INGENIEUR
Huawei's AI Portfolio
AI Application
HiAI Service
General APIs
Advanced APIs
ModelArts
HiAI Engine
Pre-integrated Solutions
MindSpore
TensorFlow
PyTorch
PaddlePaddle
…
Ascend-Tiny
Consumer Device
Ascend-Lite
Public Cloud
Ascend
Private Cloud
Ascend-Mini
Edge Computing
Framework
Chip
Enablement
CANN ( Compute Architecture for Neural Networks )
Ascend-Nano
Application
Enablement
Ascend-Max
IP & Chip
Industrial IoT Device
All Scenario
We’ve seen several technological revolutions since
AI-triggered change has just
18th century.
Each has
a huge impact
on a new
● ●
stream the
of new
advances
in had
technology
until
Only about
of retailers
have invested in
begun.
Finding
the 2% right
problem
organizational
structures,
processes,
and
workforce
GPT emerges. Nevertheless, it is important to keep is more or
deployed AI.
important
than devising
will change jobs and skills in a way that’s
● ●
in mind skills.
that But AI AI is
not a cure-all. No technology
About 5% of smart city implementations
a
novel
solution
quite different from the previous revolutions.
can solve every problem. We need to focus on
are using AI.
● ●
areas where
AI
can
create
the
most
value,
not
on
In 2017, roughly 10% of smartphones
Change can mean good news for some and bad news
Previous revolutions led to huge demand for
problems
that AI is not equipped to solve. Finding for others, especially
on the
market were equipped with AI
when the changes first start to
repetitive routine tasks, such as operating equipment
the right in problem
is
more
important
than
devising
capabilities.
emerge.
textile mills, and running car and phone assembly
● ●
a novel lines.
solution.
The supply and demand ratio of AI talent
AI will greatly boost automation in almost all
is just
Some people worldwide
might get excited
about 1%.
the new, once-
aspects of an organization. This means there will be
unimaginable
functions
that
AI
will
make
possible.
much less demand for jobs that handle repetitive,
The gaps between stellar
achievements and
These
people
will
feel
a
strong
urge
to
drive
large- forces that will
routine
tasks.
lukewarm adoption are the driving
Inspiring Gaps
scale
AI adoption.
And there forward.
will also be those
push
the industry
I find who
these gaps to be
Demand for data science jobs will keep rising, including feel anxious about underperforming AI projects, or
To get started, we need to take a look at where we very inspiring.
who worry about the reliability and security of AI
those for data scientists and data science engineers
are today with AI.
applications. These are the ones who will remain
with basic know-how in data science. The total number
The of
world
has seen significant achievements: uncertain about how to best use AI in the future.
these jobs will be much smaller than the number of
● ●
In
2017, 20,000 papers on machine Ten changes that will shape the future
jobs that handle repetitive, routine tasks.
learning were published.
If we look at the history of all GPTs, these reactions to
● ●
More
than
22 countries
have announced
a AI are
To all close
these gaps, we need the right technology,
very natural.
It’s likely
that organizations
will become
more
national
AI plan.
the right talent, and the right industry ecosystem.
diamond-shaped,
with AI systems taking the place of
are four
different
phases
along the ten
AI important changes
● ●
people at
the bottom,
they
handle
huge new There
Next,
I would
like
to discuss
In the 2017,
there
were where
more
than
1,100
productivity/adoption
We have
left the across all three
of repetitive and routine tasks.
that we have curve.
to work
on just
together
AI volumes
startups.
● ●
AI-related mergers and acquisitions totalled of these elements.
USD24 billion in 2017.
4
WINWIN ISSUE 32 12. 2018
● ●
First, we need to increase the speed of model
Venture Capitalist investment in AI reached
training.
USD14 billion in the same year.
Despite these incredible achievements, we With existing technology, training more complex
have also seen quite a few smaller figures that models often takes days, if not months. Successful
speak to lukewarm AI adoption in its early stages. innovation only happens after several rounds of
iteration. Slow model training seriously impedes
For example:
● ●
Only 4% of enterprises have invested in or application innovation. We believe that training
should be completed in minutes or even seconds.
deployed AI.
6
44
VOL
2019
VOL 77
55 JANUARY–MARCH
JUNE 2013