AI
vs
Climate Change
CCAI @
Humanitarian Reading Group - 31/01/2019
Background
(what this all all about)
CCAI @
There is still time. Let's act.
It may still technically be possible to limit warming to 1.5C if drastic action is taken now
Smith et al, Nature Communications, 15 jan 2019
How?
CCAI @
Visceralize Factual Knowledge
- Reason is not enough (we've tried facts for years)
- People need relatability
- Visualization and emotion enforce message
Influence Organizations
- Individual actions are not enough
- Can be counter-productive
- Governments need public support (and demands)
CCAI @
Executive summary
Yoshua Bengio & Jennifer Chayes
CCAI @
Vision
(long term)
CCAI @
An app.
Viral
Usable
Realistic
Accurate
Actionable
Compelling
CCAI @
Improve on other initiatives like:
CCAI @
Visualizations
· Geographically precise climate models to predict the probability of future events at a user-chosen location (e.g., floods, fires, hurricanes)
· Compelling user-controlled visualization
· Condition on
- climate change intensity
- governmental actions
- global behaviour
Actions
Write to your representatives
Compare political programs
Favor respectful companies
CCAI @
Coming up
(and what to do to help)
CCAI @
Use Generative Adversarial Networks
CycleGANs can learn to transform images even with no or very little matched pairs
Hopeful progress
CCAI @
Start small
Choose sub-goal:
- Fires
- Floods
- Draughts
- Snow
- ...
CCAI @
CycleGan Experiment with floods, Sasha Luccioni
https://github.com/sashavor/karmaGAN
Data
-
Governmental agencies (Canada + France)
-
Simulations
-
Physical
-
Video Games
-
-
Web
-
Timestamped-street view
-
Google Images scrapping
-
#10yearsChallenge
-
-
NGOs
-
Ouranos
-
GIEC
-
CCAI @
Want to help?
You like GANs
Help us train them, choose and process the data, explore hyperparameters etc.
You like models
Help us run accurate climate simulations to model likely events at a given location
You like people
Help us understand behaviors, belief propagation and have a wide impact
CCAI @
You like structures
Help us predict consequences of macroscopic change and suggest relevant actions
Reach
Out
Kris Sankaran
Post-Doc - MILA
sankaran.kris@gmail.com
Victor Schmidt
PhD student - MILA
vsch@pm.me
Ask & join the
CCAI Slack channel!
CCAI @
François Corneau-Tremblay
MSc. student - MILA
corneauf@mila.quebec
Sasha Luccioni
Post-Doc - MILA
SashaVLuccioni@gmail.com
Karthik Mukkavilli
Post-Doc - MILA
karthik.skmukkavilli@gmail.com