare for the next COVID-19 pandemic, and Cognizant's involvement with  XPRIZE to launch the Pandemic Response Challenge, a competition to create AI systems that can help societies reopen safely in the wake of COVID-19. 

Cognizant’s work was based on data from Oxford and John Hopkins and provided data-centric strategies for governments that want to support healthcare professionals in the fight against future pandemics. 

We also covered:

  • The difference between Artificial Intelligence, Artificial General Intelligence and Evolutionary Artificial Intelligence
  • Other implementations beyond XPRIZE for Evolutionary AI
  • Bret's view on General AI and how far we are away from it in practice
  • The importance of ethics and AI, and conscious bias
  • What more can be done in schools to prepare students for a world dominated by AI?
  • What Governments be doing to implement the sorts of solutions uncovered in the challenge to help our societies get back to a pre-Covid world
  • The notion of "data credits" to encourage data sharing to solve broader problems
  • How AI augments people’s decisions and finds meaning from the noise
  • The suggestion that data and access to data should be an asset on a company balance sheet
  • How AI can be used to develop better predictions for a range of business problems
  • Some practical advice on what you can be doing to better understand the power of AI and in particular Evolutionary AI
  • Bret's view on how AI can better connect with humans

My favourite quote of the episode was

"Climbers climb mountains, data scientists find meaning in data"

We've had a number of AI experts on the show, and this episode was one of my favourites because of the way Bret explained very complicated topics in an easy to understand way.

More on Bret

Linkedin
Cognizant

Your Host

For more on Andrew - what he speaks about and replays of recent talks, please visit ActionableFuturist.com follow @AndrewGrill on Twitter or @andrew.grill on

Show Notes Transcript

In this episode, we spoke to Bret Greenstein who is Global Head of AI & Analytics for leading consulting firm Cognizant.

Brett leads Cognizant’s Global Data Practice, focused on helping  Chief Data Officers transform their businesses through Data Modernization. 

Prior to this, Bret lead IBM Watson’s Internet of Things Offerings, establishing new IoT products and services. He holds patents in the areas of collaboration systems, a bachelor’s degree in Electrical Engineering and a master’s degree in Manufacturing Systems Engineering.

In this wide-ranging discussion, we uncovered new uses for AI, around the notion of "Evolutionary AI".

We also discussed how AI can help governments and countries prepare for the next COVID-19 pandemic, and Cognizant's involvement with  XPRIZE to launch the Pandemic Response Challenge, a competition to create AI systems that can help societies reopen safely in the wake of COVID-19. 

Cognizant’s work was based on data from Oxford and John Hopkins and provided data-centric strategies for governments that want to support healthcare professionals in the fight against future pandemics. 

We also covered:

  • The difference between Artificial Intelligence, Artificial General Intelligence and Evolutionary Artificial Intelligence
  • Other implementations beyond XPRIZE for Evolutionary AI
  • Bret's view on General AI and how far we are away from it in practice
  • The importance of ethics and AI, and conscious bias
  • What more can be done in schools to prepare students for a world dominated by AI?
  • What Governments be doing to implement the sorts of solutions uncovered in the challenge to help our societies get back to a pre-Covid world
  • The notion of "data credits" to encourage data sharing to solve broader problems
  • How AI augments people’s decisions and finds meaning from the noise
  • The suggestion that data and access to data should be an asset on a company balance sheet
  • How AI can be used to develop better predictions for a range of business problems
  • Some practical advice on what you can be doing to better understand the power of AI and in particular Evolutionary AI
  • Bret's view on how AI can better connect with humans

My favourite quote of the episode was

"Climbers climb mountains, data scientists find meaning in data"

We've had a number of AI experts on the show, and this episode was one of my favourites because of the way Bret explained very complicated topics in an easy to understand way.

More on Bret

Linkedin
Cognizant

Your Host

For more on Andrew - what he speaks about and replays of recent talks, please visit ActionableFuturist.com follow @AndrewGrill on Twitter or @andrew.grill on

Intro:

And welcome to the pragmatic Futurist podcast a show all about the near term future with practical advice from a range of global experts to help you stay ahead of the curve. Every episode answers the question, what's the future of with voices and opinions that need to be heard? Your host is international keynote speaker and pragmatic Futurist Andrew Grill.

Andrew Grill:

Welcome to the pragmatic futures podcast. Today's guest is Brett Greenstein, Brett leads cognisant global data practice focused on helping Chief Data officers transform their business through data modernization. The data practice provides essential capabilities to help companies architect for and adopt cloud, embrace external data, increase accessibility of data, and ensure data governance at scale. Brad is also helping to drive data transformation across cognizance business units, and participated in industry advisory boards to shape data policy, accelerate education, and share best practices. Prior to this, Brett led IBM Watson's Internet of Things offerings, establishing new IoT products and services for the industrial Internet of Things. He holds patents in the areas of collaboration systems, a bachelor's degree in electrical engineering, and a master's degree in manufacturing systems engineering. Welcome, Brett.

Unknown:

Thank you so much, Andrew, it's great to be here.

Andrew Grill:

I just realised we have a lot in common because I also have some engineering degrees. I don't have any patents on my belt. But it sounds like you came from a sort of an engineering background as well, the out of the possible Is that fair?

Unknown:

It is I think when you grow up as an engineer, you have a certain way of thinking that can be applied to business. And that's what I kind of learned through my career was really how to take systems thinking and problem solving an engineering into into business and how to marry technology and business together.

Andrew Grill:

Now we both worked at IBM some time ago and cross paths, then I'm sure but more recently joined Cognizant, what is your role at Cognizant Intel at the moment.

Unknown:

So we do everything around digital transformation, and very much I'm focused on turning data into a powerful competitive advantage for companies. And certainly over the last couple of years, the adoption of cloud the integration of external and third party data, the ability to make data accessible to all decision makers within the enterprise is transforming product engineering experiences, how companies go to market, how they handle risk, pretty much everything is being touched by data, and transformed through data and AI.

Andrew Grill:

Now I've had a lot of experts on the show the last couple of years to talk about AI from general AI to outline the multiple uses of the technology, we've had my friend mean to dial on to talk about the ethics of AI, you've introduced a new term to me evolutionary AI. First of all, what is this? And how does it differ to the AI that we see in the press?

Unknown:

Well, one, the AI you see in the press is all over the map, I think there's way too much hype, you know, around it, you have to be more focused and specific. And with evolutionary AI, we've turned some techniques in place so that AI helps shape AI. In other words, when you design a model, a neural network to do something to recognise an image or to recognise tax, or to predict something, it's a reasonably tedious task today, we can use evolutionary techniques, which basically uses models to train and refine the models and make them significantly better. And I think what we found so far is that almost anything that was designed by a person can be enhanced and optimised further by running it through effectively, scenarios to find the optimal configurations of a network. But more importantly, it actually enables simulation, and what if scenarios at scale. So we can look, for example, at what the optimal decision is, when you're setting for example, price of a product, we can look at all past decisions, we can look at all conditions that have changed, and run huge parallel simulations extremely rapidly and efficiently. And then know what the best price will be given the conditions that exist to go against a business goal you have. And that's a very, very new type of AI applications, prescriptive analytics, and, and it really is enabled by cloud, as well as these, these genetic algorithms, these ability to evolve a solution through simulation. It sounds

Andrew Grill:

like a really practical use of AI, which is something I love. And because I'm the pragmatic Futurist, I like to look at things in a very practical and actionable way. And I love how you've actually put this into action and cognitive develop the pandemic response challenge. First of all, can you tell our listeners a bit more about the challenge and why you went about doing it?

Unknown:

Yeah. And to do that I have to go back just a couple of steps during the pandemic, which obviously was top of mind for everybody. The teams at Oxford began to collate and curate data about the spread of the disease at the country level and then later at the state level. They also added something unique, which was they looked for the non medical interventions, the thing policies that countries and states were doing to help combat it for example, social distancing, limiting international travel, restricting work and things like that. That each one of those decisions has an effect on the number of cases. They collected and curated all that data they updated every single day. It was a, it was an army of volunteers. And they did it because they felt the data would help enlighten everyone on what's happening. Our data scientists who work on evolutionary AI looked at that data and saw something additional. They saw that they could build both a predictive model using it, but also an optimization engine to help recommend decisions based on varying goals. If your goal is maximum public safety, here are the policies you should implement to maximise that, and here's what effect it will have. And you'll learn a tonne by doing that. It's like for example, some policies have no effect. And they're different by country and the different by state. So we use real data trained it every single day, every night, we retrained our system based on the Oxford data, that's a reasonably new way of working. And we published it, we published papers about how it worked, we put it on the web for anyone to look at to see any state any country, you know, forward looking view. And it was really, it was powerful. It was sobering, you know, because we can see, you know, a lot of things, a lot of dynamics happening in the various locations. And then the XPrize team noticed it, and thought, what if we could enlist data science teams around the world, to build better predictive models to build better prescriptive guidance? What if we unlocked data science around this data? And so we partnered with X PRIZE to do that, and the response was overwhelming.

Andrew Grill:

So this was a competition, but was the data used in anger? And what you described is stuff that governments would normally spend billions and billions actually, running simulations for sounds like this is great open source data, did anyone actually use

Unknown:

the Oxford data was being used by many people. And because it was unique? Well, it did something very special, they created a standardised data model that allowed those policies to be turned from just raw text into something actionable. If you look at it policies were implemented and distributed through RSS feeds and websites and things like that, when you turn them into a standard data model, data scientists can do something with it, and countries and corporations were looking at the data. But to really unlock it, you needed a system, an AI based system to unlock it, by publishing how we did it. It gave people a roadmap on how to use the data. And then with X Prize, we actually standardised some rules and some frameworks and gave some standard services to make it easier for people to innovate. And then the team and one of the things you'll notice in AI projects is 90% of the work is the data work. The fact that Oxford did that. And the fact that we built a cloud based environment to start from meant the data science teams, science teams could focus on the problem. And not as much on the data part of it it really unlock their potential. Yeah, I

Andrew Grill:

mean, ai does the heavy lifting for you, because there's so much data out there. And especially in a global live pandemic, that's where AI comes into its own. So I can imagine, yes, you could apply that to pricing, you could apply to all sorts of things. Now you had more than 100 entries to the XPrize. What surprised you most about the entries you received when they

Unknown:

came from every part of the world. And so the global participation, which you'd expect in a pandemic, was also reflective of the talent, and how it's distributed around the world. You know, and teams were from everywhere. And they were also, you know, some small teams, some larger teams, some commercially driven folks, some folks who were more from academia. But I think it brings together like minds, people who see patterns and data that regular people never see, and, and challenge them to create. And there were some very passionate teams. And like, like any other group of passionate people, there were debates and arguments, there were discussions on forums, I think it was the interaction, the debating the lively energy that was into it. That was really exciting. You know, there were people who literally just focused all of their free time on doing this, partially because of the prize, but also partially because, like anything else, it's why'd you climb the mountain, it was there. Mountain Climbers climb mountains, data, scientists find meaning in data. And here's the world's largest problem presented right in front of everyone with data they could act on. And it really, I think that's what inspired that passion.

Andrew Grill:

I talk to clients a lot about the use of hackathons. And often they say, well, that's for techie people. But this sounds like it was almost a global hackathon. Would that be one way describe it,

Unknown:

it had that feeling? I think that hackathons have perhaps a scrappier vibe, there's a lot of structure to this and a lot of thought it was also measured in terms of competitive benchmarks that are absolutely not, you know, subjective. And so in a lot of ways, there's a very well thought out, structured competition, in contrast, where hackathons are often, you know, based on judges and more subjective measures, but the same energy goes into it and the same innovative thinking and also the fact that teams can solve form and can compete and progress through a competition that has that hackathon feeling to it. But I think when you do, especially for data scientists, they're always working to elevate the state of the art, you have to be able to prove what you've done is better. And so we had to come up with a lot of structure with XPrize, to figure out how to measure the results

Andrew Grill:

now with your work at IBM and Watson, and also more recently, we commerce and I'm sure you've seen so many different uses of AI. But we even you surprised that some of these teams uncover new uses you hadn't even thought about,

Unknown:

they certainly brought in new types of data, I think the the state of AI, everyone in this space kind of sees it evolving very rapidly, deep learning, reinforcement learning, you know, General adversarial networks that the technology keeps advancing. But what I see people doing is really bringing external data and other data sources, combining it together to provide more meaning. And also, we found that when we combine models, we created more resilient predictions. So a lot of people will model based on assumptions. For example, they might assume the efficacy of something, or they'll assume when we put a mass compliance rule in what percentage of people will actually comply, we learned through this that was actually more valuable to have no assumptions, and to do the models based on the data itself, the output. So we look, for example, the Oxford data updated every day. And if you put a policy in place in one location for masks, and another location for masks, people naturally based on whims and politics and just feelings will comply more in one place than another. It's cultural, there's a lot of reasons for it, you can't assume that putting a rule in actually works. So you have to look at what the effect of the rule was. What you find, for example, is that mass compliance in some places has absolutely no impact on the caseload. Because people just don't comply with it. In other places that had a huge effect. And so decisions and their impacts are very local, they're local to the culture and feeling of people and how they follow those rules. Social distancing is not followed everywhere. So even though you implement it, it doesn't mean that it will have the same effect everywhere. So we tried to really take a data first view, not an assumption based view. And because the data was updated every day, and the models were retrained every day, it it tracked really quickly with the dynamics happening in the pandemic

Andrew Grill:

brings me to interesting thought about conscious bias. So whenever we talk to AI experts, the the whole issue of conscious bias comes up. And you've uncovered a really interesting fact that rather than an assumption, which is an unconscious bias, you've said, Let the data do the talking. So stepping back, can we remove the the issue of conscious bias by allowing the data to have the control,

Unknown:

I wish that were the case, the problem is the data itself is also bias. So in the case of the pandemic, because the data was real time and not historical, you eliminate one of the elements of bias, which is historical data, which reinforces historical decisions. The pandemic data was real time and so that that did reduce that part. But in setting up the model, and I don't think bias was introduced here, but in setting up the model it you can introduce bias and even the data and the fields that you look for. And so for the for the future of bias and AI, you need to have people who work on the data being diverse themselves, so they don't bake in assumptions and amplify those assumptions in which data they collect. For example, when in HR recruiting systems, if you look at past performance, like who you hired in the past, who happened to excel, you might find patterns that actually reflect the bias of the hiring demographics you had in the past. So the people who shaped data need to have a diverse point of view, so that they don't introduce bias, you need to check it constantly to make sure you're really missing that and updating your thinking about the data. So you're not so as you think more contemporary view, historic data has inherent bias. And you've got to be aware of that. And then, of course, the models themselves reducing assumptions does reduce bias.

Andrew Grill:

So that brings me to the notion of diversity of thought, I'm a big proponent of diversity of all types, there's diversity of thought, how do you introduce diversity of thought on teams like that? So you are, you're normalising any biases that are there,

Unknown:

we do this for ourselves, we run a responsible AI Council and we help clients to set this up too. It starts by having not delegating the entire work to data scientists themselves. So when you're doing a project of any type, you've got to have representation from the business, from HR from legal from the perspectives that are inherently diverse. And then you do have to look at the diversity of your leadership team itself, to make sure they're not reinforcing, you know, past views. The other thing that comes into it is you start to have to ask yourself about the relevance and how you might influence data that is longer term, historical versus more recent weather, you're going to be reinforcing past biases. The best thing that comes from these things is actually asking the question, is it bias? Does it reinforce bias? Are we doing the right thing? Are we getting the right points of view? It's the asking of the questions that actually starts to generate the change of behaviour. And then of course, ensuring the team itself has a representation that is beyond the data scientist, which, by the way, is just one technical view, but also that group itself may not be diverse, and so it's You really have to just kind of look at this systemically. And from a governance point of view

Andrew Grill:

to x prizes over the winners have been announced. What implementations Do you see beyond X Prize for evolutionary AI?

Unknown:

evolutionary is already being used by clients. And in almost every industry, we're using it to optimise pricing and underwriting and to better test medical conditions. We've used it to help predict and prescribe, you know, cosmetic surgery, there's, there's like so many different uses, because decisions themselves, human beings, when we make decisions are not optimal. We balance for example, cost and customer set and margin and, and other factors and a human can optimise across three or four parameters. And that's it. You what you really want is a system that can look and simulate all the possible decisions and come up with a narrower set of best decisions to make. And we find the business leaders, when they're making decisions on supply chain, or inventory, or pricing, or even where to put products on shelves. They're doing it based on their past history, their own worldview, and not looking at every possible combination. When we run evolutionary AI and those optimization problems, they can literally look at millions of possible decisions. Imagine a store manager making millions of decisions doesn't happen. They decide product A belongs on the shelf at this height because it usually works until it doesn't. And then the pandemic really highlighted, I think for everybody how dynamic the world really is. We had many companies come to us during the pandemic, who said, our customers aren't behaving as they used to. They're not behaving normally. And and that's because obviously nothing was the same in 2020, as it was in 2019, or every year before. So we had to build new models, we used evolutionary AI to predict supply and demand for consumer beverages, for example. And we have come up with multiple models. And it was all about the change of behaviours. People didn't go to convenience stores anymore. people bought in bulk, people just behave differently. So how they bought changed. And you could see that, for example, in the spike in, in, in supply constraints for flour and yeast during the pandemic, when people were baking at home, consumer behaviour shifted, you have to adapt with it. And that's where tools like evolutionary AI help adapt more quickly based on the actual data.

Andrew Grill:

So when I went to MBA school, we learned in the retail environment, you had the milk at the back of the store, because that allowed people to go all the way to the back and the confectionery aisle had to have bright lights because that encouraged you to buy wood I evolutionary I debunk those theories based on fact, and maybe we'd have to rewrite the NBA marketing books because it isn't the truth,

Unknown:

I think what you have to consider is that it's it would refine the thinking the thinking was true and probably is still true at a macro level. However, at this moment, on this day, in this store, in Tulsa, Oklahoma, is that the right decision, when things are shifting, they happen locally, they happen in real time. And those MBAs didn't have at the time access to all the social data we all have now. So if you had been watching social traction around home baking and sourdough bread, you would imagine there would have been shortages of the supplies of those things. And it varies based on region and in the country or around the world. And so most of our customers were telling us these were very localised, very dynamic decisions, and they needed more guidance to make faster decisions. Milk in the back of the store probably is great. But what if you could also put something else next to it because of what's happening and social that might help drive up the sales of that, and a week from now, that won't be true. So those decisions can be much more dynamic than they've been in the past,

Andrew Grill:

my time spent in socialists, that social media is the best piece of market research you never commissioned. It's there. It's raw, it's live, the challenge is actually filtering it so that the AI systems can make sense of it, because humans don't always talk the same way. So how do you how do you train a system with something as noisy as social to make those sort of smart decisions,

Unknown:

we use natural language processing to extract data, and you have to have people who are good at having hypotheses about data and understanding which terminology and language would be consistent with certain meaning will do sentiment analysis, you know, which tells how people feel, but you also have to look for keywords. And generally there's a, what's called a bag of words, techniques, which allows words that are common and related, that you're looking for, there'll be indicative of things trending. The interesting thing about that sourdough example is since no one was looking for it before, they didn't look for it during. And I think what is really highlighted for everyone is you have to open your aperture, look for things that are trending, and then look for what's happening under the trend. So when that was trending great, you know, that was sourdough. But the question is, what are the ripple effects of that? And what do you look for? And I think all of us have to be a little more dynamic in that as well. There is a lot of data and a lot of noise. But, you know, you can see the companies that are really good at looking through the data finding meaning in that in the data. social data is very powerful. local data, for example, what's showing up in RSS feeds in towns, what's top events and things like that. It's right there for us. LinkedIn data, I everyone I talked to, I look on LinkedIn to see You know, what kind of roles are they hiring beforehand? because it tells me what their priorities are? How many people really look at that data? You really, there's so many indicators of what people care about. If you look,

Andrew Grill:

I remember my father teaching me years ago, his small business, he would look at the job ads in the local newspaper to see who was hiring, because if I were hiring that showed demand simple signals like that, that are now amplified and social can be so important. You're an AI practitioner, you've been doing this for a while, I have had a number of discussions with people about the notion of general AI. As a Futurist, I'm actually keen to learn where do you think we are? How far away are we from it

Unknown:

and isn't real. This is a topic we could talk about for days, I'd love to do that actually, with you, I got into this space because of the dream of AGI of general intelligence. However, practically speaking, we are so far away, what we really have is the simulation of intelligence. And so when you look at AI that can recognise what's in a picture, or read an X ray, or understand or appear to understand language that feels magical, it feels intelligent. When you see an AI like open AI that can generate text that actually sounds perfectly natural sounds like something you were I would, right? You say that's intelligent. But it's really pattern recognition and and simulation of us the same way that mechanical Turks and systems for hundreds of years have have represented how people behave in a way that feels real. There are companies now, for example, that are training AI systems on what we individually say, to create a virtual version of us. William Shatner recently for his for his birthday, had himself recorded in a way that that people can interact with him for generations to come. And it'll talk as if he were talking, but it's not him. And it's not actually intelligent. It's, it's mimicking the patterns of what people do the same way that I think, I know, this is gonna sound terrible, the way the parrots, you know, sound like people, but they've learned how to respond in a way that we respond to. And that's really reflective of sort of where AI is today. I think we all dream and aspire towards an AI that is more representative of really intelligent decisioning and things like that. And that's where we're all headed. We shouldn't worry too much about AI being us. It augments people's decisions. Because it's really good at patterns, it's really good at looking at wide sets of data quickly, and then sorting through it to find meaning from the noise. This is what it's best at. And it's what we all use it for today. And that's really the practical applications. And then the future, the future is still bright. But I'm not I'm not worried about the future.

Andrew Grill:

I often talk about AI being augmented intelligence, as you said, because it's not just artificial, it has so many uses. I do a lot of talks to school children, people that are coming up, and they're looking at, you know, defining what they want to do in their career. What more can we do in schools to prepare students for a world dominated by AI?

Unknown:

While the world will have a finite number of data scientists and people who build AI, there's almost an infinite number of us who will be using AI in the course of our jobs in the course of business, that MBA class you mentioned before putting milk in the back of the store, we should be teaching people what it what data can mean for business, how to look for causality in data, not correlation, to find real meaning in it, to what signals mean, how to look for them. And of course, the potential and what AI does versus what it doesn't do. I would like to see a more data and AI literate community of people coming up through school, because as they look at their disciplines, every discipline is being touched by this. I mentioned before, whether it's cosmetic surgery, or underwriting or supply chain, you know, imagine, imagine the store manager for convenience store having to consult with AI to figure out where to put things on shelves, or the dock worker deciding what order to load up a ship, the person getting us on and off an aeroplane, maybe there's an optimal way to do that, you know that AI could help with, we should be informed and augmented by systems that can recommend that the best thing to do. That means we must have better judgement, to not just listen, literally, the first I remember as a as a child, I was always fascinated with engineering and science. But my teacher reminded me common sense first, estimation first so that you know that the recommendation that comes from a system actually is close enough to be possibly right. And it's the judgement about what data and AI means and how it can be used, that will make the best business leaders of tomorrow when we looked at people building web based businesses, the leaders of those businesses were not the best webmasters, they were the people who understood the potential of the web and the network and the connectivity and the social implications of it. You know, I'm sure Jeff Bezos was not the best webmaster at Amazon. But he understood the implications of it to great degree.

Andrew Grill:

So when I was at school, they made me learn Latin, and I chose to learn French. What you're saying maybe is the data is a language and maybe data literacy should be taught so that people understand data and ask the right questions. I mean, maybe I'm sure it's happening already. But is that a subject that you should suggest school children learn?

Unknown:

Absolutely. I think that problem means that the discipline itself is not well defined yet, when people talk about data literacy, they often think, what does this data mean? Instead of what can data mean? And I think that how to look at the patterns and this and the types of data, the types of implications are going to have also how to how to recognise bias and data, how to recognise when data is good, what to do when you have poor data quality, which almost everyone has, what do you do about that I've had so many people come to me and say, Well, our data is not very good. We can't use AI, or we can't use analytics, the data is bad, but it might be systemically and consistently bad, which means you can account for it and do something about it. You know, I had a car once where the the speedometer was always five miles per hour low, because the tires were the wrong size. But I knew it was wrong. I knew by how much it was wrong. And I didn't get a speeding ticket. And I there's a lot of stuff in business that is quite similar to that. And I think getting past the fear of bad data, getting to recognise the potential of it, where to use it, how to find causal relationships, how to deal with bias, great business leaders will recognise this innately. And they will generate businesses which are highly differentiated because they're data driven at their core, not a bolt on where someone just throws another report on top of the business and says, look, we use data.

Andrew Grill:

So going back to the challenge, a lot of the outputs there could inform policy decisions, either in real time or longer term. So what more can government's be doing to implement the sort of solutions that were uncovered in the challenge to help as societies get back on track to a post COVID world?

Unknown:

Well, I think it's a challenge for governments as well as for companies, I think it's an opportunity for companies because there's a lot of data related to the pandemic, and future threats, whatever they may be, that accompanies hold. They know when people, for example, are buying cough and fever suppressant, long before people are tested for a sickness, if that spike, suddenly, that's a signal that should tell governments and companies to do something to act on it. We need to create more early warning systems within our society, that companies and governments can all participate in. And when we see signals that indicate health challenges, for example, we need to we need to share that we need to do something about it. And those signals are everywhere. They're you know, whether it's in in the local pharmacy selling significantly more fever suppressing than they have, in previous times, something's going on in that local area. You got to watch for that. And then I look at governments and I think the policies that are implemented are sometimes implemented, because they are easy, or because they are, you know, politically beneficial. However, how effective are they. And if you looked at this, we did a sort of a sensitivity analysis, when you looked at the policies in different states and countries. Some of them they implemented very strict policies that even if you had relaxed, it would have no effect on the on the pandemic. And other policies that would have strong effect weren't done for whatever reason they chose to do them. So I think just be more aware of the effect of a policy how effective it is. And then adjusting policies based on the real impact is probably something we can all learn from.

Andrew Grill:

He mentioned before about access to data. So if the trends are happening, but the issue is getting access to their data, and I've just read an article you published recently, about the need for countries and organisations to share data. And you talk about the notion of a data credit to facilitate this. What is that? And how would it work?

Unknown:

companies often think of their data as their competitive and proprietary advantage and something that shouldn't be shared. However, the most powerful insights come when you combine data. Recently, u haul published some data about the migration patterns of people over the last 12 months, they uniquely know a certain demographic, who likes to themselves, where they've moved in this country. That is a huge ripple effect on real estate prices on mortgage underwriting on the economy. And you know, even even on the census, I mean, this is the stuff that is really powerful data. By sharing it, they did something they made people aware, if we were to encourage sharing, which can be done safely with privacy doesn't mean identifying people, it means sharing what you've learned, if we could encourage sharing, we could unlock greater insights, whether it's governments or companies, and unlock economic value. the thinking behind the credits is if we were to consent, the sharing of data, the same way we incense, you know, carbon credits, for example, some countries, some companies who share more of that data and provide societal benefit should get credit for that whether it's a tax incentive or something else against targets that they might be set. Other companies might not be in a position to share, but might pay for those credits from other companies, therefore, also creating more economic value. So I think the the general incentive is if we do share, there is more value to society more economic value, we have to find ways to share safely and to encourage it.

Andrew Grill:

So a good example of this is the snowflake data marketplace. How can more companies get involved in this,

Unknown:

obviously working with snowflake, they've enabled something which is which is interesting, which allows people to store data once to authorise data through their ecosystem. And in a way really reduce the friction of sharing. Now companies have been sharing data for years through b2b marketplaces and all kinds of other technologies. It's just that it was harder to set up in the past, there was more cost because data was often duplicated in multiple places. And access control was more challenging. This is really creating almost a social network for data, which is going to allow a company to work better with its ecosystem, whether it's their suppliers, their customers, their partners, these are the kinds of things that will really unlock, I think, a lot of the potential. And if governments were to look at sharing mechanisms, and encourage and enable it in ways that were safe and secure, I think the potential value of that would be enormous.

Andrew Grill:

So if we had time again, if you and I are talking December 2019, the trending information is there, we're about to have a pandemic, it's going to be bad. How would AI have helped mitigate the impacts of COVID-19 if we had our time again,

Unknown:

so I'd like to think we did a lot during the pandemic, as soon as data was available to share best practices and methods and even, you know, predictions. Going forward, though, what I'd like to see us do because I would like to think forward looking here, there are many signals of what's coming. And sadly, there will likely be more pandemics, they'll certainly be effects of global warming, social unrest, whatever, you know, other things, or dynamics and chaos that happens around the world, we need to find those signals, we need to work together to share those signals. What Oxford did was really, really a great example of collaboration, they had an army of volunteers, they published, they did it for the good of society. Our data, scientists did the same thing. They they did this without being asked they did it without, you know, a customer telling us to do it. We just did it, because we will unlock that. And I think that's the kind of behaviours we want to see. But I'd like to see us all focused on the signals that will matter. What have we learned from the pandemic, what signals were indicative here, other sicknesses might have other signals. So we have to open our eyes to that, and other challenges will have different signals. So it's up to us, I think all of us to advocate for the power of data, the power of sharing the power of AI, the examples where it's been done, we do we do so much of this work with clients. And we share that work where the clients want to share those stories. I think sharing those stories, really unlocks economic value, and and certainly improves data literacy, literacy about AI and its potential. This is what really matters, I think, to the world.

Andrew Grill:

So just to back on your notion of data credits, we talk about various intangibles being on a balance sheet and brand equity, for example, is on a balance sheet, and sometimes it goes up and it goes down. Should data be an asset that's on a balance sheet, and access to that data is seen as an asset?

Unknown:

Without question, what we all forget is that we're each uniquely have knowledge that is valuable to someone, most of us don't have the chance to share it in a way that creates economic value. But there are companies now sitting on significant amount of value. Some of them are in a position to share it with their ecosystem, some are using it as part of their business model. But recognising the value of that data is huge. I think when I look at a company like Tesla, and the millions and millions and millions of miles that they've they've logged collecting data to enable self driving. There's an enormous value that's clearly baked into the valuation of the company. Because people recognise it. The question is, how would you measure it? How do you how do you think there's a still wide open questions for the industry. But I wouldn't be surprised in the years to come if the value of data starts to show up as a line item in balance sheets and begins to become a significant portion of the valuation of companies. And it is why you see companies talking about being coming, moving from product company to software company or to a data company. It's why the Internet of Things took off so wildly was the data coming from things has tremendous power and insights into healthcare, manufacturing, automotive almost every sector

Andrew Grill:

almost at a time. But as this is the pragmatic futures podcast, what three things can our listeners be doing to better understand the power of AI, and in particular, evolutionary AI? Three

Unknown:

things boiling down the entire world of and three things. I think the first piece is to recognise that the work in AI is very heavily data, making sure you have the right data, knowing which data matters, cleansing data, preparing data, pipeline and data. Don't underestimate that that is a huge, huge focus in any programme. I think the other part is then to bet to expect significantly better predictive models of what's coming. Anyone who's running decisions in a business, anyone who's making a decision about what's going on, who's an operations leader, shouldn't be looking at what happened last month, but they should have predictive models. It says based on everything they know, using AI, these are the things that might happen and what confidence you have in that you should expect your predictive models to include confidence levels, if there's a 57% chance this is the kind of thing that's going to happen. that informs us more. And then with evolutionary AI. We should be expecting prescriptive analytics recommendations on the very best decisions to make what is the best price to offer? What is the best thing to do in this moment, given everything we know and In the simulations, it doesn't mean you do it, it means you've got a recommendation on what your systems and your data thinks is the best thing to do. And then allowing you to use your judgement to really make that best choice,

Andrew Grill:

Wilson. So before you go, let's do a quick fire round. A short answer is a good answer. iPhone or Android, Android, PC or Mac, both biggest hope for 2021.

Unknown:

You just touched on something kind of sensitive. I hope that my children get to play with other children face to face without fear. Wilson,

Andrew Grill:

what's the one thing you won't be doing? Again, post pandemic,

Unknown:

possibly shaking hands? I still remember the very last day I shook hands with somebody, it's going to be a while till I do that, again.

Andrew Grill:

I'm thinking the same thing because even elbowing, I mean, yeah, that's all other podcasts. What are you reading at the moment?

Unknown:

It's a great question. I read a lot of things. I don't think I give you a book I there's too many things that are going on right now. But I think what I'm following is what people are talking about, especially in the world of AI and data. And I'm really following I think this this unlocking of the potential of AI to help better understand people. I'm very, we talked before about the thing with William Shatner, I'm fascinated by the fact that people are training AI, to better understand emotion to better. mimic is a strong word, but to reflect the communications that people have. And I've talked to many researchers recently who've been wanting to kind of capture themselves so that their children and grandchildren can interact with them in the future. And then I recently did a project where there's a new AI system that will take a still image and animate the image. And I took a picture of my grandfather who's passed away for quite some time, holding me as a baby. And I animated that, using that AI and shared it with my mom, and I wasn't sure if she would think it was wrong. Or she cried. It was beautiful because it It brought to life an image that has that was just a still image that we all look at, and it really brought it to life. So there's a connection between people that I think AI is beginning to fill in. And so instead of thinking of AI as becoming an intelligence, I think it just enhances the human experience in ways that I'm just fascinated by. So I, I read as much as I can around that topic.

Andrew Grill:

So probably my last question is quite pointy. And then how do you want to be remembered?

Unknown:

I'd like to think that I helped people become more aware of the implications and power of AI and perhaps inspired someone to follow a career they might not have otherwise.

Andrew Grill:

Brett, thank you so much for being on the show today. How can people find out more about you and your work?

Unknown:

We certainly published quite a bit on cognitive calm, as well. As you know, we share a lot on social. So between LinkedIn and blogs and series that we do, I think it's just important for us to keep a dialogue going in society. So we try to be very prolific in that, but certainly visit us at cognitive comm Follow us on LinkedIn, follow me on LinkedIn. We talk about these topics all the time. Thank

Andrew Grill:

you so much. I've learned a lot today and have everyone else have as

Unknown:

well. Thank you.

Intro:

Thank you for listening to the pragmatic Futurist podcast. You can find all of our previous shows at Futurist dot London. And if you like what you've heard on this show, please consider subscribing via your favourite podcast app so you never miss an episode. You can find out more about Andrew and how he helps corporates navigate a disruptive digital world with keynote speeches and C suite workshops delivered in person or virtually at Futurist dot London. Until next time, this has been the pragmatic Futurist podcast