O'Reilly,
the premier source for insight-driven learning on technology and
business, today announced the results of its 2018 Artificial
Intelligence (AI) survey, "How
Companies Are Putting AI to Work Through Deep Learning." Focused on
deep learning, a technique used primarily for supervised machine
learning, the survey explores the adoption of tools and techniques to
build AI applications and the barriers that hinder business adoption.
Findings suggest that the democratization of AI and deep learning
applications will continue, as development tools and libraries improve.
However, the shortage of AI-trained engineers and developers will
persist. For example, while 54% of respondents indicated AI will play a
big role (35%) or essential role (19%) in their organization's future
projects, lack of skilled people was the number one bottleneck reported.
Other key findings of the survey include:
-
28% of respondents are already using deep learning, but a majority
(71%) have not yet started.
-
Of those already using deep learning, most are using it to make sense
of structured or semi-structured data or text.
-
13% cited computer vision, used to gain a high-level understanding
from digital images or videos, as the application of deep learning
they are most interested in.
-
20% reported lack of skilled people as a bottleneck, when compared to
other barriers, such as hardware and compute resources, data related
challenges, company resources and culture, and accuracy and efficiency
of DL models.
-
A majority (75%) responded that their company is using an in-house or
external AI training program.
-
73% indicated they have begun using deep learning software;
TensorFlow, an open-source software library for dataflow programming,
was the most popular AI-tool cited.
-
70% said cloud services are important to the application they're
building.
-
Only 8% of respondents believe that deep learning will not play a role
in their future projects.
-
Other techniques also factor into how respondents build AI systems.
For example, 45% of respondents are using or evaluating reinforcement
learning.
"Despite some claims that AI is over-hyped, these results show that we
can expect more companies to use deep learning to improve their own
products and services in the coming year," said Ben Lorica, O'Reilly
chief data scientist and AI Conference chair. "With that will come a
tremendous emphasis on training at every level - from college degree
programs to professional trainers - as businesses seek to develop the
deep learning skills of their own staff."
You can download the full survey results here.