O'Reilly
O'Reilly

AI adoption, data cascades, and what's wrong with ML

Email sent: Apr 26, 2021 4:04pm
Plus, how to process large datasets without running out of memory.
O'Reilly
 
Data & AI Newsletter
 

1. AI Adoption in the Enterprise 2021

Every year, we conduct a survey on AI adoption in the enterprise—how developers work, what techniques and tools they use, what their concerns are, and what development practices are in place. (Special thanks if you helped us out with that survey.) Every year, there are a few surprises. Mike Loukides shares the results here.

 

2. The dangers of data cascades

A recent Google research paper, “Everyone Wants to Do the Model Work, Not the Data Work” (covered in a previous issue) describes data cascades as “compounding events causing negative, downstream effects from data issues that result in technical debt over time” and notes that 92% of the data pros interviewed had experienced at least one cascade. This post looks at four data cascade challenges outlined in the paper and frames them in the context of everyday AI development.

 

3. AI for health care equity

Regina Barzilay, Fotini Christia, and Collin Stultz describe how AI can support fairness, personalization, and inclusiveness in health care.

 

4. Why is reliable metadata so important?

Sandeep Uttamchandani explains why reliable metadata is so important and how to approach common challenges in handling metadata.

+ Sandeep is the author of The Self-Service Data Roadmap. Read it on O'Reilly or buy it on Amazon or WHSmith.

 

5. What’s wrong with MLOps?

Preconceived notions (and a little hype).

 

Strata Data Superstream Series: Creating Data-Intensive Applications

In this live online O’Reilly event, you’ll gain insight into design and engineering best practices through interactive sessions and live coding demos. Join us to learn how to make the right decisions for your applications.

Learn more

6. AI and drug discovery: attacking the right problems

“Since no one can actually dock a billion virtual molecules into a protein target, how can we reduce the problem to something theoretically manageable without throwing away the answers we want? And how will we know if we have?” The biggest problems in AI are often not computational, but in homing in on asking the right questions in the right way. This may be especially true in computational drug discovery.

 

7. Bias is not just a data problem

“A surprisingly sticky belief is that a machine learning model merely reflects existing algorithmic bias in the dataset and does not itself contribute to harm...Algorithms are not impartial, and some design choices are better than others.” Understanding model design—and its effects—opens up new, less burdensome mitigation techniques.

 

8. How to process large datasets without running out of memory

This series of articles explains how to process larger-than-RAM datasets in Python, covering code structure, data management, pandas, NumPy, database querying, and measuring memory usage.

 

New release: Data Science on AWS

In this live online O’Reilly event, you’ll gain insight into design and engineering best practices through interactive sessions and live coding demos. Join us to learn how to make the right decisions for your applications.

Read it on O’Reilly
Buy it on Amazon
 
 
Share this newsletter
Tweet Share
 

Want your own copy of this newsletter? Sign up here.

 
 
Twitter Logo Facebook Logo LinkedIn Logo YouTube Logo Email Logo
 

This message was sent to -. You’re receiving this because you’re a customer of O’Reilly or you’ve signed up to receive email from us. We hope you found this message to be useful. However, if you’d rather not receive future emails of this type from O’Reilly, please manage your preferences or unsubscribe here.

Read our privacy policy.

O’Reilly Media, Inc. 1005 Gravenstein Highway North, Sebastopol, CA 95472 (707) 827-7000

Other emails from O'Reilly

O'Reilly
Nov 22, 2021
What Google, Apple, Shopify, and Reddit are doing with data
O'Reilly
Nov 22, 2021
Limited time! A year for $299.
O'Reilly
Nov 20, 2021
Better bug catchers, autocomplete code, and Python for finance
O'Reilly
Nov 16, 2021
How to perfect bar charts in Tableau
O'Reilly
Nov 15, 2021
MTTR, SLI, SRE, IBM, and lots of other letters
O'Reilly
Nov 9, 2021
Add AI to your mobile apps