Article | May 20, 2020
We’re dealing with more data in the enterprise than ever before. Headlines blare that “data is valuable” but, that’s only true if the information you have is of high quality. The question becomes, how do you know if your data is high-quality? This post explores the concept of big data quality and why it is a challenge, why the enterprise needs it, and what solution you can use to ensure the quality of big data.
Article | August 13, 2020
The coronavirus outbreak in China has grown to a pandemic and is affecting the global health & social and economic dynamics. An ever increasing velocity and scale of analysis — in terms of both processing and access is required to succeed in the face of unimaginable shifts of market; health and social paradigms. The COVID-19 pandemic is accompanied by an Infodemic. With the global Novel Coronavirus pandemic filling headlines, TV news space and social media it can seem as if we are drowning in information and data about the virus. With so much data being pushed at us and shared it can be hard for the general public to know what is correct, what is useful and (unfortunately) what is dangerous. In general, levels of trust in scientists are quite high albeit with differences across countries and regions. A 2019 survey conducted across 140 countries showed that, globally, 72% of the respondents trusted scientists at “high” or “medium” levels. However, the proportion expressing “high” or “medium” levels of trust in science ranged from about 90% in Northern and Western Europe to 68% in South America and 48% in Central Africa (Rabesandratana, 2020).
In times of crisis, like the ongoing spread of COVID-19, both scientific & non-scientific data should be a trusted source for information, analysis and decision making. While global sharing and collaboration of research data has reached unprecedented levels, challenges remain. Trust in at least some of the data is relatively low, and outstanding issues include the lack of specific standards, co-ordination and interoperability, as well as data quality and interpretation. To strengthen the contribution of open science to the COVID-19 response, policy makers need to ensure adequate data governance models, interoperable standards, sustainable data sharing agreements involving public sector, private sector and civil society, incentives for researchers, sustainable infrastructures, human and institutional capabilities and mechanisms for access to data across borders.
The COVID19 data is cited critical for vaccine discovery; planning and forecasting for healthcare set up; emergency systems set up and expected to contribute to policy objectives like higher transparency and accountability, more informed policy debates, better public services, greater citizen engagement, and new business development. This is precisely why the need to have “open data” access to COVID-19 information is critical for humanity to succeed. In global emergencies like the coronavirus (COVID-19) pandemic, open science policies can remove obstacles to the free flow of research data and ideas, and thus accelerate the pace of research critical to combating the disease. UNESCO have set up open access to few data is leading a major role in this direction. Thankfully though, scientists around the world working on COVID-19 are able to work together, share data and findings and hopefully make a difference to the containment, treatment and eventually vaccines for COVID-19.
Science and technology are essential to humanity’s collective response to the COVID-19 pandemic. Yet the extent to which policymaking is shaped by scientific evidence and by technological possibilities varies across governments and societies, and can often be limited. At the same time, collaborations across science and technology communities have grown in response to the current crisis, holding promise for enhanced cooperation in the future as well.
A prominent example of this is the Coalition for Epidemic Preparedness Innovations (CEPI), launched in 2017 as a partnership between public, private, philanthropic and civil society organizations to accelerate the development of epidemic vaccines. Its ongoing work has cut the expected development time for a COVID-19 vaccine to 12–18 months, and its grants are providing quick funding for some promising early candidates. It is estimated that an investment of USD 2 billion will be needed, with resources being made available from a variety of sources (Yamey, et al., 2020).
The Open COVID Pledge was launched in April 2020 by an international coalition of scientists, lawyers, and technology companies, and calls on authors to make all intellectual property (IP) under their control available, free of charge, and without encumbrances to help end the COVID-19 pandemic, and reduce the impact of the disease. Some notable signatories include Intel, Facebook, Amazon, IBM, Sandia National Laboratories, Hewlett Packard, Microsoft, Uber, Open Knowledge Foundation, the Massachusetts Institute of Technology, and AT&T. The signatories will offer a specific non-exclusive royalty-free Open COVID license to use IP for the purpose of diagnosing, preventing and treating COVID-19.
Also illustrating the power of open science, online platforms are increasingly facilitating collaborative work of COVID-19 researchers around the world. A few examples include:
1. Research on treatments and vaccines is supported by Elixir, REACTing, CEPI and others.
2. WHO funded research and data organization.
3. London School of Hygiene and Tropical Medicine releases a dataset about the environments that have led to significant clusters of COVID-19 cases,containing more than 250 records with date, location, if the event was indoors or outdoors, and how many individuals became infected. (7/24/20)
4. The European Union Science Hub publishes a report on the concept of data-driven Mobility Functional Areas (MFAs). They demonstrate how mobile data calculated at a European regional scale can be useful for informing policies related to COVID-19 and future outbreaks. (7/16/20)
While clinical, epidemiological and laboratory data about COVID-19 is widely available, including genomic sequencing of the pathogen, a number of challenges remain:
1. All data is not sufficiently findable, accessible, interoperable and reusable (FAIR), or not yet FAIR data.
2. Sources of data tend to be dispersed, even though many pooling initiatives are under way, curation needs to be operated “on the fly”.
3. In addition, many issues arise around the interpretation of data – this can be illustrated by the widely followed epidemiological statistics. Typically, the statistics concern “confirmed cases”, “deaths” and “recoveries”. Each of these items seem to be treated differently in different countries, and are sometimes subject to methodological changes within the same country.
4. Specific standards for COVID-19 data therefore need to be established, and this is one of the priorities of the UK COVID-19 Strategy. A working group within Research Data Alliance has been set up to propose such standards at an international level.
Given the achievements and challenges of open science in the current crisis, lessons from prior experience & from SARS and MARS outbreaks globally can be drawn to assist the design of open science initiatives to address the COVID-19 crisis. The following actions can help to further strengthen open science in support of responses to the COVID-19 crisis:
1. Providing regulatory frameworks that would enable interoperability within the networks of large electronic health records providers, patient mediated exchanges, and peer-to-peer direct exchanges. Data standards need to ensure that data is findable, accessible, interoperable and reusable, including general data standards, as well as specific standards for the pandemic.
2. Working together by public actors, private actors, and civil society to develop and/or clarify a governance framework for the trusted reuse of privately-held research data toward the public interest. This framework should include governance principles, open data policies, trusted data reuse agreements, transparency requirements and safeguards, and accountability mechanisms, including ethical councils, that clearly define duties of care for data accessed in emergency contexts.
3. Securing adequate infrastructure (including data and software repositories, computational infrastructure, and digital collaboration platforms) to allow for recurrent occurrences of emergency situations. This includes a global network of certified trustworthy and interlinked repositories with compatible standards to guarantee the long-term preservation of FAIR COVID-19 data, as well as the preparedness for any future emergencies.
4. Ensuring that adequate human capital and institutional capabilities are in place to manage, create, curate and reuse research data – both in individual institutions and in institutions that act as data aggregators, whose role is real-time curation of data from different sources.
In increasingly knowledge-based societies and economies, data are a key resource. Enhanced access to publicly funded data enables research and innovation, and has far-reaching effects on resource efficiency, productivity and competitiveness, creating benefits for society at large. Yet these benefits must also be balanced against associated risks to privacy, intellectual property, national security and the public interest.
Entities such as UNESCO are helping the open science movement to progress towards establishing norms and standards that will facilitate greater, and more timely, access to scientific research across the world. Independent scientific assessments that inform the work of many United Nations bodies are indicating areas needing urgent action, and international cooperation can help with national capacities to implement them. At the same time, actively engaging with different stakeholders in countries around the dissemination of the findings of such assessments can help in building public trust in science.
Article | March 12, 2020
The book Design, Launch, and Scale IoT Services classifies the components of IoT services into technical modules. One of the most important of these is Artificial Intelligence (AI). This article is intended to supplement the book by providing insight into AI and its applications for IoT. After many years in the wilderness, AI is back on the hype curve and will change the world again. Or, will it? AI has always been interesting, but what has changed to justify the current hype?
Article | December 11, 2020
Intelligence is a much-debated term, with varying connotations to distinct disciplines. Humans have an innate intelligence that is capable of achieving complex, integrative goals through multiple faculties. These faculties involve learning and creativity, deal with ambiguity and uncertainty, critical thinking, strategy and planning, scenario analysis, and more. Humans have an evolutionary mind that is capable of drawing inferences and insights.
Creating machines, bots, or capabilities imbued with human-like intelligence has fascinated humans for a long time and has been the subject of active technical effort since John McCarthy coined the term ‘Artificial Intelligence’ (AI). Interest in AI has waxed and waned, with unrealized hype leading to a long AI winter. However, recent advances, such as Hinton’s backpropagation based deep neural networks for ImageNet that match human accuracy for image recognition, have revived hope and optimism for the advent of ‘Artificial General Intelligence’ (AGI).
AGI is about emulating or even exceeding, human levels of intelligence. At the moment, it is more of a pipe dream in the realm of sci-fi movies like Terminator. Silicon Valley leaders and scientists like Elon Musk, Bill Gates, and Stephen Hawking have predicted a dystopian, even Frankensteinian, world with recursively- improving technological singularity potentially turning against the humans.
Strong Vs. Weak AI
Weak or narrow AI is categorized as mimicking a specific human ability to perform a well-defined task. Humans seem to have become pretty good at aspects of narrow AI lately, such as natural language processing (NLP), image recognition, machine translation, and detecting fraudulent credit card transactions. In the words of Andrew Ng, any task that takes a few minutes of human cognition can be automated with supervised machine learning and the help of labeled data. Recent advances in machine and deep learning have upped the ante on weak AI. For example, DeepMind’s AlphaFold can solve the intractable problem of predicting a protein’s folding structure from its amino acid sequence, thereby circumventing years of laborious work. This goes far beyond narrow AI into the gray zone.
Strong AI, or artificial general intelligence, can solve present-day ‘AI-hard’ problems that require a complex interplay of human cognitive abilities. For example, understanding the nuances of language is hard, but humans are slowly making strides. Some human skills are multifactorial, such as driving that requires image recognition, fine motor skills, or estimation with a high degree of situational awareness. A point has been reached where a self-driving car with level five autonomy can emulate that with simultaneous localization and mapping (SLAM) while being vulnerable to getting tricked at the same time.
Leading voices have articulated several benchmarks for having accomplished AGI, such as:
Turing test: If a human and machine are indistinguishable most of the time while conversing with another human. With OpenAI’s GPT-n series, that is probably not far away.
A bot or computational system successfully passes grad school.
An AGI bot becomes a productive member of society, possibly paying taxes while performing a complex job.
Emulating the Human Brain
Unraveling the human brain is as enigmatic as solving the mysteries of the cosmos. With approximately 100 billion neurons interconnected through a quadrillion synapses, leading to 100 trillion synaptic updates per second (SUPS), the human brain is inordinately complex to simulate. Other than the interconnectedness of the brain, its evolutionary neurophysiology at the molecular and cellular level requires a level of chemical, physical, and biological understanding that leaves one confounded. How the three-pound mass of mostly fat, protein and water, with neurons firing in a chemical soup, allows cognitive abilities is quite hard to fathom.
All the advances in artificial neural networks, IoT sensing, 5G bandwidth, real-time big data, GPUs or TPUs, and storage put together get nowhere close to creating a computational system that has characteristics of sentience, self-awareness, sapience, and consciousness. Some even argue that there can be no human-like intelligence and consciousness without the accompanying embodiment.
Challenging as that may be, the advances in narrow AI are quickly adding up, with a bottom-up approach, to an impressive array of well-defined and compartmentalized human abilities. While AGI is the holy grail, the key point is that such pursuits are enabling scientific and technological advances that are the sweet spot of enabling human-in-the-loop technologies that augment humans instead of replacing them. Progress will likely stay in the augmentation zone for the next couple of decades, as Ray Kurzweil’s prediction of AGI comes true by 2045. Others argue that humans may not accomplish AGI in this century at all. But there is little disagreement over the fact that AI is likely to create US$15 trillion of economic value by 2030, with US$6 trillion being attributed to deeplearning alone. Individuals, societies, and businesses have to brace for that impact.
How Can Businesses Prepare and Respond to General AI
China is leading the AI frontier, as much as due to its lack of regulatory and ethical oversight as to its dogged commitment to winning the AI supremacy race. The US is not far behind whereas other nations occupy different positions on the leaderboard. Expertise in AI is likely to shake up the global economic and geopolitical order in the future world. While individuals grapple with the widespread displacement of world labor markets, enterprises need to sense and respond as well to ensure they thrive in a world replete with AI.
Here are some steps they can take to ensure they are not sidelined in a world of sustained disruption and mere transient advantages:
#1 Create a vision of yourself in the future world of AGI. Make small bets to preserve strategic options in aspects of your business potentially exposed to general AI.
#2 Make big, bold moves on narrow AI for quick wins. This will instill confidence and purpose to respond to general AI as it comes of age. Embrace AI augmentation as opposed to resisting it.
#3 Put your digital maturity on the front burner and prioritize digital transformation initiatives. Be a digital leader, not a laggard.
#4 Data maturity is a precursor to digital maturity. Invest in advantaged data with internal data or external data from partnerships, acquisitions, or ecosystem orchestration. AI is contingent upon data and algorithmic advances.
#5 Democratize technology by expanding it beyond the traditional IT organization of the company.
#6 Embrace a digital culture with rapid test-and-learn abilities. Don’t ostracize failure as long as you pivot fast and fail cheaply.
#7 Institutionalize innovation incubation. Also, explore open innovation models by partnering with other businesses and institutions.
#8 Orchestrate between exploitation and exploration strategies – the former for the here and now and the latter for the future.
#9 Deploy a forward-thinking governance framework that can orchestrate across near, mid-, and long-term growth.
#10 Deploy your workforce in fluid, agile, self-organizing teams that can ‘flow to the work’.