How Democratized Artificial Intelligence Can Move Manufacturing to a New Evolution Pace
devices and sensors for a cheap
investment across networks and
geographies is making the equipment
data a gold mine to:
4. The evolution of affordable multi-
purpose
robotics
includes
new
operational technologies needing data
to run, 3D printers that cannot run
without files describing complex
products and connected tools that need
instructions on the fly to operate
correctly. For example, a connected
smart handheld drill 11 needs to get the
right information to apply the right
torque depending the part that must be
assembled.
High-speed
camera
controlling systems need to record and
process high volumes of images to
detect image pattern as fast as possible.
This is now possible using AI in the cloud
using TensorFlow® or Caffe® for
instance. 12
a. Understand good, bad and
changing usages of products
b. Understand the relationship
between equipment and their
environments
c. Identify some common, normal
and abnormal patterns.
2. The
mobile
and
high-speed
communication networks are available
for a cheap price (1Gbps at $158/month
– Public Cloud pricing), making the cost
of moving the data affordable to place it
in cheap storage systems in the cloud
(Data storage goes from $1M to 2 cents
per gigabyte) 10 ; almost 6 cents in the
cloud
(including
the
needed
infrastructure for the storage system).
These digital and operational technology
revolutions are starting and will provide
enough data to use AI as an enterprise and
on a large scale. Of course, AI can be used
with only local data for specific and
dedicated purpose. Large-scale data can
train AI algorithms to build a next stage of
learning analytics and make them
affordable. These can then be used locally in
a more accurate and efficient way to fix
3. Cloud offers inexpensive elasticity with
computing. There is no need to think
about size of data or the size of
computing power like it was necessary
10 years ago for on premise solutions,
which were limited due to the inherent
cost of infrastructure to run high
performance computing. Right now, the
cost versus the return on investment is
becoming obvious in most cases.
10 Computer
World 2017: https://www.computerworld.com/article/3182207/data-storage/cw50-data-storage-goes-from-1m-
to-2-cents-per-gigabyte.html?nsdr=true
11
https://blog.bosch-si.com/categories/manufacturing/2016/02/airbus-factory-future/
12
https://en.wikipedia.org/wiki/TensorFlow; http://caffe.berkeleyvision.org/
https://www.youtube.com/watch?v=cef7C2uiDTM (from 13: 20 to 16:14)
- 8 -
November 2017