Asias AI agenda
How Asia is speeding up global
artificial intelligence adoption
A briefing paper in association with
Asias AI agenda
Preface
Preface
Asias AI agenda is a briefing paper by MIT Technology Review. It is
based on surveys of business leaders from across the Asia Pacific
region conducted between September and October 2016. Further
insights were gained through in-depth industry interviews and are
included in this report. We would like to thank all participants in this
research project as well as the sponsor, human capital management
solutions provider ADP.
MIT Technology Review has collected and reported on all findings
contained in this paper independently, regardless of participation or
sponsorship.
November 2016
MIT Technology Review, 2016. All Rights Reserved.
MIT Technology Review
Contents
1. Executive summary
2. Introduction: AI and Asia
3. Positively inclined
4. Human Capital and AI
5. Aware, but unprepared?
10
6. Conclusion: Putting the AI in Asia
11
7. Survey demographics and firmographics
12
8. Infographics on AI
14
MIT Technology Review, 2016. All Rights Reserved.
Asias AI agenda
Executive summary
1. Executive summary
sia-based senior executives at global
firms believe that the impact of artificial
intelligence (AI) and robotics on their business
performance in Asia will be profound and positive
and will be felt sooner than we may think.
Outside of global robotics industry leaders Korea
and Japan, most of Asia currently lacks the depth
of technical skills and R&D facilities needed to keep
pace with AI development. However, China, India,
and other large Asian economies generate a copious
amount of data, a tremendous natural resource
that is critical to pushing AIs capabilities forward.
Ironically, given the commonly held view that AI will
be responsible for disintermediation of jobs at all
levels, it is Asias massive human capital dividendthe
billions of constantly Internet-connected workers and
consumersthat will propel AI development in the
region farther and faster.
MIT Technology Review surveyed over 60 Asiabased senior executives to gather perspective on the
impact of AI and robotics on Asias business landscape.
Additionally, two dozen senior HR professionals
were polled to assess the impact on jobs in Asia
and the future of their roles in particular. Several
in-depth interviews were conducted with AI industry
technologists, investors, and application developers.
The key findings of this research are as follows:
Overwhelmingly, respondents feel technological
advancements in AI and robotics will have very
positive effects on most industrial sectors in Asia.
They also believe even more strongly that these
technology advancements will specifically improve
their own firms competitiveness.
Executives also feel positively about AIs ability
to add value to their own industries, with the
exception of financial service executives, who
are generally less confident that automation and
machine learning will create purely constructive
benefits for their industry.
Most C-suite respondents feel AI will significantly
improve their own competitiveness in Asia,
especially their process efficiency and their ability
to delve into customer data to achieve better
insight. Again, financial industry responses lagged
the average.
Only a small percentage of respondents are
currently investing in AI development in Asia. 25%
of respondents, however, indicate that their firms
have already made investments at a global level
and another 50% are considering doing so.
Some 70% of HR executives feel that AI and
robotics adoption will result in significant job losses
in Asia over the next five years. Unsurprisingly,
nearly all of these respondents feel these
technologies will have a major impact on their roles
and functions in the future.
Specifically, HR managers and talent professionals
feel that their roles will evolve into broader, and
more strategic, productivity management roles;
two-thirds of them say their roles will encompass
the management of both human and artificial talent
in the next five years.
MIT Technology Review, 2016. All Rights Reserved.
MIT Technology Review
2. Introduction: AI and Asia
Asias quest for technology innovation and sustained economic growth drives
tremendous investment in AI
peculation on the future of artificial intelligence
(AI) and robotics technologies often posits
a binary outcome: either AI will profoundly
increase efficiency and convenience for the worlds
economies and its people, or it will irrevocably disrupt
(if not destroy) nearly every established industry and
the livelihoods they all support.
The reality will lie between these two extremes.
Based on research gathered from surveying Asian
business leaders and human resources and AI
professionals, this report argues that AIs future will
cleave much more closely to the positive outcome.
Moreover, this future appears to be approaching
quickly: advances in deep learning and the rapid
expansion of process automation in such diverse
sectors as manufacturing, transportation, and
financial services mean that AIs impact is growing
exponentially with each passing year. Decision makers
in all organizations must now begin to understand
how AI will alter their own operational processes and
those of suppliers, partners, and customers.
Asias business landscape is poised not only to
benefit greatly from AIs rise, but also to define it.
Asias quest for technology innovation and sustained
economic growth drives tremendous investment in
the space. Japan and Korea are looking to infuse all
manner of consumer electronics with intelligence.
China is attempting to complement and extend, rather
than replace, human labor to maintain its dominant
role in global manufacturing. China is also striving to
build up its intellectual property reserves to ensure
that its own firms are globally competitive in the
creation of AI and robotic technologies. Singapore
and Hong Kong are in a race to become leading
financial technology hubs, seeding investment in AI
tools in financial service applications.
MIT Technology Review, 2016. All Rights Reserved.
MIT Technology Review has conducted surveys
and interviews across Asia in developing this report,
which assesses how recent advancements in AI and
robotics are impacting Asias business landscape,
looking at the efforts that Asian governments,
research institutes, and venture capitalists are making
to hasten their development.
Asias business landscape is
poised not only to benefit greatly
from AIs rise, but to also define it
Asias AI agenda
Positively inclined
3. Positively inclined
Executives feel positively about AIs ability to add value to their industries
urvey respondents were asked whether they
felt robotics and AI would have a constructive
or a destructive effect on several market
segments; their responses were ranked from 1 (AI
adoption would result in the destruction of jobs
and processes) to 5 (AI would bring a significant
increase in the industrys value and efficiency).
Across industry sectors, responses were positive
over 3.8. They were higher in more technologydependent industries such as information
technology and communications (ITC), logistics,
and manufacturing.
Figure 1: AIs impact on industries in Asia
(1 being least positive, 5 being most positive)
Chemicals*
3.7
Commodities*
3.5
3.9
3.9
Consumer
goods, Retail
3.7
Financial
services
3.9
Information
technology &
communications
4.1
4.1
4.6
Manufacturing
4.1
Pharma,
Healthcare*
4
4
Professional
services
3.7
3.5
3.6
Property,
Construction
4.4
Transport,
Logistics
4.1
1
Respondents within the sector
Average
* Sectors for which only the Average metric had a statistically significant number of respondents.
Source: MIT Technology Review
MIT Technology Review, 2016. All Rights Reserved.
MIT Technology Review
Most CEOs surveyed feel AI will
bring positive benefits to their own
industryexcept executives
from the crisis-plagued financial
service industry
Tellingly, respondents were even more inclined to
see positive benefits from AI for their own industry:
most industry participants ranked the impact of
AI on their own sector higher than the average.
Respondents from one sector, however, were actually
more pessimistic: those from the financial service
industry on average saw AI as a positive influence,
yet many also felt that automated processes and
machine-based transactions would destroy value in
their industry.
Banking industry fears of artificial intelligences
destructive potential were also revealed when
respondents were asked to rate the impact that
these technologies would have on Asias industrial,
policy, and competitive landscape. Overall, survey
respondents felt that advances in artificial intelligence
would significantly boost Asias competitiveness as a
manufacturing and service center, benefit government
policy makers in their attempts to boost innovation,
and increase overall industry growth prospects. The
only industry cohort that did not feel as positive as the
respondent average was the one from the financial
service industry. On average, industry respondents
felt most encouraged about the positive impact AI and
automation would have on their own competitiveness,
ranking it 4 out of 5. Once again, the optimism from
respondents in retailing, ITC, and manufacturing
industries was higher than the averagewhile
sentiment from the financial service industry was
noticeably lower.
The relative caution and pessimism among
financial sector respondents could have to do with
past experiences. The 2007 global financial crisis
may not have started in Asia or hit the regions banks
and financial institutions as hard as it did elsewhere,
but the fallout, and the lingering regulatory and
compliance burden it created, plague Asian banks
to this day. Moreover, mini-crises fomented by new
technologies constantly punctuate the industry
landscape; flash crashes, for example, are fast,
sharp stock market plunges brought about by the
combination of human error and algorithmic trading.
Financial industry respondents may have these events
in the back of their minds when considering what
artificial intelligence means for their future.
Modest reticence from finance respondents
aside, AI has the potential to affect a number of
Asias development challenges, from food security
to public safety, transportation networks, and health
care. Lin Yuanqing, the director of the Institute of
Figure 2: AIs impact on Asias business landscape
What degree of impact will advancements in AI and robotics have over the next five years?
(1 being least positive, 5 being most positive)
5
4
3
3.7
3.5
Asias manufacturing
competitiveness
Asias service
industries
3.9
3.9
2
1
Source: MIT Technology Review
MIT Technology Review, 2016. All Rights Reserved.
Asias tech and innovation
My industrys
government policies
growth prospects
My companys
competitiveness
Asias AI agenda
Positively inclined
Figure 3: AIs impact on growth and competitiveness
What degree of impact will advancements in
AI and robotics have over the next five years?
(1 being least positive, 5 being most positive)
My industrys
growth prospects
My companys
competitiveness
Consumer/
Retail
3.9
4
Financial
services
3.6
3.6
Information
technology &
communications
4.3
4.2
Manufacturing
Deep Learning at Baidu Research (known as Baidu
IDL), believes that all this is coming to a head at
once: It is impossible to point to an industry that
will be first to adopt AI. Public transportation,
logisticsnearly every critical infrastructure platform
can benefit from it, and they are all interconnected.
AI will come to all industries at once, and it will come
sooner than we think.
AI industry executives feel that two linked factors
will allow autonomous and intelligent applications
to proliferate somewhat simultaneously. One is the
growth of big data in Asia, fed by many hundreds of
millions of people in densely populated cities who are
connected to the mobile Internet. Data is the most
important resource for successful machine learning
development, says Professor Zhang Yue, a machine
language researcher at the Singapore University of
Technology and Design. And mobile data, large in
volume and rich in context, provides AI developers
with a large amount of useful data. The harnessing
of big data through analytics gives rise to the second
factor driving AI development: firms are increasingly
willing not only to use analytics to increase their
own business performance, but to borrow and share
automated process insight across sectors.
A successful ecosystem for AI development
requires four input factors, according to Baidu
IDLs Lin: Big data, the continuous production of
algorithms, massive computation power, and big
applications. Lin describes this last factor as the
most crucial: applications that attract a large number
of users quickly drive usage, interaction, and data
creation to create a positive development loop for
deep neural networks, which are increasing in scale
and capability as processing power becomes ever
cheaper and more plentiful.
4
4.2
Professional
services
3.6
4.1
Source: MIT Technology Review
Data is the most important
resource for successful machine
learning developmentand
mobile data provides AI
developers with a large amount of
useful data.
Zhang Yue, Professor, SUTD
MIT Technology Review, 2016. All Rights Reserved.
MIT Technology Review
4. Human capital and AI
Tremendous change is imminent in the profession of human resources
n addition to compiling senior executives
views about AIs impact on the Asian business
environment, we surveyed two dozen senior human
resource executives in AsiaHR directors, recruiting
consultants, and talent acquisition headsto
understand their views on AIs potential to disrupt
talent management.
The results revealed an assumption that
tremendous change is imminent in the profession of
human resources: 70% of respondents feel that AI
and robotics will lead to substantial job losses in Asia
over the next five years, underscoring a long-standing
anxiety that many in Asia feel when considering
the impact of new technologies. The pace of work
displacement in Asia will be at a much faster rate,
because of the relatively higher percentage of
low-skilled jobs in the labor force relative to more
developed economies says Tak Lo, a partner at Hong
Kongbased AI accelerator [Link].
In his view, this not only heightens anxieties for
talent managers within enterprises but influences
government policy makers as well. Asian governments
are particularly suspicious of the threat AI poses to
their efforts to transform skills in the labor force,
he says, noting that governments should be more
focused on ways to retrain displaced workers rather
than defending existing jobs from disintermediation.
The views of the industry professionals surveyed
suggest that enterprises may soon be shifting
their practices in human resource management to
achieve that focus. Nearly all respondents felt that
their professions would be significantly altered with
the advent of AI. However, they also felt that these
changes would be positive. Rather than the HR
function narrowing as robots replace workers, the
majority of respondents thought that it would be
expanded to oversee the management of both human
and machine productivity.
Figure 4: Human resources
Will advancements in AI and robotics lead to substantial job losses in Asia over the next five years?
(% of respondents)
Yes (70.8%)
No (29.2%)
Will these technologieshave a major impact on the role and function of the HR chief?
(% of respondents)
Yes (87.5%)
No (12.5%)
How will the role of the HR functionchange with advancements in AI and robotics?
(% of respondents)
HR will be tasked with managing overall productivity,
and therefore both man and machine (66.7%)
Source: MIT Technology Review
MIT Technology Review, 2016. All Rights Reserved.
HR will keep managing people,
and only people (33.3%)
Asias AI agenda
Human capital and AI
AI industry professionals echo this positive view
of the technologys ability to enhance, rather than
replace, jobs. The notion that robots will soon steal
jobs at any skill level and decimate entire professions
is an exaggerated one, according to Baidu IDL
director Lin Yuanqing. Even the best robotic financial
analyst is only going to prepare a report at 70% of the
capability and insight of its human counterpart, he
says, though admitting that it will happen in a fraction
of the time. This is not scary. All our work lives will
be complemented, for the betterwe are not going
to be replaced. Despite AIs rapid advances (or
perhaps because of them), skepticism remains, and
the industry is taking pains to address the doubts.
Recently, an industry coalition called the Partnership
on Artificial Intelligence to Benefit People and
Society was formed by some of the worlds largest
technology firms, including Facebook, Google, and
Amazon. But while this groups almost apologetic
name suggests that its main function is to stem
the rise of job-robbing machines, the initiatives
primary role is to engender collaboration around
data gathering and analysis to accelerate cognitive
learning and drive further AI insight.
This is not scary. All our work lives
will be complemented, for the
betterwe are not going
to be replaced.
Lin Yuanqing, Director, Baidu IDL
MIT Technology Review, 2016. All Rights Reserved.
10
MIT Technology Review
5. Aware, but unprepared?
Only a small percentage of companies are currently investing in AI development in Asia
ecision makers at Asian-based firms are firmly
convinced of the benefits AI and automation
technology will bring to their own prospects
and the regions economy as a whole, but most
have not yet committed resources to capitalize on
these beliefs. A small percentage of respondents
indicate that they have invested in AI in Asia, and
nearly 20% say they have robotics and automation
commitments in the region. Larger percentages
indicate that they have AI or robotics investments
at a global level, and fully half indicate that they are
considering AI investments.
While multinational firms may not be committing
resources to AI development in Asia yet, domestic
industries across the region are most certainly doing
so, driven by a traditional response of policy makers
in the region, according to [Link] Tak Lo: The
external shocks that can happen create risk that
Figure 5: AI investment plans
(% of respondents)
20
13
43
50
25
AI (analytics,
machine
learning or other
AI capabilities)
Automation/
Robotics
13
Asian governments feel they need to take to mitigate
with technology progression, he says.
In China, the hope is to use automation to bolster
labor productivity as the countrys workforce ages
and shrinks. Over the longer term, China hopes to
build a domestic manufacturing capability in industrial
robots and, by extension, in process automation and
artificial intelligence. Singapore, by contrast is into a
little bit of everythingfintech, autonomous driving,
automated customer support, says Tak. Autonomous
driving is considered one of the most likely AI-enabled
applications, observes Professor Zhang of the
Singapore University of Technology and Design, but
largely as an anchor for establishing a broad-based
AI innovation platform. Singapore is aligning its policy
objectives across multiple sectorsurban mobility,
financial data science, health carein an effort to
elicit the synergies across them and speed up AI
development overall, he says.
Asia as a regional economy is poised not only to
benefit greatly from advancements in AI and robotics
technologies, but to also define themand perhaps
to lead their future development. AIs rise will create
a seismic shift in the processes that senior managers
use to manage talent, growth, and productivity across
nearly every industry sector. Asian businesses stand
to leverage AIs rise faster, thanks to the virtuous
cycle created by Asias technology investments and
the organic growth of big data.
18
18
We have invested globally
We are considering investment
We have invested in Asia
We have no plans to invest
Source: MIT Technology Review
MIT Technology Review, 2016. All Rights Reserved.
AIs rise will create a seismic
shift in the processes that senior
managers use to manage talent,
growth and productivity, across
nearly every industry sector
Asias AI agenda
Conclusion
6. Conclusion: Putting the AI in Asia
A holistic view of productivity development across an entire firm and its assets
is emerging
sian business leaders need to adopt an
extreme openness to successfully capitalize
on todays AI and automation trendsa
willingness to actively collaborate in data analytics
projects and share best practices and insight
resulting from their own process automation efforts.
These technologies may have powerful effects on
an individual firms competitiveness (as business
leaders assume in the survey results), but the quest
for competitive advantage cannot blind business
leaders to the greater benefits of collaboration, in
much the same way that competitive AI developers
themselves are increasingly pooling resources
to accelerate their collective progress. It will also
require leaders to champion process improvements
based on automation and machine learning that are
integrated into their firms broader talent management
programs. AI may start to disintermediate roles and
responsibilities across Asias economies, but it will
enhance and redefine far more, while increasing the
productivity of all workers.
The research undertaken for this report
has elicited three primary take-aways for senior
executives managing businesses across Asia:
AIs big bang. The C-level executives surveyed and AI
industry experts interviewed are largely aligned in the
belief that AI and robotics will impact every industry
significantly, positively, and all at once. Rising investment
in process automation in Asian manufacturing tends
to focus much of the industrys attention, with good
reason: three Asian countriesChina, Korea, and
Japanpurchased nearly half the worlds shipments of
250,000 articulated robots last year. But importantly,
survey respondents largely felt that most other sectors,
from retail to logistics, would benefit equally from AI
technology. AI industry executives feel that big datas
growth in Asia will allow autonomous and intelligent
applications to proliferate across almost all industries in
the region more or less simultaneously.
Aware, but unprepared. C-level executives surveyed
also firmly believe that process automation will not
only improve their own business performance but
impact industry growth overall. Yet only a small
percentage of respondents indicate that they are
currently investing in AI or robotics in Asia, and
although many more say they are considering
investment, it is clear that business leaders in the
region must take more active steps to prepare for AIs
rapid rise.
Change Asias workers can believe in. Most Asiabased human resource professionals surveyed feel
AI and automation will soon create significant job
losses in the region. At first glance, this outlook paints
a bleak picture for Asias workers, who mostly live
in economies where automation threatens the very
entry-level jobs upon which they depend to build skills
and advance careers. But when asked what sort of
impact this would have on the profession overall, most
HR industry participants believed that their roles would
evolve and expand so that they would manage and
develop both human capital and inputs from machines.
The emergence of a holistic view of productivity
development across an entire firm and its assets,
coupled with the high hopes Asia-based senior
executives have for AIs influence on their firms
success, means it is likely that business leaders
will focus their investment and adoption efforts on
complementing and enhancing job functions with
automation, rather than eliminating them.
MIT Technology Review, 2016. All Rights Reserved.
11
MIT Technology Review
7. Survey demographics
and firmographics
ver 60 Asia-based senior decision makers
responded to our 10-question survey. We
sought their opinions on advancements in
the artificial intelligence and robotics industries, and
we asked how these industries would impact growth
prospects for Asian business.
Our respondents were senior regional decision
makers from over a dozen industries, with large
representation from the financial service industries,
consumer goods and retail, and professional
service sectors. More than half of our respondents
were CEOs, and a further 20% of respondents
were CFOs or other C-suite executives. The
majority of our respondents were executives at
large multinationals headquartered outside of Asia;
roughly equal percentages reported that their head
offices were in Europe, North America, or the AsiaPacific region. Over 70% of respondent firms had
global annual revenues of over US$1bn, and over a
quarter represented firms with turnovers in excess of
US$10bn annually.
Figure 6: Respondent demographics and firmographics
(% of respondents)
North America
Asia
14
.3
12
.7
Australasia
$500m to $1bn
Source: MIT Technology Review
MIT Technology Review, 2016. All Rights Reserved.
$1b to $5bn
$5bn to $10bn
27
14
.3
27
4.
8
11
.1
15
.9
$50m to $500m
> $10bn
Other Director-level
Size of firm
<$50m
7.9
9.
5
1.6
11
.1
Director of Sales and/or Marketing
33
.3
CXO
3.
CIO
Other
7.9
Transport, Logistics
28
.6
CFO
1.6
1.6
Pharma, Healthcare
34
.9
Europe
7.9
Hotels, Leisure
Media, Marketing
Property, Construction
CEO/Managing Director/President
HQ
Financial services
Manufacturing
54
Roles
Engineering
9.
5
Professional services
14
.3
1.6
20
.6
Consumer goods, Retail
Information technology & communications
1.6
Chemicals
6.
3
14
.3
Industries
1.6
12
Asias AI agenda
Survey demographics and firmographics
Within the region, most respondents were
based in Southeast Asia, where their firms
regional headquarters were located, with
sizable numbers based in Greater China and
South Asia as well.
Respondents managed diversified
businesses across Asia. A third reported direct
manufacturing presence in the region, roughly
equally present across China, Southeast Asia,
and South Asia. Over half reported sales,
distribution, and investment activities in each
of those Asian regions. Respondents managed
diversified businesses across Asia. A third
reported direct manufacturing presence in the
region, roughly equally present across China,
Southeast Asia, and South Asia. Over half
reported sales, distribution, and investment
activities in each of those Asian regions.
Figure 7: Regional presence
(% of respondents)
14.8
16.9
3.3
9.6
42.2
52.5
Regional HQ
Respondent
location
27.9
21.7
9.6
1.6
ANZ
Greater China
ASEAN
India/South Asia
East Asia (Korea, Japan)
Source: MIT Technology Review
Figure 8: Regional activities
(% of respondents)
ANZ
Greater China
7.9
Third-party
distribution or
franchises
12.7
25.4
Manufacturing
presence
Direct wholesale
and/or retail
activities
ASEAN
South Asia
27
30.2
34.9
44.4
36.5
54
42.9
25.4
27
27
39.7
34.9
33.3
19
39.7
47.6
14.3
Investment
activities
East Asia
Source: MIT Technology Review
MIT Technology Review, 2016. All Rights Reserved.
13
MIT Technology Review
What Is AI?
A Time Line of AI
It has grown up out of many different
disciplines and has many denitions.
Here is ours.
After a century of ups and downs,
artificial intelligence is getting smarter.
rticial intelligence, as we
use the term in this report, is
an evolving constellation of
technologies that enable computers to
simulate elements of human thinking
learning and reasoning among them.
Regular improvements in Googles
search algorithm, for example, come
frommachine learning, a type of AI that
programs systems to learn from data,
nd patterns in it, and make predictions
about it. The same technology has
been pivotal to voice and image
recognition as well as advances in selfdriving cars. Integral to many recent
improvements in the eld is a form of
machine learning called deep learning.
Looselymodeled on the way neurons
and synapses in the brain change
as they are exposed to new input, it
has been used independently or in
combination with other AI approaches
to help machines tackle tricky tasks and
exhibit something resembling intuition,
in some cases performing tasks better
than humans.
1914
1943
Neuroscientist Warren McCulloch and
logician Walter Pitts present a calculus
based on neuron-like logic units that can
be connected together in networks to
model the action of a real brain.
1956
John McCarthy, Marvin Minsky, and
Claude Shannon organize a summertime
research meeting at Dartmouth that
brings together the leading thinkers
on information theory, articial neural
networks, and symbolic logic, christening
the eld articial intelligence.
1960
Frank Rosenblatt demonstrates the
Mark I Perceptron, an attempt to create
an articial neural network for image
recognition that the New York Times
calls the rst step toward a computer
able to walk, talk, see, write, reproduce ...
and be conscious of its existence.
Placing That Face
As of 2015, groups
reported different results
using AI techniques like
deep learning on large
data sets, but computers
are sometimes better
than humans at
recognizing a face.
Percentage error in face recognition
14
MIT Technology Review, 2016. All Rights Reserved.
1950
In a paper that helps establish a practical goal
for articial-intelligence research, Alan Turing
proposes a game to answer the question Can
machines think? He predicts that by 2000
computers will be able to pass as human more
than 30 percent of the time.
1958
Oliver Selfridge presents a paper in England
describing Pandemonium, a new model of a
neural network based on lower-level data
demons working in parallel with higher-level
cognitive demons in order to perform pattern
recognition and other tasks.
1961
Marvin Minsky publishes his foundational paper,
Steps Toward Articial Intelligence.
3.67
3
2
1.65
0.53
0.37
Microsoft
Source: MIT Technology Reviews AI Takes Off Business Report
In what would come to be described as the
worlds rst computer game, Spanish inventor
Leonardo Torres y Quevedo debuts El
Ajedrecista, a machine that can automatically
play chess thanks to a simple algorithm built into
its mechanical design.
Facebook
Chinese
University of
Hong Kong
Google
Human
0.23
Baidu
Asias AI agenda
Infographics on AI
2000
Cynthia Breazeal designs a sociable
humanoid robot named Kismet that is
able to express emotion and recognize
cues from interaction with humans.
1987
Ernst Dickmanns and collaborators equip
a Mercedes van with video cameras,
microprocessors, and other electronics to
demonstrate autonomous driving at almost 60
miles per hour. After much other AI research falls
short, DARPA cuts the projects budget.
1979
A backgammon program developed by
Hans Berliner defeats the reigning world
champion in a match, the rst time a
computer has defeated a champion-level
competitor in an intellectual game.
1966
Joseph Weizenbaum demonstrates ELIZA,
the worlds rst chat program, which is able to
converse using a series of preprogrammed
phrases, sometimes to comic effect.
2004
2011
DARPA sponsors its rst Grand
Challenge, which pits research teams
against each other to design driverless
vehicles capable of independently
traversing the Mojave Desert.
IBMs Watson defeats Jeopardy! champions
Ken Jennings and Brad Rutter in a televised
two-game, three-night face-off that ends with
the computer amassing more than three times
the winnings of its human competitors.
1997
2012
IBMs Deep Blue chess computer
avenges its prior defeat to world
champion Garry Kasparov in a tense
match commemorated in a documentary
lm, The Man vs. the Machine.
A team from Geoff Hintons lab wins
the ImageNet Large Scale Visual
Recognition Challenge with deeplearning software that could within ve
guesses identify a thousand types of
objects about 85 percent of the time, a
huge improvement in accuracy.
1984
2014
Douglas Lenat begins the Cyc project, an
ambitious attempt to create a commonsense knowledge base that can
eventually become self-educating. Little
progress is seen for decades.
Google acquires DeepMind
Technologies, a small London-based
startup focused on deep learning,
a relatively new eld of articial
intelligence that aims to achieve tasks
like recognizing faces in video or words in
human speech.
1972
AI takes a hit when philosopher Hubert
Dreyfus publishes What Computers
Cant Do, a manifesto challenging the
predictions of AI researchers, and scientist
James Lighthill pens a pessimistic review
of progress in AI research in the U.K.,
leading to funding cuts.
2016
Googles AlphaGo decisively beats the
world champion of the complex board
game Go.
Improving Translation
Translation quality
15
By adding industry- and company-specific words, sentences,
and translated documents, Microsoft has created tools that
help users improve translation.
Translation is
done without
customization
Focused
by type of
content to be
translated
Microsoft Translator
Models
Standard
Category
Adds industry
or business
specific words
only
Dictionary
Translated
sentences
added to
improve
tone and
terminology
1,000-5,000
Parallel Sentences
With this many
translated
sentences to
work with,
system begins
learning new
terms in their
business
context
5,000-50,000
Parallel Sentences
Source: MIT Technology Reviews AI Takes Off Business Report
MIT Technology Review, 2016. All Rights Reserved.
MIT Technology Review
[Link]
@techreview @techreview_asia
amea@[Link]