“There’s no such things as survival of the fittest. Survival of the most adequate, maybe. It doesn’t matter whether a solution’s optimal. All that matters is whether it beats the alternative.”
Let them eat cake
One of the biggest concerns for the near future is the wave of progress dubbed “the 4th industrial revolution”. Unlike demographic and GDP shifts, the 4IR does not appear in statistical projections. The narrative spins from many different sources – economists trying to understand technology, technologists trying to understand human resource management, and computer scientists relentlessly pursuing improvement at all costs. The general impression is that of a change as vast and sweeping as the original Industrial Revolution was, potentially with even greater impacts.
As with the first Industrial Revolution, the key concern appears to be automation: human beings being replaced at their jobs by machines that can do them better, faster and cheaper. As Andrew Ng, co-founder of Google Brain, Coursera and deeplearning.ai, says: “Much has been written about AI’s potential to reflect both the best and the worst of humanity. For example, we have seen AI providing conversation and comfort to the lonely; we have also seen AI engaging in racial discrimination. Yet the biggest harm that AI is likely to do to individuals in the short term is job displacement, as the amount of work we can automate with AI is vastly bigger than before. As leaders, it is incumbent on all of us to make sure we are building a world in which every individual has an opportunity to thrive. Understanding what AI can do and how it fits into your strategy is the beginning, not the end, of that process.”
Much of the thrust is from the explosion of cheaply implementable, effective machine learning (often lumped under the umbrella term “AI”) that began with AlexNet in 2012. In some highly limited use cases, the AI in question has demonstrated learning aptitude and skill far in excess of the best humans in the field, as with DeepMind’s AlphaGo, which beat experts of a roughly 3,000 year-old-game. The newest iterations achieve these levels of expertise simply by training against itself. In other cases, the sheer breadth of data available to the AI enables it to perform tasks in ways that humans cannot.
The second facet of change comes from the recent explosion in technology that can use these advances – from cheap processing power to advances in integration into robotics. The nature of this progress is uneven; however, the code for much of this work is available free and is open source, reducing cost and speed of adoption and opening up doors to automation across multiple sectors.
Cecille Soria, one of Philippine’s foremost experts in data law, highlights the third aspect of change – three signals that commonly appear in the APAC: increasing amounts of non-permanent work, such as in business process outsourcing, where the primary concern is to hire cheap and keep costs low; “future of work” initiatives that slowly crawl towards automation initiatives; and a narrative, or perception around the gig economy that touts small units of labor, with no permanent contract, as being “cool”. “Workers like the idea that they “own their time”“, she notes. “But as this workforce gets older, they will start to realize that the safety nets of a retirement pay and medical benefits are not there for them.” These lay the groundwork for automation-ready industries.
David Galipeau, founder of SDGx and director of the Yunus Center Near Future Lab, points out the logical next step: as the three facets merge, as labor costs go up and the cost of robotics comes down, a break-even point is reached. “This,” he notes, “will impact most low-cost economies in Asia quickly. Not only labor costs are important in this equation, the stigma of slave labor has also come at a very high cost. Therefore, as more and more NGOs are outing ‘slave labor’ practices by western companies (fisheries, fruit and other foods, retail manufacturing and clothes/shoes/etc.), there will be a push to simply get rid of humans. Robots can work 24 hrs a day and very little marginal costs. They can be transferred across borders freely and work in inhumane conditions. This would solve two main issues for western producers but will create massive economic migration issues in Asia Pacific.”
The most significant analysis on downsides of this phenomenon comes from the collaboration between the UNDP and the Economist Intelligence Unit. Many of the APAC’s economies (Hong Kong, Singapore, South Korea, Taiwan, China, Vietnam, Malaysia, Bangladesh, and Thailand) rely on export-oriented manufacturing and BPO industries built around the low cost of labour. Automation will be a severe blow to this export-oriented model. This in turn increases job insecurity, reduces the tax base available to governments, and cuts out much of the wind from under the sails of the APAC. Asian countries, which have thus far been insulated from this effect, may soon experience it in full force. Major regional businesses such as Alibaba are actively throwing their weight behind the use of robots and AI for logistics). Unless Asian governments step in, millions in Asia’s middle class could slide back into poverty. [Core: THE NEXT BIG ECONOMY].
This, coupled with expanding populations [Core: FEEDING THE BEAST] and the explosion of wealth coming into the region, drives inequality even higher, creating massive social unrest. At particular risk are Vietnam and Thailand, where manufactured goods are more than 75% of merchandise exports; Philippines, where AI could potentially replace 50% of the BPO workforce; and India, which again has relied on the low cost of labour.
Data as capital
Underlying all these phenomenal technologies is data – collected, accumulated, and now processed at scales greater than ever before. Some have compared data to oil, but that analogy rings false- oil is nonrenewable, exists in fixed quantities, and is not traded back and forth by every stratum of society – whereas capital is renewable, does not exist in fixed quantities, and is constantly exchanged between almost all agents of society.
As with capital, those who amass significant amounts of data amass social, political and executive power, which, powered by technologies that arise in this new age, they then wield in what ways they wish. Sometimes the underlying goals are benevolent – for example, Aadhar, the Indian identity system running under the auspices of the Ministry of Electronics and Information Technology. Under the age-old logic of collecting data in the name of more efficiently managing a welfare state, everything from voting to pensions and money transfers are linked to a citizen’s fingerprints, iris and face photograph. A powerful push from the Indian government has turned Aadhar into possibly the largest biometric database in the world, with over 1.23 billion people in the system as of the time of writing. Given that presently, most governments in the region are reliant on costly and slow census data, integrated data ecosystems and processing technologies like this can enormously empower efficient welfare states.
However, this empowerment of the state brings with it the threat of sophisticated surveillance states and repression on a scale hitherto unseen, threatening both individualistic thinking and democracy in the name of state-nurtured communalism. China, with its social scoring and outsized economic impact in the APAC region, may very well be a science-fictional harbinger of things to come. It appears that India may follow: the new data protection policies emerging from India declare that the Indian government now has the right to access a citizen’s personal communications, whether stored on an Indian platform or elsewhere. It appears that states that can collect vast amounts of this data-capital, while managing themselves efficiently, manage to put a power differential so high between the governing and the governed that citizenry can be kept suppressed.
Attitudes to this state ownership differ among the APAC. Peggy Liu of JUCCCE highlights that the concept of privacy is very different in China, where the Western liberal belief of a right to privacy is nonexistent, and therefore the fears of surveillance statism don’t necessarily exist. Cecille Soria points out that this is changing in the Philippines, where the number of complaints received by the National Privacy Commission for 2019, up to May, equaled the entire volume of complaints received for the year of 2018. In India, the civil rights outcry over Aadhar has been enormous.
Even more pertinent, and outside the paradigm of the 20th century, are non-state, multinational actors that possess this power. McKinsey points out that “cross-border data bandwidth grew by 148 times between 2005 and 2017, to more than 700 terabytes per second—a larger quantity per second than the quantity contained in the entire US Library of Congress—and is projected to grow by another nine times in the next five years as digital flows of commerce, information, searches, video, communication, and intracompany traffic continue to surge”, while Facebook reports some 2.3 billion users on its platform. Such companies as this function – Facebook, Amazon, Google and the like – more like empires than factories, and the cumulative power they wield seem to have elevated them to the status of international monopolies, opinion-makers and threats to existing, democratic governance structures.
Move fast and break things?
This startling narrative opens up the question – should this be just because we can? Our present economic order breeds inequality, both at social and corporate levels [Core: THE NEXT BIG ECONOMY]. The beneficiaries of the 4th Industrial Revolution will likely be those with significant capital or education, creating winner-takes-all economies facing “premature industrialization” as has happened in Latin America, backed by repressive surveillance that rolls back progress on human rights.
In previous Industrial Revolutions, there was the temporary phenomena of Luddites – a group that would go around smashing machines in the face of overwhelming loss of jobs and workers unable to reskill themselves into new careers. We expect a similar phase. The disruption brought about by these changes is huge, and the move-fast-and-break things approach may forget that people, too, can break, especially at the bottom of the pyramid. Governments may attempt to regulate these new Pandora’s Boxes, but may also fall prey to competition: from a game-theoretic perspective, one bad actor (if by bad we mean hypercapitalist systems that allow for these disruptions in the name of improved economic performance) forces everyone else to act the same.
The celebrated technologist Kai-Fu Lee former founding director of Microsoft Research China and former President of Google China, points out that China and the USA, spurred by private and public sector interests, are locked in competition on the AI front in precisely this manner. At a recent panel at RightsCon 2019, concern was expressed that the world would de-evolve into something similar to the “warring states” period of China – a period of constant, intense conflict and co-dependence marked by rapidly shifting alliances, both between state and nonstate actors. David Galipeau, founder of SDGx and director of the Yunus Center Near Future Lab, provides a slightly different narrative: “China is way way way behind on AI development than the US, France and Israel. They may produce the highest number of AI patents but the US (NASA, DARPA, CMU and MIT specifically) has practically all the important AI patents of the last 60 years. Where China will excel is in AI convergence with other technologies through viable applications. Remember that Chinese companies (Alibaba, Baidu, Tencent, etc.) and the academic research institutes in China are heavily subsidized and protected from any competition so there is no shortage of money to commercialize AI apps (i.e. the cost of the China Social Credit Scoring platform is huge with no profit motivation at all). There is no real regulation in China, so applications will be seen on the market very quickly.” Both seem to agree that the lack of intellectual property controls allow Chinese corporations to first copy, then drive domestic innovation through ferocious competition and the strength of a vast corps of highly skilled engineering graduates.
Can we integrate?
In the middle of dystopia emerges a new framing: instead of AI-versus-humans, AI and humans, integrated. There already exist fields where AI outperforms humans, and one of them is chess. In 1997, Gary Kasparov, the legendary chessmaster, was beaten by IBM’s Deep Blue – the first truly public defeat of a human grandmaster. It was a shocking as historic moment, and is credited with starting the Big Data revolution. Kasparov’s response to defeat was unusual: instead of blaming technology, he integrated with it. He championed the field of Advanced (or Centaur) Chess,. In this style of chess, the chessmasters are not AI, but humans paired with their chess engines. Computer chess expert Kenneth Regan, and (separately) Likewise, the professional Go player who initially trained Google’s AlphaGo has now learned from the AI and advanced his career by a few hundred ranks, and the US Department of Defense is funding initiatives to build frameworks for machine agents to better collaborate with humans.
This narrative of integration is now stepping out of the realm of games and becoming more powerful [Satellite: HUMANS VS AI; HUMANS + AI]. Harvard Business Review points out that in their study of 1500 businesses, the most successful were the ones that complemented “the leadership, teamwork, creativity, and social skills” of humans with ”the speed, scalability, and quantitative capabilities” of AI. This exploits a unique feature in both humans and AI technologies. Humans are far better generalists, able to pick up multiple domains and perform many different kinds of tasks, whereas AI applications are extremely specialized. The future may be this “cyborg” model, and given the coming population boom [Core: FEEDING THE BEAST] markets will find it far more cost-effective to use adaptable humans rather than discard them.
Integration brings with it its own set of challenges. It creates the need for highly adaptable skills – knowledge acquisition, retention, technology access skills – and may potentially break the existing education model by emphasizing the need for “broad” skills and opposed to “deep” knowledge. Countries with education sectors that perform badly at imparting such adaptability may rapidly find themselves on the back foot. Large-scale reskilling initiatives – possibly a paradigm shift in education – are due.
This has major implications on more than one narrative. Firstly, as global economic power shifts towards countries in the APAC region [Core: THE NEXT BIG ECONOMY], these technologies could allow countries to capitalize on this change faster and close skill gaps. For example, most statistical analysis nowadays has been reduced to a few lines of code; more sophisticated machine learning allows even fresh computer science graduates to create predictive models far superior to what expert statisticians of a previous decade might have achieved. This may even enable economies with increasing numbers of dependents to maintain their workforce output in the face of a shrinking worker class. Japan is an example of this model: not only do they plan for robots to usher at the Olympics, even service jobs generally thought to require a human – such as elderly care – can be offloaded, at least in part, to machinery, in the face of worker shortages.
In 2017, The State Council of China released the “New Generation Artificial Intelligence Development Plan” which outlines AI as a national priority; it calls for China to match the US by 2020 and to be the world’s premier artificial intelligence innovation center,’ by 2030. New startups are rising, possibly as a by-product of this policy: an increasing number of Chinese AI startups are achieving $1 billion “unicorn” status. It is highly likely that China and the US will experience problems and the benefits first by introducing AI into markets of high technological sophistication and competition. They may emerge from the 2030s far more technologically sophisticated than any other nation in the region. China is already the world’s #1 robotics producer.
Other countries are not standing idle. Singapore, in particular, has announced $150 million to be made available for AI investments. By AI patent applications, Japan and the Republic of Korea still remain in the top 10, despite only accounting for a fraction of the development compared to the US and China. While automation is projected to disrupt labour markets by eliminating jobs in the short term in certain sectors, the resultant boost to the economy provided by these new technologies will create demand for more jobs in different sectors. A recent study published by Oxford Economics and CISCO on the future of ASEAN jobs posits that the jobs landscape in 10 years will look very different and there will be high demand for niche technical and IT skills along with sound management and interactive skills. Several countries in the region have already started initiatives to reskill their workforce to cater to these new jobs. The Government of Singapore has introduced SkillsFuture: a national movement to provide Singaporeans with the opportunities to equip themselves a broad range of skills, regardless of their starting points. In the Philippines, the K-to-12 education program has been redesigned placing more emphasis on skills development and aligning formal education with technical and vocational education and training (TVET). The Skill India program has introduced skill vouchers where the youth can undertake skilling programs of their choice at approved providers.
Then there is the ancillary effect of these technologies being abused for power. It is common to paint China as the antagonist, as a supposedly dystopian surveillance state that rigorously persecutes minorities while quantifying everything from your travel privileges to your dating. But the other economic giant also seems to be trending this way: for example, in India, a new proposed Data Protection Act gives Indian intelligence agencies unprecedented access to the data on every citizen. Even less authoritarian states and champions of democracy do this stuff. In the 2008 American presidential election, the Obama campaign drew “from voter registration records, consumer data warehouses, and past campaign contacts” to build a prediction model for every single voter in the United States – one that assigned scores to every person based on the likelihood of them a) voting b) voting for Obama. By 2012, the Obama campaign was even more sophisticated: led by Dan Wagner, the campaign’s team of engineers and programmers “began the election year confident it knew the name of every one of the 69,456,897 Americans whose votes had put [Obama] in the White House”. And then, of course, there is corporate America’s far-reaching credit scores, which American media somehow seems to ignore when they talk about China’s systems.
There seem to be no exceptions to this AI-powered game: only good players and better ones. Whether we like it or not, the minarchist state is effectively a myth, relegated there by the technology that is available and used by people trying to stay in power – which would explain why Chinese surveillance systems are now showing up in Ecuador. Needless to say, the UN Declaration of Human Rights stands violated.
The underlying value of data has led to ‘data protectionism’. For example, China and Vietnam have regulations which necessitate all domestic data be physically located in their countries, and India is considering a similar move (the earlier norm was to let data move freely between countries since it allowed cheaper processing). While this is welcome from a privacy perspective, it can also stifle domestic innovation and growth if these regulatory responses become too restrictive. The impact of the ‘data as resource’ narrative and the ‘data protectionism’ response on economic growth and digital innovation is still unclear.
These narratives feed into increasing militarization in the region [Core: THE DOGS OF WAR]. With the advent of the 4th Industrial Revolution, the way countries address conflict – both external and internal – is changing. As governments and infrastructure become increasingly connected in the APAC, particularly in developing countries, cyberwarfare is becoming an increasingly powerful threat in the region. Cyberwarfare attacks are extremely hard to predict, because this is the realm of criminal activity, and all we know are the hackers who are caught in some fashion. With the accumulation of data-capital and an increasingly connected future, the potential for being hacked, and the disruption thereof, is expected to be enormous when paired with both a powerful weapon and a lucrative target for cyberwarfare operations [Satellite: IoT AS A THREAT].
- The human + AI narrative [Satellite: HUMANS VS AI; HUMANS + AI] appears to be particularly popular in Japan: from Japanese literary AI co-writing (with a human) a novel that made it to the shortlist of a literary award, to robot guides at the Tokyo train station, to the Japanese government investing a chunk of a $920 million grant in the pursuit of ‘cyborg’ research, the general sense is more of acceptance than of fear. Writers lay this difference at the feet of differing social norms and sensibilities. It’s no surprise, then, that Japanese companies make up five of the ten largest robotics manufacturing corporations in the world. South East Asian countries and China, in sharing these sensibilities, may prove to be a market that keeps Japan at the cutting edge of such work – AI assistants are already helping human bank associates in Malaysia to identify customer emotions, smart wearable devices are giving superhuman strength and endurance to human workers at Hyundai Motor Company in South Korea, and China has the most number of industrial robots in operation in the world – making East Asia a potential bleeding edge of future cyborg narratives, as often seen in works of the cyberpunk strain of science fiction.
- Interestingly, there is a counterweight: the giants in the AI are Amazon, Baidu, Facebook, Google, IBM, Microsoft, Tesla Motors and Nvidia – the majority of which are American corporations. Despite the economic pendulum swinging towards APAC, overall AI investment remains strongest in the old world order: an analysis of 10,000 AI startups reveals that the APAC region has yet to see the kind of AI startups and venture capital that North America and Western Europe see. These private investments in AI technologies make all the government initiatives look like scraps from Oliver Twist’s table, and may very well keep America ahead in back-end technological sophistication – if not invisible social integration.
- A second narrative is the ‘AI gig economy’ . The gig economy has shown benefits to APAC countries by providing part-time labor [Core: THE NEXT BIG ECONOMY]. However, cheap and easily performed tasks may be automated, leading to even less protection for gig economy workers. This increasing lack of security offered may force more workers to unionize and fight back to stave off poverty. The conversation around jobs inevitably leads to education: what skills will be crucial to face the workplace of the next several decades? And what will this mean for the Asia Pacific Region? The National University of Singapore has launched a Lifelong Learners Initiative where every enrollment is valid for 20 years. Non-profit initiatives like the 10 Minute School in Bangladesh are revolutionizing education with simple tech and thousands of volunteers. Additional impetus for modular, flexible syllabi with low barriers to entry are rising in the form of MOOCs. A pertinent question, asked by Lev Kaye of CredSpark, is how the advantages are distributed among generalists and specialists.
- Aspects of gender inequality may become entrenched in AI systems and left unchecked, unless thorough frameworks and actors are in place to investigate such bias. While many believe it will bring gender equality closer, examples already exist of AI recruiting tools in Amazon that discriminate against women. A rising narrative is that as AI takes over more jobs, women will have the most to lose. There is an increasing counterthrust from civil society around issues of bias. While AI initiatives from civil society have not slowed down the AI arms race, they have proven key to highlighting concerns around implementation, especially with regard to aspects of governance. One notable example is the ProPublica case study on the KOMPASS system, which highlighted a type of sentencing discrimination against African-Americans. A more equitable future may very well depend on civil society being able to understand and interrogate technology in this manner.
- Algorithmic social control becomes an even more plausible reality. Both negative and positive narratives are touched on in the realm of science fiction. The negative narrative, examining the creation of China’s social credit systems, FICO scoring systems, and the technology available today, posits a future where hidden algorithms and overwhelmingly complete surveillance sort humanity out into subtle approximations of castes and dictate human goals. On the flip side, their positive narratives include the creation of more efficient welfare states by taxing wealth, using hitherto unavailable insight into financial transactions and individuals empowering themselves through technology to exit dominating states.
- European policy forms a counter-narrative that strikes at the core of these developments. GDPR, in particular, is interesting: it cracks down on the secondary use of data, where data is used for purposes other than which it was originally collected for. The vast bulk of machine learning datasets can be said to fall under secondary use. GDPR would thus hamper localized AI progress. Policymakers inspired by Europe’s example may change the APAC region’s prospects with laws like these. In addition, as pointed out at RightsCon, the creation of the European Union allowed smaller European countries to negotiate better with nonstate actors; the accumulation of data-capital may empower regional integration initiatives like BIMSTEC and ASEAN, alongside with regional norms and frameworks, if for no other reason that to have better negotiation abilities.
- In the midst of this, there stands the narrative of education and the reskilling challenges of the future. Anshul Sonak, Senior Director (Education & Innovation Initiatives) at Intel,
points out that this pressure may, in a sense, force a return to an older form of work, one where people didn’t necessarily rely on universities, but worked, learned, unlearned, worked in a messy, cyclical, but also highly resilient manner. That is one possible path. The other is, of course, the rise of modular, flexible learning, as exemplified by MOOCs today – while problems have been highlighted in their adoption, significant progress has already been made in adopting 4IHR tools for the task. Squirrel, a Chinese startup, has unveiled a system that, by breaking a curriculum into small units and analyzing gaps in students’ knowledge, has shown itself capable of designing and delivering a syllabus customized to the individual student – and this system already runs across 2,000 learning centers in 200 cities, a network reportedly as large as New York’s entire public school system. It could very well be that reskilling and education fears – large;y based on implicit assumptions of how our human-led education systems function – may very well be minimized by a new wave of personalized, AI-led teaching.
This report has been written by Yudhanjaya Wijeratne, Merl Chandana, Sriganesh Lokanathan and Shazna Zuhyle of LIRNEasia with commissioning by the UNDP Regional Innovation Centre (RIC) as an exploratory and intellectual analysis; the views and opinions published in this work are those of the authors and do not necessarily reflect or represent the official position or policy of the RIC, United Nations Development Programme or any United Nations agency or UN Member States.