Computer scientists have reached impressive breakthroughs in AI. In China, the use of AI will particularly play an important role in the shift from a capital-intensive model to a consumption-based economy.
05.10.2018 | 10:41 Uhr
Artificial intelligence (AI) is a set of computational technologies that are inspired by the ways human use their nervous systems and bodies to sense, learn, reason and take actions. Sensors, including microphones and cameras, collect data in the external environment from our day-to- day interactions. Algorithms are coded to condition machines to gradually learn and make inferences based on past data. Over the years, AI has grown tremendously in complexity, transforming from handling process- driven tasks to data-drive ones.
The term AI was officially born in 1956 at the Dartmouth Summer Research Project on Artificial Intelligence. Between the late 1960s and early 1980s, theories such as heuristic search, computer vision, natural language processing, mobile robotics, machine learning and artificial neural networks emerged. However, by the mid-1980s, AI still saw no significant practical success. The gap was in part because of the lack of direct access to environmental signals and data, and in part because of over-emphasis on characterizing true or false logic and thereby overlooking uncertainty. Interest in AI began to drop and funding dried up. The “AI winter” loomed for the next decade and started to see a new boom in the 1990s.
Technological advancements
According
to a Zignal Lab report from May 20171, 90% of the world’s data was
created in the last two years thanks to media and social intelligence.
2.5 Exabytes of media data are now produced daily. Coupled with improved
quality and wide availability of different hardware over the years,
data can now be fully and accurately captured, processed and shared.
Sophisticated software programs have also been developed to interpret
the output of algorithms and present data in easily digestible
visualizations, allowing humans to interface, analyze and synthesize
insights without the necessity of prior trainings in AI. Moreover,
computing and storage power have grown exponentially while costs are
reduced dramatically, further accelerating the processing efficiency and
the pace of innovation.
Another reason for this accelerating momentum hinges on the economic reality faced by the world’s major economies. The example of China is a telling one. The world’s second largest economy has relied heavily on its vast labor market and significant capital investment to sustain its economic growth over a long period. These two levers were the traditional drivers of production, yet they can no longer sustain the steady march of prosperity enjoyed in the past three decades. China’s demographics are turning from a tailwind to a headwind. With an aging population – nearly 50% of it is now middle-aged – China seems likely to fall well short of the workforce numbers needed to sustain economic growth at current productivity levels. The government acknowledges the increasing need for China to embrace a new growth model that relies less on a capital-intensive model (fixed investment and exporting), and more on private consumption, services and innovation to drive higher quality and more sustainable economic growth. China has to undergo structural reforms to address challenges arising from the past high-speed growth, such as excess capacity in numerous industries. The most realistic alternative for maintaining momentum would be to sharply accelerate productivity growth. The use of artificial intelligence is thus set to play an important role in boosting productivity. AI can augment labor by complementing human capabilities, offering employees new tools to enhance their natural intelligence. In our view, a significant part of China’s economic growth from AI will come, not from replacing existing labor and capital, but in enabling them to be used much more effectively. As AI technology continues to evolve and value slowly outweighs the cost of adoption, these emerging markets are more willing to adopt AI to solve some of their most pressing issues.
In the Middle East, the dependence on energy is pushing governments to diversify and leverage technology more. Even though gulf countries focused on cost-cutting after the oil price collapsed from its peak price in 2014, most countries today still have fiscal breakeven oil prices above the average trading Brent price. In Saudi Arabia, foreign exchange reserves collapsed from a peak of $750 billion USD in 2014 to below $500 billion USD in 2017, as the government had to draw down the reserves to cover a budget deficit caused by low oil export receipts.
China consumer - AI changing how people shop online
China
is an interesting case in which AI-powered tools are already widely
adopted in people’s daily lives. For instance, Taobao, Alibaba’s leading
online shopping website with over 500 million users, applies AI across
the entire customer journey: from product discovery to purchase
decision, delivery and after-sale service. AI redefines the shopping
experience for millions of China’s online shoppers and merchants. Some
of these developments may be subtle to the consumers, but they actually
significantly improve the consumption experience. For instance, after
opening the shopping app, users will see virtual storefronts that
display information tailored to them as individual shoppers based on
their unique characteristics and preferences. The system is smart enough
to use users’ browsing history, shopping history and other interesting
behavioral traits that they leave online to create a list of products
and advertisements appropriate for each potential shopper. The product
search results also become uncannily precise. All of these features are
implemented in real time with the aim of increasing the opportunity to
match what a person wants with what is available. These AI-powered tools
are capable of increasing conversion rate (sales) by as much as 20%.
Customer-service chatbot is another example how AI is slowly changing how consumer service are delivered. Ali Assistant, Taobao’s customer chatbot, now handles more than 90% of customers’ queries. It can conduct both spoken and written queries, functioning as both a customer- service representative and personal shopping assistant. Ali Assistant not only can provide answers on specific transactions, it can also recommend products based on text, photo or voice description. The chatbot function not only assists in lowering cost and increasing customer service efficiency, it also provides a more friendly-efficient environment along consumers’ shopping journeys and increases the chance that consumers come back to the platform for future consumption. All these amazing capabilities are only made possible with the emerging AI technologies in the fields of voice recognition and natural language processing.
China has a gigantic online
shopping segment, which creates a substantial challenge – how to rapidly
deliver millions of packages daily to customers’ hands on a consistent
basis at a manageable cost? This is where AI can substantially improve
productivity. The combination of consumer behavioral data, real-time
warehousing and logistical data, as well as automation of warehouses
allow delivery companies to speed up the delivery time and optimize the
utilization of infrastructures (e.g. warehouses and trucks).
One of
the best examples of this in China is Alibaba’s Cainiao system, a smart
logistic system that can identify which box is required to pack items of
different sizes or weight, thus optimizing the use of materials. This
AI-driven solution already reduces the use of packing materials by more
than 10%. With the help of AI, the system can also determine the fastest
and most cost-effective delivery routes to speed up parcel delivery. In
addition, it helps online and offline merchants to forecast product
demand, and thus prepare and allocate the appropriate level of
inventory. As a result, merchants can reduce working capital cost and
minimize the loss of revenue due to out of stock popular items.
AI’s
ability to efficiently handle large volumes of data to generate a useful
prediction or action also applies to other logistics needs. For
instance, China’s food delivery market (with 18 million orders per day),
represents nine times the size of the US market and faces a daunting
challenge to deliver time-sensitive goods (such as hot food) to feed
hungry clients quickly. Meituan-Dianping, a leading food delivery
application in China, uses the logic discussed previously in its AI
engine to generate the most time-efficient delivery routes. Meituan’s
o2o Real Time Logistic Dispatch System is an engine that uses predictive
modeling to support millions of orders between restaurants and
customers every day with an average delivery time of less than 30
minutes.
Transport generally poses greater problems in China than in
the rest of the world because of the country’s high density. For
example, some of its major cities are as big as five cities in other
parts of the world. Didi Chuxing, a leading ride-sharing application in
China, operates on a scale which is five to six times larger than that
of its US competitors in its home market. The company uses AI, by taking
into account weather, car numbers, customer profile and road conditions
to forecast rider demand and car supply 30 minutes in advance with over
80% accuracy. By deploying AI, the system dispatches drivers in advance
to meet potential demand and provides the most efficient route so that
drivers can pick up customers and deliver them to their destinations in
the shortest time. In Hangzhou, the government partners with Alibaba to
deploy a smart city system to manage the city’s traffic flow by using a
combination of AI powered traffic light and accident detection systems.
The city’s traffic congestion is reduced by 10% as a result.
In
January 2018, Seegene Inc., a South Korean biopharmaceutical company
that manufactures In Vitro Diagnostic (IVD) products, became the first
company in the world to have successfully developed diagnosis reagents
using a newly created AI based system. The company set up the system
with data on pathogen and disease information accumulated over the last
15 years. This approach simplified a complex research and development
process through its self-developed algorithm and virtual experiment,
reducing the development period from a typical one year to just four
days. In general, a researcher needs to examine 200 to 300 cases of data
each year to develop a new drug. An AI system can parse through more
than 1 million dissertations and clinical test data for 4 million people
during the same period.
Besides Seegene, many other Korean
biopharmaceutical firms are actively leveraging AI in their research and
development process. In December 2017, CJ Healthcare Corp signed an
agreement with genome and exome data analysis firm Syntekabio Inc. to
co-develop a new anticancer immunotherapy using an AI model. Dong-A ST
Co. has been collaborating with the u-Health Information Research
Institute at Ajou University since 2016 to develop new drugs by
analyzing data on patients’ medical records.
In
June 2017, Gazprom Neft, an integrated oil and gas company in Russia,
signed a Cooperation Agreement with Yandex, Russia’s leading internet
search company, on implementing AI-enabled projects in the oil and gas
sector. The two parties are developing Russia’s first integrated
platform for the processing and interpretation of seismic data, a
platform to support the entire seismic-survey cycle. AI helps reduce the
significant costs that are typically incurred in managing disjointed
data and modules. The end goal of the platform is to develop cognitive
assistants that will process information and carry out calculations in
order to provide engineers with pre-prepared solutions for further
actions.
AI ecosystem beyond platform companies
Semiconductor
With
more data collected, demands for computing power and storage memory
have skyrocketed. In the driverless cars industry, Google and Intel
estimated that there would be 4TB of raw data, or 400GB of compressed
data, collected in an average 1.5-hour driving day. The large amount of
data increases demand for servers, which are needed to store and process
data with low latency. Alliance Bernstein estimated last year that
cloud servers would grow 50% and enterprise servers would grow 25% by
2025.
Within DRAM memory alone, analysts estimate that revenue
derived from assisted driving and data centers will increase to $24
billion by 2030. Memory is just one example in the downstream
semiconductor sector that is set to gain from the AI revolution. All
parts of the value- chain, including silicon wafer, chips, and foundry,
will benefit from the megatrend. Semiconductor strongholds Taiwan and
Korea, and even newcomer China, should benefit.
One of the most exciting AI applications is the eventual shift to autonomous driving. Electric vehicles are often deemed the ideal solution for autonomous vehicles, from both an environmental and an engineering standpoint. Adding autonomous driving equipment to a car adds weight, aerodynamic drag, and electrical power consumption, leading to increased fuel consumption. However, with electric vehicles, emissions can be reduced by 55% to 65%. On the engineering side, there are much fewer moving pieces in an electric vehicle, resulting in less room for mechanical failure.
Should the world reach a 100% electric vehicle adoption, demands for
many metals will skyrocket. The vast opportunity lies in not only the
battery packs but also the body and the motor. These estimates do not
yet include the additional demand created from complementary
applications such as grids and charging infrastructure, applications
that accelerate the need for metals. While electric vehicle adoption is
still nascent, investment and consolidation in the metal industry have
picked up significantly.
Cobalt, an essential metal in cathodes, is
constrained by supply in tonnage and origin. Almost 60% of the world’s
unrefined cobalt output in 2017 came from the Democratic Republic of
Congo, whose output was more than 10 times that of the second producer
Russia. The price of cobalt has nearly tripled in the last two years,
surging 129% in just 2017. Players along the value chain scramble to
secure additional supplies. GEM, a China-based battery producer and
recycler, struck a deal in March this year to buy a third of the cobalt
output from projects owned by international commodities major Glencore
over the next three years.
On the lithium front, even though it is
not as supply-constrained as cobalt is, suppliers have quickened the
pace in reaching greater concentration in an already oligopolistic
market. In May this year, China-based Tianqi Lithium agreed to buy a
23.8% stake in Chile’s Sociedad Quimica y Minera for $4 billion USD. The
acquisition will boost Tianqi’s global share of the metal’s output to
18 % from 13%.
In addition to cobalt and lithium, other EM mining
firms such as Norilsk Nickel (Nornickel) in Russia and Vale in Brazil
are also principal players in the race for electrification. Nornickel is
the world’s leading producer of nickel, palladium and copper. Vale is
the largest producer of iron ore and nickel in the world, and an
important producer of manganese, copper and cobalt.
For most of the EM countries, especially in Latin America, South East
Asia and India, digitalization and 4G are still novel concepts. For
example, the average 4G penetration in Latin America is just 17% and
only 26% of total subscribers are on post-paid plans. However, as the
economy recovers and as rollout for end-applications using digital
services accelerates, growing evidence shows that customers are willing
to pay more for better connectivity and network coverage. As 4G
penetration deepens, the percentage of post-paid subscribers should
therefore rise, enhancing the value proposition and monetization of
telecom operators.
In more advanced EM countries, such as China and
Korea, operators are moving into 5G deployment and value-added services
such as payment, e-commerce, Internet of Things and data center, riding
the AI growth wave.
China
is an interesting case as to why so many AI-related investment
opportunities are emerging. Firstly, China has a large data pool across
many verticals and industries. The country has over 900 million
smartphone users with easy internet access (four times the number in the
US or India), the market is also an eager adopter of leading
technologies in their daily lives. For instance, the total value of
mobile payment transactions in China is more than 10 times that in the
US. Digital penetration is very high across a range of day-to-day
applications, ranging from retail and travel to entertainment and local
services. This large volume of data generated by various applications
form the backbone of AI development.
Secondly, China has a large pool
of low-cost engineering talent. Every year, 3-4 million students
graduate in math, science and engineering or technology related
disciplines vs. approximately 500,000 in the US. Today, China has the
world’s highest number of R&D personnel, three times the number in
the US. China’s large economy and booming business opportunities mean
Chinese technology companies can now compete with western firms in
hiring the most talented engineers. Chinese companies have opened
research labs in Silicon Valley and offer comparable salaries.
Last
but not least, the Chinese government is highly supportive of AI
development and is already setting the goal of positioning the country
to be a major AI innovation hub in the next decade. In 2015, the Made in
China 2025 policy was announced by the State Council – the first
ten-year action plan that calls for green, innovative and intelligent
manufacturing in China. In the “Internet+” Action Plan, the plan listed
AI as one of its 11 key focus areas. In 2016, the Chinese government
published the Three-year Implementation Plan for “Internet Plus”
Artificial Intelligence. This identifies six specific areas of support
for AI development, including capital funding, system standardization,
IP protection, human capital development, international cooperation and
implementation arrangement. The importance of AI was reiterated in 2017,
with the government’s New Plan on Artificial Intelligence Development,
in which China’s goal is to catch up with global leaders and achieve
world-leading positions in AI by 2030 by solving issues such as a lack
of high-end computer chips, software and trained personnel. The
government looks set to play a growing role through policy support and
regulation. It expects China’s overall AI industry to be synchronized
with international development, and to lead the global market in
system-level AI technology and applications.
Most EM countries have favorable policies toward AI adoption. According to SAP, more than half of 1,500 AI research projects in Russia in the past decade were paid for by the state2. Hungary and Poland have actively supported the growth of start-ups by setting up special economic zones, investing in infrastructure, and providing tax breaks. In the UAE, the government just appointed its first minister for AI in October 2017 and it launched a $270 million Dubai Future Endowment Fund in the same year. Saudi Arabia shares a similar ambition to invigorate the economy through innovation in the National Transformation Program 2020 and Vision 2030 programs announced in 2017. In November 2017, Saudi Arabia’s Crown Prince Mohammed bin Salman also pledged $500 billion to build a new, high tech city called Neom on the Kingdom’s Red Sea Coast.
For
EM markets as a whole, some existing structural issues should get
resolved in the short to medium term. Currently, lack of talents is the
most commonly cited reason for the gap in AI development that exists
between developed and emerging markets. Many countries also lack
infrastructure. However, the long-term attractiveness of AI investment
and the rapid adoption in EM are clear trends. Many foreign owned
companies have long deployed resources in EM. For example, in India 58%
of the companies are using AI at work at scale4. The sector is dominated
by foreign firms such as Accenture, Microsoft, and Adobe, all of which
have their innovation centers in India. Foreign presence in the country
is an important step to AI adoption as it helps raise awareness and
educate the local workforce.
In China, the country is gradually
narrowing the gap with developed markets in some key technology building
blocks. China has already outstripped the US in terms of AI research
publication and citation. It is already spending more than the European
Union in terms of R&D and the spending gap with the US continues to
narrow quickly. Today, China has over 600 AI start-ups, making it
already second only to the US. Since 1999, China has invested about USD
10 billion in AI, again bringing the nation second to the US (with USD
16 billion). As detailed previously, an increasing number of leading AI
talents are coming back to China to work for local companies. China is
seeing the rapid emergence of domestic technology leaders, such as
Baidu, Alibaba and Tencent. These companies have participated in over
300 AI-related equity deals since 2014, over 50% of which have been
outside China. They have created investment and research centers around
the globe including in the US and Israel. AI-related investments in
China are also slowly diverging beyond the popular fields (e.g. computer
vision and voice recognition)
In short, despite a number of challenges, AI could bring many new opportunities to emerging markets. A number of markets, such as China, are already progressing at the forefront. More companies from the region are well positioned to benefit from AI, either as technology drivers of AI or as the early adopters of AI in traditional industries.
Diesen Beitrag teilen: