Is Human Talent the Limiting Factor? Financial Inclusion, New Data and AI
by Keyur Patel and Marshall Lincoln
|
Published by the Center for Financial Inclusion, April 2019 |
The shortage of qualified data scientists has a ripple effect on financial inclusion beyond staffing requirements.
There’s much anticipation that combining new data and artificial intelligence will transform the pace of financial inclusion and the scale at which it’s possible.
It’s easy to see why. AI is being used to rapidly perform tasks that would otherwise require endless hours of human labor. It can, for example, sift through and analyze enormous volumes of non-traditional sources of data to expand access to credit. It can digitally verify identities, detect fraud, and improve customer engagement.
Institutions view AI as a tool that could greatly expand the number of customers they serve without having to scale their workforces by anywhere near as much.
This reasoning is compelling, but it obscures a wider point: the technologies that drive automation require people with specialized technical skills to implement, monitor, and refine them. These people are, for now at least, in very short supply.
We’re the authors of an upcoming report investigating the risks that will arise as financial institutions increasingly embrace AI (to be published by the Centre for the Study of Financial Innovation [CSFI] later this year). Our research suggests that one of the most fundamental of these risks – and a threat to institutions, their customers, and the financial system as a whole – is a pervasive skills gap.
There’s much anticipation that combining new data and artificial intelligence will transform the pace of financial inclusion and the scale at which it’s possible.
It’s easy to see why. AI is being used to rapidly perform tasks that would otherwise require endless hours of human labor. It can, for example, sift through and analyze enormous volumes of non-traditional sources of data to expand access to credit. It can digitally verify identities, detect fraud, and improve customer engagement.
Institutions view AI as a tool that could greatly expand the number of customers they serve without having to scale their workforces by anywhere near as much.
This reasoning is compelling, but it obscures a wider point: the technologies that drive automation require people with specialized technical skills to implement, monitor, and refine them. These people are, for now at least, in very short supply.
We’re the authors of an upcoming report investigating the risks that will arise as financial institutions increasingly embrace AI (to be published by the Centre for the Study of Financial Innovation [CSFI] later this year). Our research suggests that one of the most fundamental of these risks – and a threat to institutions, their customers, and the financial system as a whole – is a pervasive skills gap.
IMPACT - Part 3: Become a world-class communicator
by Marshall Lincoln and Keyur Patel
|
Published by Data Science Central, April 2019 |
his is part 3 of a 3 part series: “How to make your mark on the world as a talented, socially conscious data scientist.”
“Seek first to understand - then to be understood” - Stephen Covey
As a data scientist, you may often feel you live in an entirely different world than your stakeholders. Your experiences, skill sets, and interests may be vastly different (at odds, even) with those who rely on you. To be an exceptional data scientist, you must seek to understand others - not demand that they seek to understand you.
When you join a new organization, you will be inundated with hundreds of new terms, acronyms, processes, and the general enormity of that specific business apparatus which is yet unknown to you. You must seize upon these unknowns with relish, and regard the lot of them as an opportunity to learn, to educate yourself, and to improve yourself as a data scientist and businessperson. If you do not, you will likely fail to understand what really matters, and fail to deliver a product that really makes a difference.
In business, ignorance is not bliss. Ignorance is an inexcusable product of laziness and short sightedness, and serial ignorance is a fireable offense.
It’s not the ‘not knowing’ that is a problem. It is ‘knowing that you don’t know, and not seeking to fix that.’ You must be willing to ask questions. You must be willing to look stupid. You must be willing to do so often. As a data scientist, your job is to be as well versed in your company’s business model as your CEO. This simply won’t happen if you do not ask questions.
“Seek first to understand - then to be understood” - Stephen Covey
As a data scientist, you may often feel you live in an entirely different world than your stakeholders. Your experiences, skill sets, and interests may be vastly different (at odds, even) with those who rely on you. To be an exceptional data scientist, you must seek to understand others - not demand that they seek to understand you.
When you join a new organization, you will be inundated with hundreds of new terms, acronyms, processes, and the general enormity of that specific business apparatus which is yet unknown to you. You must seize upon these unknowns with relish, and regard the lot of them as an opportunity to learn, to educate yourself, and to improve yourself as a data scientist and businessperson. If you do not, you will likely fail to understand what really matters, and fail to deliver a product that really makes a difference.
In business, ignorance is not bliss. Ignorance is an inexcusable product of laziness and short sightedness, and serial ignorance is a fireable offense.
It’s not the ‘not knowing’ that is a problem. It is ‘knowing that you don’t know, and not seeking to fix that.’ You must be willing to ask questions. You must be willing to look stupid. You must be willing to do so often. As a data scientist, your job is to be as well versed in your company’s business model as your CEO. This simply won’t happen if you do not ask questions.
IMPACT - Part 2: Be a problem solver first, an engineer second.
by Marshall Lincoln and Keyur Patel
|
Published by Data Science Central, March 2019 |
This is part 2 of a 3 part series: “How to make your mark on the world as a talented, socially conscious data scientist.”
The world needs innovative leaders, critical thinkers, and pragmatic business people who know how to use data science methodologies to solve real problems.
This requires a thorough grasp of the business model, operational challenges, and markets in which your product or service operates. First and foremost, it requires a deep understanding of the customer you’re serving. All too often, the ambitious and talented data scientist forgets about the real world implications of their hyper-optimized ML model. (Hyper-optimized for what, exactly?)
Adversarial neural networks are not to be fetishized.
The wheat farmer does not lie awake at night thinking about how to make a better plow. He’s thinking about how to grow more wheat on the same amount of land. If a better plow will help him increase yields - and he knows how to make one - then he will develop that plow. But the plow is only the means to an end. In the same way, the compute, storage, modeling, visualization, and networking technologies at a data scientist’s disposal are tools in a toolbox - not an end unto themselves.
You don’t improve your product by tweaking hyper-parameters on your regression model to get a better MSE - at least, not necessarily. Improving your error metric is only useful insofar as your model is correctly aligned to deliver a practical solution to the real world problems at hand. Before you start comparing the performance of a random forest versus a support vector machine versus a neural network, be damn sure you are optimizing for the right problem.
The data scientists who will transform our world and solve humanity's greatest challenges are the ones who can wrap their heads around complex challenges and apply their technical expertise toward solving them. Crucially they are they then able to communicate with their technical and non-technical colleagues, their customers, investors, and their partners - to coordinate end-to-end testing, delivery, and operationalization of products which provide useful solutions to real problems.
The world needs innovative leaders, critical thinkers, and pragmatic business people who know how to use data science methodologies to solve real problems.
This requires a thorough grasp of the business model, operational challenges, and markets in which your product or service operates. First and foremost, it requires a deep understanding of the customer you’re serving. All too often, the ambitious and talented data scientist forgets about the real world implications of their hyper-optimized ML model. (Hyper-optimized for what, exactly?)
Adversarial neural networks are not to be fetishized.
The wheat farmer does not lie awake at night thinking about how to make a better plow. He’s thinking about how to grow more wheat on the same amount of land. If a better plow will help him increase yields - and he knows how to make one - then he will develop that plow. But the plow is only the means to an end. In the same way, the compute, storage, modeling, visualization, and networking technologies at a data scientist’s disposal are tools in a toolbox - not an end unto themselves.
You don’t improve your product by tweaking hyper-parameters on your regression model to get a better MSE - at least, not necessarily. Improving your error metric is only useful insofar as your model is correctly aligned to deliver a practical solution to the real world problems at hand. Before you start comparing the performance of a random forest versus a support vector machine versus a neural network, be damn sure you are optimizing for the right problem.
The data scientists who will transform our world and solve humanity's greatest challenges are the ones who can wrap their heads around complex challenges and apply their technical expertise toward solving them. Crucially they are they then able to communicate with their technical and non-technical colleagues, their customers, investors, and their partners - to coordinate end-to-end testing, delivery, and operationalization of products which provide useful solutions to real problems.
How AI can help solve some of humanity’s greatest challenges – and why we might fail
by Marshall Lincoln and Keyur Patel
|
Published by KDnuggets, February 2019 |
AI represents a step change in humanity’s ability to rise to its greatest challenges. We explore three areas in which AI can contribute to the UN’s Global Goals - and why we could fall short.
In 2015, all 193 member countries of the United Nations ratified the 2030 “Sustainable Development Goals” (SDG): a call to action to “end poverty, protect the planet and ensure that all people enjoy peace and prosperity.” The 17 goals – shown in the chart below – are measured against 169 targets, set on a purposefully aggressive timeline. The first of these targets, for example, is: “by 2030, [to] eradicate extreme poverty for all people everywhere, currently measured as people living on less than $1.25 a day”.
The UN emphasizes that Science, Technology and Innovation (STI) will be critical in the pursuit of these ambitious targets. Rapid advances in technologies which have only really emerged in the past decade – such as the internet of things (IoT), blockchain, and advanced network connectivity – have exciting SDG applications.
No innovation is expected to be more pervasive and transformative, however, than artificial intelligence (AI) and machine learning (ML). A recent study by the McKinsey Global Institute found that AI could add around 16 per cent to global output by 2030 – or about $13 trillion. McKinsey calculates that the annual increase in productivity growth it engenders could substantially surpass the impact of earlier technologies that have fundamentally transformed our world – including the steam engine, computers, and broadband internet.
AI/ML is not only revolutionary in its own right, but also increasingly central to the foundation upon which the next generation of technologies are being built. But the pace and scale of the change it will bring about also creates risks that humanity must take very seriously.
Our research has led us to conclude that AI/ML will directly contribute to at least 12 of the 17 SDGs – likely more than any other emerging technology. In this piece, we explore potential use cases in three areas which are central to the Global Goals: financial inclusion, healthcare and disaster relief, and transportation.
In 2015, all 193 member countries of the United Nations ratified the 2030 “Sustainable Development Goals” (SDG): a call to action to “end poverty, protect the planet and ensure that all people enjoy peace and prosperity.” The 17 goals – shown in the chart below – are measured against 169 targets, set on a purposefully aggressive timeline. The first of these targets, for example, is: “by 2030, [to] eradicate extreme poverty for all people everywhere, currently measured as people living on less than $1.25 a day”.
The UN emphasizes that Science, Technology and Innovation (STI) will be critical in the pursuit of these ambitious targets. Rapid advances in technologies which have only really emerged in the past decade – such as the internet of things (IoT), blockchain, and advanced network connectivity – have exciting SDG applications.
No innovation is expected to be more pervasive and transformative, however, than artificial intelligence (AI) and machine learning (ML). A recent study by the McKinsey Global Institute found that AI could add around 16 per cent to global output by 2030 – or about $13 trillion. McKinsey calculates that the annual increase in productivity growth it engenders could substantially surpass the impact of earlier technologies that have fundamentally transformed our world – including the steam engine, computers, and broadband internet.
AI/ML is not only revolutionary in its own right, but also increasingly central to the foundation upon which the next generation of technologies are being built. But the pace and scale of the change it will bring about also creates risks that humanity must take very seriously.
Our research has led us to conclude that AI/ML will directly contribute to at least 12 of the 17 SDGs – likely more than any other emerging technology. In this piece, we explore potential use cases in three areas which are central to the Global Goals: financial inclusion, healthcare and disaster relief, and transportation.
IMPACT: Choose a domain which enables you to create scalable solutions to
meaningful global problems
by Marshall Lincoln and Keyur Patel
|
Published by Data Science Central, February 2019 |
Data science - especially AI & ML - is positioned to transform the world
Artificial intelligence, big data, and the rapid emergence of the ‘data economy’ represent both a fundamental shift in the way many industries are conduct business, and a massive boost to the global economy.
According to the McKinsey Global Institute, Artificial Intelligence alone could increase global GDP by about 16 per cent by 2030. In absolute terms, this represents additional economic activity on the order of $13 trillion, annually. To put this into context, McKinsey predicts that AI will contribute to global productivity increases at twice the annual rate of the digital revolution in the 1990s and 2000s; and four times the rate of growth engendered by the introduction of the steam engine in the second half of the 19th century.
Current forecasts show that those who need AI most will benefit least
Though the benefits of the AI revolution will be felt across every country on earth, they will not be shared equally. Nor will the risks and negative externalities of this revolution be imposed upon individuals, or countries, in proportion to its benefits.
As McKinsey states, “A key challenge is that adoption of AI could widen gaps between countries, companies, and workers.”
If you’re a data science practitioner or student reading this article, you may well stand to benefit from the AI revolution. You’re making what seems like a logical decision to secure your economic future - a decision that few would begrudge you. Yet it’s an uncomfortable truth that you stand to gain from will cause others to lose their livelihoods and incomes - often those who in a position where they can least afford to.
Artificial intelligence, big data, and the rapid emergence of the ‘data economy’ represent both a fundamental shift in the way many industries are conduct business, and a massive boost to the global economy.
According to the McKinsey Global Institute, Artificial Intelligence alone could increase global GDP by about 16 per cent by 2030. In absolute terms, this represents additional economic activity on the order of $13 trillion, annually. To put this into context, McKinsey predicts that AI will contribute to global productivity increases at twice the annual rate of the digital revolution in the 1990s and 2000s; and four times the rate of growth engendered by the introduction of the steam engine in the second half of the 19th century.
Current forecasts show that those who need AI most will benefit least
Though the benefits of the AI revolution will be felt across every country on earth, they will not be shared equally. Nor will the risks and negative externalities of this revolution be imposed upon individuals, or countries, in proportion to its benefits.
As McKinsey states, “A key challenge is that adoption of AI could widen gaps between countries, companies, and workers.”
If you’re a data science practitioner or student reading this article, you may well stand to benefit from the AI revolution. You’re making what seems like a logical decision to secure your economic future - a decision that few would begrudge you. Yet it’s an uncomfortable truth that you stand to gain from will cause others to lose their livelihoods and incomes - often those who in a position where they can least afford to.
Does fintech have the right green credentials?
by Keyur Patel
|
Published by Financial World, October 2018 |
Financial institutions are being urged onto the front line of the world’s fight against climate change and ecological devastation. It is a call not only from environmental advocacy groups, but increasingly intensified by policy makers, intergovernmental organisations, and activist investors. It is also percolating through the financial industry itself – a commercial imperative as well as an ethical one. Leading insurance companies, for example, are divesting from coal-linked assets as evidence mounts that natural disasters caused by climate change are fundamentally threatening their business models.
What traditional financial services firms do is, however, only one part of the picture. A report published last year by the United Nations Environment Programme noted: “Fintech is not just another topic in the green finance space – it is a lens on the future of the financial system itself”.
In principle, greenfield financial firms, without the burden of legacy assets and set up in a lower-cost virtual environment can steer the financial sector towards financing low-carbon, sustainable economic activities. Their most important application could be in emerging economies. There fintechs can help “leapfrog over many stages of development”, says the UN. In other words, they can help power economic growth sustainably from the bottom up while simultaneously giving tens or hundreds of millions of people access to financial services for the first time.
To take one example, M-Kopa, a Kenyan solar energy company, sells off-grid home solar systems in East Africa. Its customers, who reside largely in rural areas where on-grid electrical infrastructure can be prohibitively expensive, pay a deposit for the system and a small daily fee ($0.50) over a year, at which point they own it. M-Kopa has connected well over half a million homes to solar power to date – which it says will save over 600,000 tonnes of carbon emissions over four years.
What traditional financial services firms do is, however, only one part of the picture. A report published last year by the United Nations Environment Programme noted: “Fintech is not just another topic in the green finance space – it is a lens on the future of the financial system itself”.
In principle, greenfield financial firms, without the burden of legacy assets and set up in a lower-cost virtual environment can steer the financial sector towards financing low-carbon, sustainable economic activities. Their most important application could be in emerging economies. There fintechs can help “leapfrog over many stages of development”, says the UN. In other words, they can help power economic growth sustainably from the bottom up while simultaneously giving tens or hundreds of millions of people access to financial services for the first time.
To take one example, M-Kopa, a Kenyan solar energy company, sells off-grid home solar systems in East Africa. Its customers, who reside largely in rural areas where on-grid electrical infrastructure can be prohibitively expensive, pay a deposit for the system and a small daily fee ($0.50) over a year, at which point they own it. M-Kopa has connected well over half a million homes to solar power to date – which it says will save over 600,000 tonnes of carbon emissions over four years.
Come the Revolution
by Keyur Patel
|
Published by Financial World |
The first industrial revolution saw mechanisation transform the world, machines powered by steam and water which not only replaced individual craftsmen, but also built new machines. The second saw electricity make power widely and instantly available and drove a further step in mechanisation: the production line. The third, starting in the mid-20th century, ushered in the digital age.
Now we are on the precipice of a fourth – and, for better or for worse, it will fundamentally change our livelihoods. Once the domain of futurists with feverish imaginations, it is a narrative that has become ensconced in mainstream thinking: artificial intelligence (AI) could take over a huge chunk of the work humans currently do within a generation.
Why the sudden surge in attention? AI – in all its various guises, such as machine learning, deep learning and natural language processing – is evolving much faster than most observers expected even a decade ago. Traditional thinking assumes that the jobs vulnerable to automation are low-skilled. But the breadth and complexity of activities that are now within the reach of AI will present a formidable challenge to virtually every profession: medicine, engineering, law – even art and music.
It should come as no surprise that financial services – a sector that, at least in terms of processes, if not in terms of underlying activity such as risk management – has been entirely transformed by technology over the past few decades, is considered especially ripe for disruption. According to a recent report by the consultancy PwC, just under a third of British financial and insurance jobs are “at potential high risk” from AI-driven automation by the early 2030s. That’s a higher proportion of jobs than across the economy as a whole, the study finds.
So far, these disruptive forces have been somewhat subdued. Large banks aren’t generally inclined to be early adopters of AI. They’re hamstrung by their organisational complexity, regulatory restrictions, and dependence on legacy systems that are hugely expensive to overhaul.
But such cautiousness is unlikely to persist for much longer. As the technology continues to improve, there are applications for which AI-based systems are clearly superior to human brainpower alone – for instance, the detection of fraud and money laundering. Perhaps even more pertinent are economic factors. While AI is getting cheaper to implement as it comes more widely used, banks’ payroll costs are mounting.
Now we are on the precipice of a fourth – and, for better or for worse, it will fundamentally change our livelihoods. Once the domain of futurists with feverish imaginations, it is a narrative that has become ensconced in mainstream thinking: artificial intelligence (AI) could take over a huge chunk of the work humans currently do within a generation.
Why the sudden surge in attention? AI – in all its various guises, such as machine learning, deep learning and natural language processing – is evolving much faster than most observers expected even a decade ago. Traditional thinking assumes that the jobs vulnerable to automation are low-skilled. But the breadth and complexity of activities that are now within the reach of AI will present a formidable challenge to virtually every profession: medicine, engineering, law – even art and music.
It should come as no surprise that financial services – a sector that, at least in terms of processes, if not in terms of underlying activity such as risk management – has been entirely transformed by technology over the past few decades, is considered especially ripe for disruption. According to a recent report by the consultancy PwC, just under a third of British financial and insurance jobs are “at potential high risk” from AI-driven automation by the early 2030s. That’s a higher proportion of jobs than across the economy as a whole, the study finds.
So far, these disruptive forces have been somewhat subdued. Large banks aren’t generally inclined to be early adopters of AI. They’re hamstrung by their organisational complexity, regulatory restrictions, and dependence on legacy systems that are hugely expensive to overhaul.
But such cautiousness is unlikely to persist for much longer. As the technology continues to improve, there are applications for which AI-based systems are clearly superior to human brainpower alone – for instance, the detection of fraud and money laundering. Perhaps even more pertinent are economic factors. While AI is getting cheaper to implement as it comes more widely used, banks’ payroll costs are mounting.