Share
Embed
Can Software Save Science?
Keynote presentation at AMS 2020 Symposium on Advances in Modeling and Analysis Using Python.
https://ams.confex.com/ams/2020Annual/meetingapp.cgi/Paper/370521
Ryan Abernathey
January 15, 2020
Transcript
35 slides
-
C a n S o f t wa r e
S av e S c i e n c e ? R y a n A b e r n a t h e y A M S 2 0 2 0 -
A b o u t M e ( v 1
) !2 • Physical Oceanographer, PhD. From MIT • Associate Professor at Columbia / LDEO • Output: papers • 45 Peer Reviewed Publications H-index of 16 • NASA New Investigator Award, Sloan Fellowship in Ocean Sciences, NSF Career Award … -
• Output: software • Core developer of Xarray • Core
developer of Zarr • Co-founder of Pangeo • Creator of lots of other software tools (pyqg xgcm, xrft, xhistogram, …) A b o u t M e ( V 2 ) !4 -
• Anxious • Competitive • Severe Imposter syndrome • Highly
stressed → Not fun to be around • (despite all the privilege in the world) A b o u t M e ( v 1 ) !5 “confounded face” -
• Relaxed • Having Fun • Great relationships with collaborators
• Highly psyched → Fun to be around A b o u t M e ( v 2 ) !6 “grinning face with smiling eyes” -
I s i t j u s t m e
? !7 -
a n e c d o t e 1 !8
Setting: Columbia Classroom, Fall 2018, Research Computing for Earth Science, Matplotlib Lecture -
a n e c d o t e 1 !8
Setting: Columbia Classroom, Fall 2018, Research Computing for Earth Science, Matplotlib Lecture -
a n e c d o t e 1 !8
Setting: Columbia Classroom, Fall 2018, Research Computing for Earth Science, Matplotlib Lecture -
a n e c d o t e 1 !8
Setting: Columbia Classroom, Fall 2018, Research Computing for Earth Science, Matplotlib Lecture -
a n e c d o t e 1 !8
Setting: Columbia Classroom, Fall 2018, Research Computing for Earth Science, Matplotlib Lecture -
S c i e n c e h a s
B i g P r o b l e m s !10 http://blogs.nature.com/news/2014/05/global-scientific-output-doubles-every-nine-years.html -
S c i e n c e h a s
B i g P r o b l e m s !11 https://www.nature.com/news/1-500-scientists-lift-the-lid-on-reproducibility-1.19970 -
S c i e n c e h a s
B i g P r o b l e m s !12 -
A r e w e i n a r at
r a c e ? !13 -
A r e w e i n a r at
r a c e ? !13 A rat race is an endless, self-defeating, or pointless pursuit. The phrase equates humans to rats attempting to earn a reward such as cheese, in vain… The term is commonly associated with an exhausting, repetitive lifestyle that leaves no time for relaxation or enjoyment. -
A r e w e i n a r at
r a c e ? !14 Papers Scientists -
Pa p e r ≠ K n o w l
e d g e !15 What is the best way to communicate scientific knowledge in 2020? -
W h at i s a s c i e
n t i f i c D i s c o v e r y ? !16 New Data Inferences + Measurements (satellite, weather balloon etc.) Simulations, data analyses, equations, visualizations, interpretations. 99.9% of the time, this part is encoded in computer code. Old Data Paleoclimate discovery New proxy record Climate dynamics CMIP6 / ERA5 data analysis Theoretical GFD Equations and simulations -
W h at i s a s c i e
n t i f i c D i s c o v e r y ? !16 New Data Inferences + Measurements (satellite, weather balloon etc.) Simulations, data analyses, equations, visualizations, interpretations. 99.9% of the time, this part is encoded in computer code. Old Data Paleoclimate discovery New proxy record Climate dynamics CMIP6 / ERA5 data analysis Theoretical GFD Equations and simulations There are discoveries with no data. Are there discoveries with no code? -
P r o p o s a l !17 Scientists
should devote more effort to software and less to papers. Paper 1 Paper 2 Paper 3 Typical Ph.D. Output Current Proposed Data Product Software Paper 1 -
• Encourage all scientists to become responsible participants in the
open source ecosystem. ‣ Contribute to “upstream” tools on which their research relies (not just code; bug reports, docs, examples, tutorials, etc.) • Empower all scientists as software / data engineers ‣ Learn best practices for sharing, testing, documenting, and collaborating on research software ‣ Use these practices to transform how we share knowledge S p e c i f i c s !18 -
E C O S Y S T E M !19
aosp SciPy Credit: Stephan Hoyer, Jake Vanderplas (SciPy 2015) Project Project Project Project Project -
R e t h i n k i n g
S c i e n t i f i c O u t p u t !20 Paper data data code Status Quo Future existing data new data Code Docs -
Cloud computing environment C o n t i n u
o u s I n t e g r at i o n o f S c i e n c e ? !21 Example project: “Assessing Global Sea Level Rise” Analysis Code Data Data Docs Execute code; generate outputs “Sea level has risen 10 cm since 1992…” -
O P E N S O U R C E
C U LT U R E !22 vs. -
O P E N S O U R C E
C U LT U R E !22 vs. “SciPy is the best conference ever!” “I look forward to SciPy all year long!” -
• Open source culture is organized around community projects (e.g.
scikit-learn, metpy, etc.) • Individual achievement is not emphasized; Collaboration is essential • Progress is atomic, rapid, and public, with instant feedback and gratification (“agile”) • End result is a product, something others can really use and benefit from O P E N S O U R C E C U LT U R E !23 Pangeo Community Meeting (2019) -
• Open source culture is organized around community projects (e.g.
scikit-learn, metpy, etc.) • Individual achievement is not emphasized; Collaboration is essential • Progress is atomic, rapid, and public, with instant feedback and gratification (“agile”) • End result is a product, something others can really use and benefit from O P E N S O U R C E C U LT U R E !23 Pangeo Community Meeting (2019) -
• Open source culture is organized around community projects (e.g.
scikit-learn, metpy, etc.) • Individual achievement is not emphasized; Collaboration is essential • Progress is atomic, rapid, and public, with instant feedback and gratification (“agile”) • End result is a product, something others can really use and benefit from O P E N S O U R C E C U LT U R E !23 Pangeo Community Meeting (2019) -
• Open source culture is organized around community projects (e.g.
scikit-learn, metpy, etc.) • Individual achievement is not emphasized; Collaboration is essential • Progress is atomic, rapid, and public, with instant feedback and gratification (“agile”) • End result is a product, something others can really use and benefit from O P E N S O U R C E C U LT U R E !23 Pangeo Community Meeting (2019) • Mainstream scientific culture is organized around the individual scientist • Success is attributed to individual brilliance / hard work • Submit / revise / publish cycle takes years! (“waterfall”) • End result is a paper, impact measured in citations -
C a r e e r Pat h s !24
Software engineering skills are in high demand! Contributing to open source builds a portfolio that can open many doors. -
S u m m a r y !25 Proposed Data
Product Software Paper By placing software at the center of our scientific practice, we can… • Create more robust, reproducible, extensible scientific knowledge • Have more fun and feel more connected to a community • Train a better workforce and create new job opportunities