March 26, 2018
slides available at https://slides.com/chenevert
Math teacher at ISEN Lille
Computer Science department
the digital transformation?
la transformation digitale ??
in French: numérique (please)
Digital transformation is the change associated with the application of digital technology in all aspects of human society (Wikipedia)
Profound changes in industry
1760 – 1830
1870 – 1914
1980 – ?
2016 – ??
From quantitative to qualitative
Many other parameters in IT double every 1.5 to 2 years:
Exponential growth means change is happening fast (and will happen even faster)
thresholds are regularly attained
Evolution of computing power
With great power comes great responsibility
We have more computing power at our disposal during any given year than we had for the entire history of mankind up to the year before
What are we going to do with it??
An example: evolution of music recording and distribution
LP: 47 minutes of analog audio
1982: Compact disc
74 minutes of digital music
(~ 700 MB)
132 117 177 233 70 100 230 63...
10000100 01110101 10110001 11101001 01000110 01100100 11100110 00111111
1993: MPEG-1/2 Audio Layer III
Achieves x 10 compression (1 min = 1 MB)
But: requires computational power (for real-time decompression)
1997: File sharing
2001: MP3 players
128 GB ≈ 3 months of music
≈ 3200 LPs
2007: Streaming audio
We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run. (Roy Amara)
The Internet of Everything (Cisco)
The Internet was conceived as a decentralized network of computers
Blockchain: a decentralized and authenticated ledger
How to achieve trust and consensus in a network of unreliable nodes?
Overview (taken here)
To be able to add a block to the chain:
need to solve a computational problem (proof-of-work)
in a process called mining (rewarded by Bitcoins)
... rise of the World Computer?
essentially means: everything is now a computer
A "smart system" is usually composed of:
Every digital interaction generates information, often in the shape of metadata
Storage being cheap, this data is often stored for future data analysis / mining
Example: internet trackers
Google Flu trends: monitoring flu from search queries
Computing-as-a-Service: tasks are performed on remote server
Allows businesses to externalize and optimize their computing resources
France IA report
Reproducing high-level human-like cognitive activities:
In recent years: increase in computing power led to impressive achievements in machine learning using so-called deep neural networks
Idea: instead of implementing deductive rules (like an adult would learn a language), let the AI make up its own inductively (like a child would do)
Works quite well! And the funny thing is we don't really understand why
See: Stéphane Mallat, data scientist @ Collège de France
Object classification in real-time
Classification through training
Through training, the networks learns how to recognize
Once it has this understanding, it can be used "in reverse" for generation.
The trolley problem
Should we divert the tramway to kill a single person instead of 5?
Many possibilities and opportunities ahead as things continue to evolve.
You tell us!