Investing in Artificial Intelligence

Transcription

1 Published: January 2018 Investing in Artificial Intelligence An overview of the rapidly evolving AI landscape Tuhin Ghosh, PhD, CFA Chief Investment Officer Raymond Mok, CFA, CPA Portfolio Manager

2 Introduction From robots that make espressos to algorithms that match you with an Uber driver, the world is fast becoming automated. As we go from robots that assemble cars to cars that drive themselves, we re increasingly tasking smart computer algorithms with critical decision-making. At the heart of all this is the technology known as artificial intelligence. AI as it s commonly known involves engineering intelligent computer programs to complete tasks and goals without human intervention. It s a topic that s grabbing headlines these days. The idea that we can emulate human cognitive capabilities through intelligent software to solve problems both mundane and high-stakes is fueling people s imagination about what s possible with AI. Indeed, AI has the potential to deliver far-reaching benefits that go beyond the economic calculus as it can be applied to solve the world s most challenging problems, from saving lives to building greener and smarter cities all at a massive scale. What does it mean to invest in AI? Investing in AI gives you exposure to the market for software, hardware, and services that deliver sophisticated cognitive capabilities to solve problems for people and businesses. AI today is less about automating manual work and more about automating cognition and bringing intelligence to systems and processes. 2

3 Artificial intelligence at a glance Just what is AI? If you ve ever clicked on a product recommendation on Amazon, searched for something on Google, or asked Siri to find you the nearest gas station, you ve interacted with AI. AI is a vast field made up of many specialized technologies that draw from diverse fields like computer science, mathematics, neuroscience, and linguistics. It can be loosely defined as systems that can work, perform tasks, and achieve goals without human intervention. This means AI is not strictly drones and robots but rather the underlying ecosystem that powers these systems. Many of today's successful AI deployments are powered by a subfield of AI known as machine learning. This technology involves feeding the system a learning algorithm, exposing it to massive amounts of data to train it so it can eventually figure out on its own how to predict, interpret, and act on the data. For instance, you might feed an image recognition system terabytes worth of photos, to recognize the faces in those images. It s essentially software writing software, in the words of Nvidia CEO Jen-Hsun Huang. From language translation to image recognition, the ability of machines to interpret data is inching towards human-like accuracy, giving rise to more mainstream implementations. AI applications such as speech recognition used by virtual assistants like Amazon s Alexa and Netflix s recommendation engines are made possible by a branch of machine learning called deep learning. The Future of AI Let s look under the hood of self-driving cars. These vehicles are enabled by a powerful ecosystem that perceives, reasons, navigates, and monitors, culminating in a driver-free experience. Perception Reasoning Driving High Definition Mapping Object Localization AI Computing Source: NVIDIA Investor Presentation May

4 Artificial intelligence is being driven by the explosion of data AI was built on decades of research that started in academia and other non-commercial research institutions. But the advances in the field have accelerated in the last few years, giving way to a vibrant AI ecosystem amid a convergence of the following key developments: The advent of big data. The volume of data generated by ever-growing number of sources machines, devices, software, online services has grown exponentially. Data is the raw material required to train and improve machine learning algorithms. Advances in computational power. Massive computational power allows machines to do everything from recognizing objects to translating speech in real time. Powerful graphics processing units (GPUs) are key here because they enable complex computations in learning algorithms, making AI applications scalable. Investments made in AI. Over the years investments made by tech giants are starting to bear fruit. These include improvements in the design of learning systems, system architectures and software infrastructure. Cost reduction. Dramatic declines in data storage costs from $569/GB in 1992 to $0.03/GB in

5 Investment Implications Companies that enable, supply, and build data networks and infrastructure stand to benefit in this landscape. These include providers of high-performance computing semiconductor chips, connectivity and network components, analytics solutions, and sensory devices. Today, the sheer volume and variety of data that is generated by humans, software and machines far surpass our ability to process and consume them. In this Internet of Things and Everything as a Service era, we have more data-generating sources than ever. To put this into perspective, 90% of the data in the world today was created just in the last two years. 2 An additional AI opportunity lies in the fact that only 9% of the world s data is structured or tagged. 3 Machine learning and deep learning will be essential to processing and making sense of this raw, unstructured data. Big data is fundamentally changing how companies operate and compete; those who determine the best ways to get value from this data are poised to win. Global Data Volume and % Structured Data is Becoming More Structured Source: KCPB Internet Trends Report 2017 It will soon become impossible for conventional engineering solutions to handle the increasing amounts of available data. Machine learning offers the ability to extract certain knowledge and patterns from a series of observations. -Gartner 4 5

6 Democratizing AI through AI-as-a-Service AI is largely still in its early days despite the research efforts of the last 60 years. Some AI technologies, however, are hitting their stride and being applied at a greater scale than ever. In fact, AI drives the businesses of the world s five most valuable companies: Apple, Alphabet, Microsoft, Amazon and Facebook. Facebook and Google have built billion-dollar businesses by getting people to click the ads they serve on their respective platforms. Behind this magic are algorithms that predict at dizzying speed and scale which ads to serve each person who visits their platform. Now, these behemoths want to further monetize their algorithms by offering them to other businesses at scale, typically delivered via application programming interfaces (APIs). AI-as-a-service (AI-aaS) landscape Source: Goldman Sachs Artificial Intelligence Report Nov 14, 2016 Among AI s latest buzzword is democratization. The biggest AI players are all in it to democratize AI by making their artificial intelligence and machine learning technology available to anyone who wants to add AI capabilities to their technology stack. If businesses can directly access AI capabilities, they can leverage smart algorithms to transform their business processes and customer experiences. This means no longer will only the biggest companies have access to the large-scale data and computing resources that are required to train and create sophisticated AI algorithms. Smaller companies can access these services too, leveling the technological playing field for all. Investment Implications There is an expanding array of AI applications delivered through AI-as-a-Service. Tech giants like Amazon, Google, IBM and Microsoft currently dominate the fast-growing machine learning cloud services market (see next page). As mentioned above, these services can take the form of APIs that allow users to quickly access, leverage, and scale AI technologies developed by the tech behemoths. 6

7 Companies can tap machine learning APIs, for instance, that enable speech recognition for their chatbot application. In this case, Microsoft might offer an API for its Azure Cloud offering to quickly connect to its remote data centers and utilize capabilities such as machine learning and speech recognition. In short, access to AI products and services is expected to add valuable insights and capabilities to drive strategy and execution across every operational domain, from relationship management to marketing and sales to supply chain management. As AI adoption increases, companies offering AI services and platforms should stand to reap the rewards. 7

8 The case for the Artificial Intelligence theme Both businesses and governments yearn to master AI, spelling an inevitability for AI to become a dominant technology this century. In 2015, the U.S. government invested $1.1 billion in unclassified R&D in AI-related technologies, and the 2016 spend has been estimated at $1.2 billion. 5 Meanwhile, China released a national plan in 2017 to position itself as the de facto AI leader by On the commercial front, tech titans have doubled down on AI through a variety of vehicles over the past few years as they seek to dominate this space. These activities include acquisitions, internal investments and establishing venture funds that invest solely in AI startups. As far as market forecasts, current estimates for AI's revenue potential and growth rates are bullish by any measure. IDC expects the global AI market to reach $57.6 billion by 2021, compared to $8.0 billion and $12.0 billion in 2016 and 2017, respectively. This represents a compound annual growth rate (CAGR) of 50% over the forecast period. 7 These projections underline a profound technological trend that has found firm footing as AI moves decisively into mainstream commercialization. AI in numbers: Value of global AI market in 2021: $57.6 billion Source: IDC Annual Projected AI Market Growth, % Source: IDC Incremental contribution to U.S. GDP by deploying AI in 2035E: $8 trillion Source: Accenture Journal articles mentioning "deep learning" or "deep neural network" by country of publication Labor productivity gains resulting from AI deployment: +40% Source: Accenture Source: Pitchbook US venture capital investing in AI has inflected this decade (capital raised in $mm) Source: US National Science and Technology Council 8

9 While it s still early days for AI, there s no denying that some businesses are already reaping the benefits of AI while others are trying to learn and experiment with the technology (see Appendix 2). These dynamics are poised to accelerate AI s adoption: AI is past the tipping point. Businesses must embrace and learn AI or risk getting left behind. There s increasing recognition that business transformations are made through software, especially intelligent software. The race is on to make the physical world digitalized and outcomes optimized autonomously. Beyond the tipping point AI is already a key underlying capability that powers many key consumer and enterprise applications. From here, it is poised to become even more pervasive, ushering in a new computing paradigm that includes digitizing the physical world and training machines to analyze, interpret, and predict events faster and better than humans ever could. Transformative and disruptive The modern information age has only just begun and AI promises to accelerate its evolution. Remember when 10 years ago the tools to create and publish online content became highly accessible and cheap? That gave way to the capabilities and catchphrase: every company is a media company. In the last few years, the mantra has become: every company is a software company. This means businesses need to drive their business models and competitive advantages through software platforms. In the words of Microsoft CEO Satya Nadella: What you want to achieve in business is fundamentally changing technology. build systems of intelligence, built on systems of record to create a feedback loop, driving performance and efficiency." 8 If you believe that businesses are in fact becoming like software companies, then AI will be front and center of this transformation. Artificial intelligence technologies will be the most disruptive class of technologies over the next 10 years due to radical computational power, near-endless amounts of data, and unprecedented advances in deep neural networks; these will enable organizations with AI technologies to harness data in order to adapt to new situations and solve problems that no one has ever encountered previously. -Gartner 9

10 AI in a digitized, smarter world AI is the next logical frontier in the current information age that seeks to make everything smarter by digitizing the physical world. Businesses will treat their assets from assembly plants to customers as data to be processed, analyzed, and acted on to generate certain outcomes. We see the potential for AI and machine learning to reshuffle the competitive order across every industry. -Goldman Sachs Your drive to transform businesses is setting the technology agenda. What you want to achieve in business is fundamentally changing technology. -Microsoft CEO Satya Nadella Early evidence suggests that AI can deliver real value to serious adopters and can be a powerful force for disruption Companies cannot delay advancing their digital journeys, including AI. -McKinsey and Co. In the next 10 years, we will shift to a world that is AI-first. -Google CEO Sundar Pichai 10

11 Risks to the Artificial Intelligence Theme AI in numbers: Political risk: Concerns around the displacement of human labor, and the supply and demand gap between skilled and non-skilled workers may engender policy responses that dampen further investment in automation. Security and data risk: Security threats and regulations focused on data privacy and the emergence of open source solutions could hamper the growth of data and AI. Investment risk: The impact of artificial intelligence technology remains restricted to existing applications such as online search and recommendation engines, forcing companies to write down expensive R&D expenditure. Implementation risk: This can entail risk related to poor execution of AI strategies and scarcity of talent (e.g., data scientists, engineers and statisticians). A successful execution requires a transformation along many dimensions from creating a data ecosystem that enables the training, choosing the right AI tools, and adapting outdated processes and culture. 40% Percent of the world s top 20 companies that are classified as technology up from 20% in 2012 Source: KCPB Internet Trends Report 2017 $13.7 billion Amount of capital raised by AI & machine learning startups in 2016, up from $600 mm in 2005 Source: Goldman Sachs, Pitchbook 2.5 quintillion Amount of bytes of data created everyday. 90% of the world s data has been created in past 2 years Source: IBM 11

12 Investing in companies focused on artificial intelligence Artificial intelligence industry High barriers to entry with wide, technology-driven moats. Because AI technologies are highly technical and sophisticated, they require years to develop and need to be backed by significant investments in talent and resources. Intellectual property will often be protected by patents. All these factors will limit copycats and provide an edge to incumbents over new entrants. Network effect will drive additional efficiencies. Cash-rich large technology companies will be able to attract and retain key AI talent as well as acquire startups. The accumulation of a variety of AI technologies, user base, and customers will allow these companies to reinvest and redeploy resources to lower costs and develop future technologies. Look for healthy industry dynamics to persist. The market is dominated by tech companies with billions of dollars in revenues, healthy profit margins, and large cash reserves equating to a high margin of error for risk-taking. Investing in companies delivering artificial intelligence We are in the early days of a long-trend as the potential of AI is just being realized. This means companies that can successfully develop and deliver essential AI capabilities that unlock economic value are poised to benefit. These are the AI segments that we have identified as relevant to the AI investment theme and stand to benefit from further advancements in this field. Intelligent Software Artificial intelligence solutions such as analytics and customer relationship management AI Algorithms Machine learning algorithms that provide recommendations based on users likes and preferences AI Chips Semiconductors and processors used in artificial intelligence Intelligence Consulting Consulting services and intelligence platforms for automation, behavior informatics, etc. Autonomous Machines Companies providing solutions for connected cars and location-based platform services such as advanced driver assistance systems AI Infrastructure Network and connectivity for artificial intelligence solutions Sensors and Perception Software and sensors for voice and pattern recognition Consumer Products Consumer-focused products that utilize automation, learning solutions, etc. Healthcare Solutions Companies in the healthcare domain using machine learning and artificial intelligence for diagnosis 12

13 Motif Capital: Portfolio Construction Motif s thematic investment philosophy is underpinned by technology, helping us process and derive investment insights in an objective, systematic manner. The portfolio construction process starts with theme discovery and theme validation through rigorous fundamental and quantitative research. Motif then employs the same systematic, data-driven approach to create the theme universe and identify thematic exposure. Our starting point is to use publicly available data such as government reports, industry publications, and company filings. We also utilize alternative data sets, which have traditionally not been used by investors, and often come raw and unstructured. We leverage technology and a variety of computing techniques to extract valuable information from these data sources. Because artificial intelligence will continue to evolve, it is challenging to create a portfolio which expresses our interpretation of the theme and the key beneficiaries. Through a combination of structured and unstructured data sources, and conventional and unconventional investment analysis techniques, we can achieve our goal. Our Artificial Intelligence portfolio was built to provide systematic and targeted exposure to the firms which participate in the various end markets and verticals listed in the aforementioned AI segments. These companies are crucial to the AI ecosystem and will benefit as the theme plays out. Our approach to constructing this portfolio includes these key steps: Identify US-listed stocks and ADRs of companies which develop or provide AI-related products and services Segment companies by their focus within the AI technology ecosystem Determine each company s exposure to the AI theme Representative Companies: Alphabet Inc (Google) Google offers online services such as search engines that leverage AI algorithms to optimize search results and ad serving. It is also a leader in autonomous vehicle research. Baidu Inc Often called the Google of China, Baidu offers similar web services as Alphabet. It has invested heavily in AI, and offers an AI suite called Baidu Brain. Intel Corporation Intel is a chip maker with applications that help power data centers and smart "internet of things" devices. Nuance Communications Inc Nuance is a technology company which provides products and services related to voice recognition and natural language processing, which enables crucial sensory capabilities for AI machines. NVIDIA Corporation NVIDIA is a GPU manufacturer whose products are used for high performance computing, enabling advanced AI techniques such as machine learning and parallel processing. Weight each company by its market capitalization adjusted for exposure to the Artificial Intelligence theme 13

14 Appendix 1 AI in everyday life Here are some examples of AI technologies and how they re being used today: AI is already driving our daily lives in transformative ways. From the morning when you wake up and say into your phone Ok, Google to fetch your , you ve already touched AI. Ok, Google is using speech recognition, and your Gmail inbox is showing s only from the people you know because machine learning was used to detect and block spam. Gmail has also sorted your mail and put them into categories like Primary (from people you know) and Promotions (from vendors who offer you deals and discounts). Still in bed, you open your Facebook app. Your news feed shows you only those posts that Facebook thinks will be interesting to you (machine learning). And when you upload to Facebook the photo of the late-night snack you shared with your spouse, you see his or her face light up as someone to tag (that s facial recognition at work). You now figure out how to navigate your commute by opening Google Maps and see that avoiding the freeway takes the shortest time. Google is predicting this using AI. You haven t even started your work day but you ve already put AI to work. AI technology What it does Applications Machine learning (ML) Rather than programming specific instructions and rules to reach a goal, the machine is trained using large amounts of data and algorithms that give it the ability to learn how to perform the task. Recommendation engines that predict which products customers might like. You ve seen this on sites like Netflix, Spotify and Amazon. Neural networks Information processing architecture based on the structure of a human brain, consisting of nodes that can learn and solve problems. Can model nonlinear relationships between inputs and outputs. Pattern recognition tasks and projections, such as forecasting weather and predicting stock prices. Deep learning A specific type of machine learning utilizing multilayer neural networks, with each successive layer using the output from the previous layer as input. Identify patterns in unstructured data, e.g. detect fraud, screen for cancer. GPU-accelerated computing This type of computing involves a graphics processing unit (GPU) with multiple computing cores to support parallel computing. Enables and accelerates machine learning, deep learning and advanced analytics. Natural language processing (NLP) This involves processing text to analyze and derive meaning from human language. Translation, topic analyses, detection of trending sentiments. Speech recognition Receives and responds to human speech. Speech-enabled interfaces like the Siri personal assistant. Machine vision Inspecting and analyzing images and videos in an automated way. Monitoring systems in areas like security and healthcare to detect unusual events and conditions so preventative measures can be taken. 14

15 Appendix 2 AI in action The AI space is breeding companies that design AI systems for specific industries. Within the insurance industry, a slew of startups are building systems for everything from smart claims management to actuarial management to fraud detection. In healthcare, the AI providers span the gamut from applications that predict hospital readmissions to using machine vision to generate better views of human tissues. Notably, McKinsey and Co. looked at AI deployments in retail, electric utilities, manufacturing, healthcare, and education and found that AI can improve forecasting and sourcing, optimize and automate operations, develop targeted marketing and pricing, and enhance the user experience. 9 Here are a few examples of AI learnings and deployments. Drug discovery Pharma giant Pfizer is using IBM s Watson to support its research on using the body s immune system to help fight cancer. Pfizer researchers use Watson to analyze massive volumes of disparate data sources, test hypotheses and uncover insights for realtime interaction and safety assessment. Investing In recent years, a few hedge funds have adopted artificial intelligence to help them spot trends and trading opportunities in the securities markets. A handful of investment firms are relying exclusively on AI to maneuver the stock market; they include San Francisco-based Sentient, Rebellion Research in New York, and Aidyia in Hong Kong. Supply chain optimization Global food supplier Danone uses a system made by ToolsGroup that uses machine learning to analyze demand variability and improve demand planning. Danone improved performance in front-end forecasting and back-end production planning, including a 30 percent reduction in lost sales and spoilage. 10 To drill or not to drill? Engineers from Halliburton use Microsoft machine learning to drive efficiencies for the energy company s exploration and production operations. Applications include applying deep learning to understand oil reservoir characteristics, modeling and simulation, and applying voice and image recognition and virtual reality to create digital representations of physical assets. 11 Better patient outcomes At Walter Reed Medical Center, the Department of Veteran Affairs is using AI to predict medical complications and improve treatment of severe combat wounds, leading to better patient outcomes, faster healing, and lower costs. Smart public transit Cities are now starting to implement responsive dispatching and routing used by ride-sharing services, and linking it with public transit scheduling and tracking software for timely transport services that are faster and cheaper. AI in financial services FICO, a U.S. financial services company that scores consumer credit worthiness known as FICO scores, relies on machine learning for operational tasks such as: compiling FICO scores predicting fraud personalized marketing offers creating services that help lenders make credit decisions. 12 Recommendation engines Recommendation engines use algorithms that learn from data. The engine makes recommendations for each consumer based on his or her behavior and preferences then predict what people might like. These engines are widely used by the likes of Netflix, Amazon, YouTube or wherever you see You might like Preventing cyberattacks before they happen Cybersecurity companies such 15

16 Appendix 2 as Palantir and Proofpoint detect cyber security risks by mining large amounts of data and making predictions using AI techniques like machine learning. Palantir has been hired by the U.S. government to ward off cyber hackers. Smart logistics Logistics and transport companies like FedEx are using AI in some capacity, and all are looking for ways to increase operational efficiencies and improve the customer experience, according to FedEx lead data scientist Clayton Clouse. 13 Every aircraft now has thousands of sensors, each of which generates readings about once per second, says Carnegie Mellon computer scientist Jaime Carbonell who oversees the Aerospace Data Analytics Lab sponsored by Boeing. That s creating vast quantities of data that overwhelms any analytical strategy we now have. So we re developing scalable algorithms that process it in the cloud. Everything from safety and predictive maintenance to in-flight performance characteristics and the aging of the aircraft can then be better tracked and understood

17 Appendix 3 Focus areas of AI with most investments Rank Enterprise Insights Enterprise Insights Enterprise Insights Enterprise Insights Enterprise Insights 2 Retail Product Recommendations NLP / Sentiment Analysis Retail Product Recommendations Virtual Assistant Solutions for Financial Institutions 3 NLP / Sentiment Analysis Retail Product Recommendations NLP / Sentiment Analysis Retail Product Recommendations Retail Product Recommendations 4 Virtual Assistant Virtual Assistant Virtual Assistant Solutions for Financial Institutions Virtual Assistant 5 Drug Discovery & Marketing Analytics Marketing Analytics Marketing Analytics Marketing Analytics Medical Diagnostics 6 Computer Vision Computer Vision Drug Discovery & Medical Diagnostics Biometrics & Facial Recognition Security Medical Dictation 7 Drug Discovery & Medical Diagnostics Solutions for Financial Institutions Solutions for Financial Institutions NLP / Sentiment Analysis AR / VR 8 Medical Dictation Medical Dictation AR / VR Computer Vision Drug Discovery & Medical Diagnostics 9 AR / VR AR / VR Computer Vision Robotics & Autonomous Systems Biometrics & Facial Recognition Security 10 Biometrics & Facial Recognition Security Cybersecurity Cybersecurity AR / VR Computer Vision Source: (Quid) 17

18 MOTIFCAPITAL.COM References 1 Deloitte University Press, "From Exponential Technologies to Exponential Innovation," IBM, "What Is Big Data?" November KPCB Internet Trends Report Gartner, "Top Trends in the Gartner Hype Cycle for Emerging Technologies, 2017," August 15, Executive Office of the President National Science and Technology Council Committee on Technology, "Preparing for the Future of Artificial Intelligence," October Reuters, "China Aims to Become World Leader in AI, Challenges U.S. Dominance," July 20, International Data Corporation 8 Computer Weekly, "Every Business will be a software business," March McKinsey Global Institute, "Artificial Intelligence: The Next Frontier?" June ToolsGroup, Danone case study Journal of Petroleum Technology, "Halliburton-Microsoft Strategic Alliance Hopes To Speed Industry Transformation," September 11, FICO, "Artificial Intelligence: Find it Right in Your Own Backyard." 13 AI Business, "The AI Trust Fall: Understanding The Human / AI Relationship With FedEx," September Wired, "AI Wields the Power to Make Flying Safer - and Maybe Even Pleasant," March Disclosures Investing in securities involves risk to be considered prior to making investment decision. This report is made available for informational purposes only and should not be considered personalized investment advice for any particular investment or investment strategy. Neither Motif Capital Management, nor Motif Investing, Inc., assess the suitability or the potential value of any particular investment. You are responsible for understanding the risks involved with investing in securities and for all investment decisions you make. Investments in developing industries and focused sectors and segments are subject to additional risks and increased volatility which an investor should consider prior to making an investment decision. Examples provided are for illustrative purposes only and not intended to represent results an investor can expect to achieve. Past performance is no guarantee of future results. Data contained herein from third-party providers is obtained from what are considered reliable sources. However, accuracy, completeness or reliability cannot be guaranteed. Any assumptions, views, and opinions expressed are based on data collected from those sources and should not be implied to be a fact pattern that will be carried out as described. Several economic, political, and economic activities, unforeseen at the time of the publication, can have a dramatic impact on the prices and performance of stocks in the marketplace. Motif Capital Management, Inc., is an SEC registered investment advisory firm, located in San Mateo, CA., and is a wholly-owned subsidiary of Motif Investing Inc., the parent company, which is a registered broker-dealer (member SIPC) Motif Capital Management, Inc. All rights reserved 18