2. IA: pourquoi et comment
Naviguer l’émergence
de l’intelligence artificielle
Sylvain Carle
Associé @ RealVentures
ARIM - MRIA
Avril 2018
3.
Mon nom est @sylvain
Je viens de l’internet.
J’ai déjà dit que j’étais
un coureur des bois numérique.
Il y a 20 ans je venais du futur.
J’espère que bientôt je viendrai du passé!
4.
Mon nom est @sylvain
Je viens de l’internet.
J’ai déjà dit que j’étais
un coureur des bois numérique.
Il y a 20 ans je venais du futur.
J’espère que bientôt je viendrai du passé!
6. The Internet is providing the fabric on which the
Information age is built and provides the network
infrastructure to turn software into massively scalable
platforms, whether centralized or decentralized
(blockchain) enabling new services, products,
experiences and business models;
Mobility is enabling 3 billion people to interact with
their information and with each other anywhere at
anytime and;
Connectivity added to other devices such as TVs,
watches, cars, drones, clothes, robots, homes, AR/VR
head mounted displays offers new ways to interact
with and experience the world’s information;
Continued improvements in Computing efficiency for
different types of workloads whether in the cloud or at
the edge is enabling new platforms and applications
to emerge such as robotics, VR/AR, AI, blockchain,
smart devices, etc.
The storage, usage and display (UI) of Big Data we
and our businesses generate enable faster, more
impactful, data driven decisions.
The digital revolution is now disrupting all aspects of
society including retail, transportation, education,
healthcare, financial services, work, real estate, energy,
government, human communication, manufacturing,
genomics, nanotechnology, etc.
La (r)évolution numérique touche l’ensemble des secteurs
7. The Internet is providing the fabric on which the
Information age is built and provides the network
infrastructure to turn software into massively
scalable platforms, whether centralized or
decentralized (blockchain) enabling new services,
products, experiences and business models;
Mobility is enabling 3 billion people to interact with
their information and with each other anywhere at
anytime and;
Connectivity added to other devices such as TVs,
watches, cars, drones, clothes, robots, homes, AR/VR
head mounted displays offers new ways to interact
with and experience the world’s information;
Continued improvements in Computing efficiency for
different types of workloads whether in the cloud or
at the edge is enabling new platforms and
applications to emerge such as robotics, VR/AR, AI,
blockchain, smart devices, etc.
The storage, usage and display (UI) of Big Data we
and our businesses generate enable faster, more
impactful, data driven decisions.
The digital revolution is now disrupting all aspects of
society including retail, transportation, education,
healthcare, financial services, work, real estate,
energy, government, human communication,
manufacturing, genomics, nanotechnology, etc.
La (r)évolution numérique touche l’ensemble des secteurs
8. A I
D ATA U I
C O N N E C T I V I T YM O B I L E
I N T E R N E T
S O F T WA R E
2 0 0 8 2 0 11 2 0 1 4 2 0 1 7
L’évolution du stack
9. Google acquired Deepmind for $650M, launched Alphago,
GPU cloud, AI datacenter optimization, tensorflow, imagecloud,
translate, self driving cars, Health, and training all engineers in
AI. Notable quotes: Sundai Pichai - “we will evolve in computing
from a mobile-first to an AI-first world”; Eric Schmidt –
“Machine learning will cause every successful IPO win in 5
years”, donated $4.5M to MILA in MTL for AI research;
Facebook open sourced AI framework, opened Messenger
to bots, acquired Wit.ai, newsfeed powered by AI, launched
Facebook AI Research (FAIR) lead by Yann Lecun, vision to
build an AI agent per user;
Amazon: launched Echo (home voice platform), Alexa
(intelligent personal assistant), AI Cloud, acquired Evi for $26M
in 2012), acquired Angel.ai for chatbots, working on drones;
Microsoft acquired Maluuba, launched GPU cloud, Cortana,
Microsoft AI Research, smart bot framework, invested in
Element AI and donated $7M to MILA in MTL for AI research;
Apple acquired Siri for $400M, recruited Ruslan
Salakhutdinov from Carnegie Mellon (ex-UofT) to lead AI
research team, started open sourcing AI research, publicly
stated importance of AI to success;
Others worth mentioning: Tesla, Uber (Acquired Otto
for $680M, Geometric Intelligence) and leading car
manufacturers with self-driving cars (GM bought Cruise for
$1B), Nvidia and Intel (Nervana for $250M) for AI
semiconductors, Salesforce (Metamind and Tempo AI
acquisitions, Einstein initiative), Baidu, Tencent, Alibaba,
Samsung (Viv), Netflix, IBM (Watson), Twitter (Magic Pony), GE
(Bit Stew Systems);
Every successful company will require applied AI research to win and the
market leaders are showing the way, acquiring 140 AI companies since 2011
including 40 in 2016 and hiring all the AI talent they can find…
12. Not for redistribution 12
Element AI as co-founded by Yoshua Bengio, (founding father of AI renaissance) JF Gagné (co-
founder of Planora, Chief Product officer at JDA Software), Nicolas Chapados (PhD, co-founded Apstat
with Yoshua, Planora and hedge fund) and Real Ventures to build the world’s leading applied AI
research lab and implementation platform to launch AI first solutions in partnership with large
corporations and innovative startups;
Element helps companies make money with AI from fundamental research, to applied research, to
application/solution development to implementation, monetization and/or joint ventures;
Element’s research lab is uniquely connected to the world best academic ecosystems including
MILA, McGill, Poly, UofT, UBC, Microsoft and Cortana Research through its fellowship program and
partnership with Microsoft;
Element completed initial round of founding co-led by Microsoft Ventures and Real Ventures in
October ’16;
Element provides Real Ventures with credibility in the AI space, access to its network in AI academia
and corporations, due diligence support, proprietary deal flow and applied AI research support for
We are recognized leaders of the Canadian AI
ecosystem as a result of co-founding applied AI research
14. 1. La capacité d’accumulation (collecte, entreposage)
2. La rétrospective (voir, comprendre ce qui s’est passé)
3. L’analyse des signaux en temps réel (aggrégation et alertes)
4. Pouvoir émettre des recommendations (données et actions passées)
5. Capacité de prédictions (avec haut degré de certitude)
6. Prescription et automatisation (pur numérique et instrumentation)
Le modèle de maturité des données massives (big data)
Inspiré de https://en.wikipedia.org/wiki/Capability_Maturity_Model
15. Le modèle de maturité des données massives (big data)
1. La capacité d’accumulation (collecte, entreposage)
2. La rétrospective (voir, comprendre ce qui s’est passé)
3. L’analyse des signaux en temps réel (aggrégation et alertes)
4. Pouvoir émettre des recommendations (données et actions passées)
5. Capacité de prédictions (avec haut degré de certitude)
6. Prescription et automatisation (pur numérique et instrumentation)
16. 1. IA générale (AGI) vs IA appliquée (narrow, specialized)
2. La définition de l’IA change tout le temps…
“John McCarthy, who invented the name Artificial Intelligence, noted, the
definition of specialized AI is changing all of the time. Specifically, once a task
formerly thought to characterize artificial intelligence becomes routine —
like the aforementioned chess-playing, or Go, or a myriad of other taken-for-
granted computer abilities — we no longer call it artificial intelligence.”
3. IA comme “intelligence augmentée”
Le contexte actuel pour de l’intelligence artificielle
http://cacm.acm.org/magazines/2012/1/144824-artificial-intelligence-past-and-future/fulltext
https://stratechery.com/2017/the-arrival-of-artificial-intelligence/
17. Le contexte actuel pour de l’intelligence artificielle
https://jods.mitpress.mit.edu/pub/issue3-case
19. Les nouvelles capacités par l’intelligence artificielle
Le point de bascule: quand la machine dépasse l’humain moyen dans la
reconnaissance et l’interprétation des signaux dit “intelligents”. Le
language, le monde qui nous entoure (prolifération de senseurs).
1. La lecture de textes (NLP sémantique).
2. La compréhension de la voix (NLP audio).
3. La reconnaissance visuelle (CV, classification).
4. Niveaux multiples d’interprétation et d’abstraction.
5. Apprentissage profonds, non-supervisé, par contre-exemples.
6. Mise en réseau et en comm
20. Les attributs qui propulsent l’intelligence artificielle
La “tempête parfaite”.
1. Publication ouverte. https://arXiv.org
2. Un graphe des chercheurs facile d’accès.
https://scholar.google.ca/scholar?q=deep+learning+paper
3. La plupart de librairies de code en logiciel libre.
4. Beaucoup de cours gratuits en ligne.
5. Une communauté accueillante (l’esprit pédagogue de Y. Bengio)
6. Instituts de recherches “plusse meilleurs” : CIFAR, MILA, IVADO.
7. Accès aux données de références et nouvelles données
24. Quelques liens (spécifiques à la recherche marketing)
1. AI: Friend or Foe? 5 Tips to Add Automation to Market Research
https://www.insightsassociation.org/article/ai-friend-or-foe-5-tips-add-automation-market-research
2. Faster, cheaper and more efficient: AI-powered market research is here
https://martechtoday.com/faster-cheaper-and-more-efficient-ai-powered-market-research-is-now-here-207596
3. How Artificial Intelligence Will Affect Market Research in 2018
https://knect365.com/insights/article/49d59f88-3255-4938-b9b7-2332169d828b/how-artificial-intelligence-will-affect-
market-research-in-2018
4. Artificial Intelligence Will Be a Disruptive Force In Market Research
https://www.martecgroup.com/artificial-intelligence-in-market-research/
5. 4 Reasons Why AI-Powered Market Research Should Be in Your Toolkit
https://aboveintelligent.com/4-reasons-why-ai-powered-market-research-should-be-in-your-toolkit-5aaabf77c423
6. The Future is Now: Leveraging AI in Market Research
https://www.insightsassociation.org/conference-session/leveraging-artificial-intelligence-colgate-
palmolive%E2%80%99s-innovation-research-CRC2016
25. Quelques liens (généraux)
1. Building an AI Startup: Realities & Tactics. http://mattturck.com/2016/09/29/building-an-ai-startup/
2. O’Reilly Artificial Intelligence Newsletter. http://www.oreilly.com/ai/newsletter.html
3. Highlights from the O'Reilly AI Conference in New York 2016. https://www.oreilly.com/ideas/keynotes-from-ai-new-
york-2016
4. The Competitive Landscape for Machine Intelligence (2016). https://hbr.org/2016/11/the-competitive-landscape-for-
machine-intelligence et https://www.oreilly.com/ideas/the-current-state-of-machine-intelligence-3-0
5. The US Administration’s Report on the Future of Artificial Intelligence. https://obamawhitehouse.archives.gov/blog/
2016/10/12/administrations-report-future-artificial-intelligence et Stratégie France IA - http://www.enseignementsup-
recherche.gouv.fr/cid114739/rapport-strategie-france-i.a.-pour-le-developpement-des-technologies-d-intelligence-
artificielle.html
6. 5 Big Predictions for Artificial Intelligence in 2017. https://www.technologyreview.com/s/603216/5-big-predictions-for-
artificial-intelligence-in-2017/
7. A Sneak Peek at the State of AI 2016. https://medium.com/swlh/a-sneak-peek-at-the-state-of-ai-2016-d5d079e0c4de
8. Big Data et Intelligence Artificielle (panel IVADO des startups IA de Montréal). https://www.youtube.com/watch?
v=g66X1lQpYZk&feature=youtu.be#t=72m49s
26. IA: pourquoi et comment
Naviguer l’émergence
de l’intelligence artificielle
Sylvain Carle
sylvain@realventures.com
ARIM - MRIA
Avril 2018