SLIDE 1
About me
- Cornell University, Ph.D. – CCSL Lab
- Founded Nutonian in 2011
- Eureqa = AI Software, >50,000 users
- Cited in > 500 medical, scientific and
Enterprise Using automation to extract meaning from data Michael - - PowerPoint PPT Presentation
Enterprise Using automation to extract meaning from data Michael - - PowerPoint PPT Presentation
Future directions of AI in the Enterprise Using automation to extract meaning from data Michael Schmidt, Ph.D. About me Cornell University, Ph.D. CCSL Lab Founded Nutonian in 2011 Eureqa = AI Software, >50,000 users
SLIDE 2
SLIDE 3
“Computer Program Discovers Laws of Physics”
–New York Times
Schmidt M., Lipson H. (2009) "Distilling Free-Form Natural Laws from Experimental Data," Science, Vol. 324, no. 5923, pp. 81 - 85.
Nature News
SLIDE 4
- 5
- 4
- 3
- 2
- 1
1 2 3 4 5 5 10 15 20 25
y = x2 x y
- 5
- 4
- 3
- 2
- 1
1 2 3 4 5 5 10 15 20 25
SLIDE 5
- 6
- 4
- 2
2 4 6 0.2 0.4 0.6 0.8 1
y = 0.02 x cos(4 x) + 1/(1 + exp(-4 x)) x y
- 6
- 4
- 2
2 4 6 0.2 0.4 0.6 0.8 1
SLIDE 6
SLIDE 7
7
SLIDE 8
The world obeys mathematical relationships – from physics to business operations Modern AI can deduce these hidden patterns automatically from data Machine Intelligence
SLIDE 9
Test and Find Structure
k1 θ2 – k2 ω1
2 – k3 ω2 2 + k4 ω1 ω2 cos(θ1 – k5 θ2)
+ k6 cos(θ2) + k7 cos(θ1) – k8 cos(k9 θ2) – k10 cos(k11 – k12 θ2) k1 θ2 + k2 k1 ω1 ω2 – k2 cos(θ1 – θ2)
Accurate Simple Complex High error
SLIDE 10
Model search
Population
- f Models
Better Models Variation Error Metric
Complexity Parsimony / Simplicity Model front Error
SLIDE 11
The science under the hood
High Error Simple Accurate Complex Optimal Solutions
SLIDE 12
Robot Scientist
Schmidt M., Lipson H. (2009) "Distilling Free-Form Natural Laws from Experimental Data," Science, Vol. 324, no. 5923, pp. 81 - 85.
Algorithms distill laws of physics from chaotic systems (published in Science 2009)
SLIDE 13
Getting the right result
Neural networks Evolutionary Search Computational Effort Test-set Accuracy
SLIDE 14
Massively Parallel
Search Kernel
Computation tests billions of independent models on the data
Search Kernel
- Low bandwidth -- transferring solutions
- High latency -- no control flow dependencies
Compute Server 1
Search Kernel Search Kernel
CPU Cores
Search Kernel Search Kernel
Compute Server 2
Search Kernel Search Kernel
CPU Cores
Search Kernel Search Kernel
Compute Server N
Search Kernel Search Kernel
CPU Cores
...
SLIDE 15
- Predict finish positions of the 2016 Kentucky Derby
- Expose relationships between running style, speed,
and trainer record
- Predicted winner, and 4 out of top 5 horses
– Winning Exacta (30:1 odds), – Winning Trifecta (87:1) – Winning Superfecta (542:1)
Machine intelligence in action
- 1. Nyquist
- 2. Gun Runner
- 3. Exaggerator
- 4. Creator
- 5. Mohaymen
- Standardized live odds probability
- Speed over the past two races
- Post position
- Racing style
- Track conditions
http://performancegenetics.com/machine-learning-algorithm-crushed-kentucky-derby/
SLIDE 16
Example
SLIDE 17
Confidential and Proprietary. 17
Demand forecasting for pharmaceuticals
7/21/2016
SLIDE 18
Confidential and Proprietary. 18
Optimizing crop yield
7/21/2016
SLIDE 19
Confidential and Proprietary. 19
Determining causes of customer churn
7/21/2016
SLIDE 20
Twitter: @Nutonian Blog: http://blog.nutonian.com Michael Schmidt Founder & CTO michael@nutonian.com
Conclusions
www.nutonian.com
- Machine intelligence extracts meaning from data
- Some companies employing machine intelligence today
- Many new applications ahead of us