Grande conférence AQT avec Yoshua Bengio (Directeur du MILA et professeur à l'Université de Montréal) sur l'intelligence artificielle et l'apprentissage profond animée par André d'Orsonnens, PDG de Druide Informatique.
5. AI: The Upcoming Industrial Revolution
First industrial revolution:
• Machines extending humans’
mechanical power
Upcoming industrial revolution:
• Machines extending humans’
cognitive power
• From the digital economy to the
AI economy
• Predicted growth at least 25%/yr
• All sectors of the economy
6. A new revolution
seems to be in the
work after the
industrial revolution.
And Deep
Learning is at
the epicenter of
this revolution.
7. Initial breakthrough in deep learning
A Canadian-led trio at
CIFAR initiated the deep
learning AI revolution
• Fundamental
breakthrough in 2006:
first successful recipe for
training a deep supervised
neural network
8. AI, Machine Learning, Deep Learning
• Putting knowledge into computers
• Much knowledge is intuitive, uncommunicable
• Machine learning (ML) approach to AI
• Deep learning: a ML inspired by brains, learning
multiple levels of representation, different levels of
abstraction
8
10. Data is the New Oil
• Because AI is based on ML, successful AI
applications require DATA – lots of data
• The first step in any project:
– what data is available and what data is
needed, do we need to collect more, do we
need to label it?
12. 2012-2015: breakthrough
in computer vision
• Graphics Processing Units
(GPUs) + 10x more data
• 1,000 object categories,
• Facebook: millions of faces
• 2015: human-level
performance
13. 100%
90%
80%
70%
74.2
88.3
93.3
96.4
2012 2013
~ level of human
accuracy
94.9%
Use of
Deep Learning
over
Conventional
Computer Vision
Google
NYU
84.7
201520142011
U.Toronto
Microsoft
ImageNet Accuracy Still Improving
Top-5 Classification task
15. Still Far from Human-Level AI
• Industrial successes mostly based on supervised
learning
• Learning superficial clues, easy to fool trained
networks
• Still unable to discover higher-level abstractions
16. From Labs to Products
• Even if fundamental AI research were to
stop today, there would be at least 10 yrs
reaping the benefits of today’s science
• Scaling up efforts: more and better data,
engineering of models, computing power,
creative deployment of technology
17. 17
Through automation, Artificial Intelligence will cause major labor disruption
while creating new markets and opportunities
SOURCE: The Economist, BAML, Press Search
Low-Medium risk High risk
53% 47%
Existing jobs
in the U.S.
Likelihood of automation in the near future
100% = 138mn jobs
Industries impacted Examples of opportunities
Automotive &
Transport
▪ Autonomous vehicles to create a $87bn solutions market
Aerospace &
Defense
▪ Drone systems integration to create $82bn in positive economic impact and generate more
than 100,000 jobs
Financial
Services
▪ Robo advisors expected to have up to $2.2tn in AUM productivity gains from automation to
create $600-800bn in annual economic impact by 2025
Healthcare
▪ Global market for telehealth to reach $34bn
▪ Global market for medical robotics to reach $18bn
Agriculture ▪ Global agribot market to reach $16.3bn
Nearly half the jobs in the U.S. face high risk of automation
AI will create new markets and opportunities in connected industries
PRELIMINARY
18. Embracing AI in industry:
Not just a question of
competitivity,
a question of survival
19. Keys to the Success of Montreal as
an AI Hub
• Critical Mass of Talent
• Over 160 deep learning scientists in
Montreal’s universities, soon 200
• International visibility of a productive
fundamental research group
• Generating many graduate students
• Leadership, presence in the community
• Visible & substantial government investment
• VCs willing to invest outside of traditional
techno-hubs
20. Why Invest in AI Innovation: create the value
here to compensate for labor disruptions
• Advances in AI will disrupt the job markets, with some studies
estimating that half of the jobs are at risk over the next
decade.
• The new markets for products and services derived from AI
will also correspondingly create immense wealth.
• To compensate for the disruption (social safety net, re-
training, etc.) we need enough of the wealth creation to
happen under our tax base.
21. Entrepreneurs & AI
• Start thinking about how ML could be exploited to
improve current services & products, and create new
products and services
• Need a data strategy
• Need talent (that is tough), but smart engineers can learn
the skills if given time and resources
• Connect with academic researchers and take advantage
of training events & precompetitive research at institutes
like MILA
22. Be Ambitious – Tolerate Risk
• Canada’s business has been too conservative
in the past, unlike our friends in The Valley
• Canada, and in particular Montreal, benefit
from a unique intellectual asset
• It could be turned into an economic asset to
build an AI ecosystem here, if all the actors
collaborate: governments, academia, SME,
start-ups, VCs
Editor's Notes
Advances in Colonoscopy, An Issue of Gastrointestinal Endoscopy Clinics
Doug rex 2015