This document provides an overview of artificial intelligence (AI) in 3 main sections:
1. It explains what AI is, including how machine learning works using examples, neural networks, and deep learning. AI systems are able to learn on their own from large amounts of data.
2. It discusses some of the challenges with AI, including potential biases in data and systems that can cause issues if not addressed. AI may also replace some jobs but will create new types of work for humans.
3. It argues that humans still have important roles to play in supervising AI to ensure it is developed and used responsibly and for the benefit of humanity. Close collaboration between AI and human experts is needed to manage emerging
5. If you’re looking for AI tech:
Two very good ways to learn :
• DeepLearning.ai
(from Andrew Ng) (Stanford, Google Brain)
• Udacity.com
(from Sebastian Thrun) (Stanford, Google X)
7. Roald Sieberath
Training
Computer Science Engineer (UCLouvain)
Business : Solvay IACE,
Stanford EPGC + VC Unlocked, Louvain MBA
and business
Entrepreneur
Startup coach Investor
11. It *is* a paradigm shift (this time)
Classical computing
• Formal logic
• Results: defined
• Data = tables
• Program though languages
• Science is ± settled
• Coding talent is in demand
AI Computing
• Fuzzy logic
• Results: statistical chances
• Data = big, heterogenous
• Program through learning
• Science is progressing
• AI talent is very rare
12. AI will DECIDE for us
• Our insurance policies
• How our car drives
• Our HR assessment
• What ads we see
• Influence our purchase
• …
• “Social ranking” (China)
• …
13. It can SHAPE your career
• It will become a dominant strand of IT
• How much do you want to be part of it ?
14. What AI is not (until ~2030)
AGI
Artificial GENERAL Intelligence
(autonomous, self-conscious,… )
40. Beware the data pipeline
Don’t focus too much
on the algorithm
It’s only a modest part
of the total solution
AI feeds on DATA and PROCESS
GIGO :
Garbage IN
Garbage OUT
41. Reinforcement learning (RL)
Lots of applications in games :
AI playing game,
rewarded by score.
Learns by playing
thousands of games.
46. “New” AI :
Deep Learning Neural Networks (NN)
• NN were already there in the 1980’s / 90’s
• (cf. MNIST, Le Cun )
• ”Neural Networks winter” : 1990-2005
• NN revival :
• Geoff Hinton (UToronto) : Deep Belief Networks
• ...
• Deep Neural Networks
• Availability of GPUs
• (performance increase 50x over a few years)
https://www.rtinsights.com/gpus-the-key-to-cognitive-computing/