• The Economics of Data Businesses
  handful of business models dominate tech today: SaaS, marketplaces, e-commerce, on-demand, social networks and so on. Most of these business models have been studied widely, both their execution and their underlying dynamics...But there’s one notable exception: data businesses. Despite the fact that many of the largest and most dominant tech firms in the world are data businesses, there are not many resources on the what, how and why of this business model...This essay is an attempt to change that...

• Machine Learning from the Viewpoint of Investors
  In this podcast, we interview two investors who focus heavily on machine learning to get their take on the state of the machine learning industry today: Leigh-Marie Braswell at Founders Fund and Davis Treybig at Innovation Endeavors. We discuss their perspectives on opportunities within MLOps and applied machine learning, common pitfalls and challenges seen in machine learning startups, and new projects they find exciting and interesting in the space...

• Compute Trends Across Three Eras of Machine Learning
  Compute, data, and algorithmic advances are the three fundamental factors that guide the progress of modern Machine Learning (ML). In this paper we study trends in the most readily quantified factor - compute...Based on observations we split the history of compute in ML into three eras: the Pre Deep Learning Era, the Deep Learning Era and the Large-Scale Era. Overall, our work highlights the fast-growing compute requirements for training advanced ML systems...

A Message from this week's Sponsor:

Data Science Articles & Videos

• Textless NLP: Generating expressive speech from raw audio
  Text-based language models such as BERT, RoBERTa, and GPT-3 have made huge strides in recent years...There is an important limitation, however: These applications are mainly restricted to languages with very large text data sets suitable for training AI models...We’re introducing Generative Spoken Language Model (GSLM), the first high-performance NLP model that breaks free of this dependence on text. GSLM leverages recent breakthroughs in representation learning, allowing it to work directly from only raw audio signals, without any labels or text...

• Why You Should (or Shouldn't) Be Using JAX in 2022
  JAX hit the scene in late 2018...DeepMind announced in 2020 that it is using JAX to accelerate its research, and a growing number of publications and projects from Google Brain and others are using JAX. With all of this buzz, it seems like JAX is the next big Deep Learning framework, right?...Wrong. In this article we’ll clarify what JAX is (and isn’t), why you should care (or shouldn't, but you probably should), and whether you should (or shouldn’t) use it...

• New Podcast: Vanishing Gradients - a data podcast with Hugo Bowne-Anderson
  A podcast about all things data, brought to you by data scientist Hugo Bowne-Anderson. It's time for more critical conversations about the challenges in our industry in order to build better compasses for the solution space! To this end, this podcast will consist of long-format conversations between Hugo and other people who work broadly in the data science, machine learning, and AI spaces. We'll dive deep into all the moving parts of the data world, so if you're new to the space, you'll have an opportunity to learn from the experts...

• How Data Science Drives Private Equity
  Jaclyn Rice Nelson talks to Drew Conway about data-driven private equity, and why it is one of the most exciting places to apply data science...From investment sourcing to due diligence and analyzing post-investment data assets, the range of challenges is matched by the rich data and potential for enormous impact...including: a) How PE investors and PE-backed companies can use data to build a competitive advantage and b) Why PE is one of the most exciting places to apply data science...

• Red Flags to Look Out for When Joining a Data Team
  Looking for new data science opportunities in this heated market? Before you accept that offer, here are some red flags to beware of. While these are from the perspective of data science, they would also apply to most tech roles...

• Perspectives in machine learning for wildlife conservation
  Inexpensive and accessible sensors are accelerating data acquisition in animal ecology. These technologies hold great potential for large-scale ecological understanding, but are limited by current processing approaches which inefficiently distill data into relevant information. We argue that animal ecologists can capitalize on large datasets generated by modern sensors by combining machine learning approaches with domain knowledge. Incorporating machine learning into ecological workflows could improve inputs for ecological models and lead to integrated hybrid modeling tools...

• DeepMind: The Podcast - AI for science
  Step inside DeepMind's laboratories and you'll find researchers studying DNA to understand the mysteries of life, seeking new ways to use nuclear energy, or putting AI to the test in mind-bending areas of maths. In this episode, Hannah meets Pushmeet Kholi, the head of science at DeepMind, to understand how AI is accelerating scientific progress. Listeners also join Hannah on a [virtual] safari in the Serengeti in East Africa to find out how researchers are using AI to conserve wildlife in one of the world’s most spectacular ecosystems...

• EvoJAX: Hardware-Accelerated Neuroevolution
  EvoJAX is a scalable, general purpose, hardware-accelerated neuroevolution toolkit. Built on top of the JAX library, this toolkit enables neuroevolution algorithms to work with neural networks running in parallel across multiple TPU/GPUs. EvoJAX achieves very high performance by implementing the evolution algorithm, neural network and task all in NumPy, which is compiled just-in-time to run on accelerators...

• Introducing Hex Tiles: Large-scale spatial data prepped and ready for analytics in minutes
  We have something spatial to announce: Hex Tiles, a next-generation tiling system that gives data scientists the ability to easily unify diverse spatial datasets, conduct on-the-fly analytics, and quickly visualize and explore big data on a planetary scale. What’s more, this can all be done in a matter of minutes and within your browser through the Unfolded Platform...

• How to Think Less About Data Visualization
  Is there a way we can think less about the process of producing charts? Are there heuristics we can follow that make data visualization feel more like speaking our first language?...Thanks to the pioneering work of the late Leland Wilkinson, the answer to these questions is Yes...

Forum*

Jobs

• (Senior) Analytics Engineer - Fabulous - Remote
  
  Fabulous is a mobile app helping thousands of people every day to change their lifestyles by integrating healthy habits into their lives. Fabulous is using a behavioral economics lens to help everyone achieve their fullest potential. We work closely with researchers based at Duke University and our advisor is Dan Ariely, author of NYT bestseller Predictably Irrational. We are looking for an experienced Analytics Engineer to consolidate the Data Science team and lead the development and enrichment of our Data Pipelines. We have a modern Data-Stack based on Fivetran, dbt, BigQuery, Amplitude, Metabase...

Training & Resources

• TMLR - A New Open Journal For Machine Learning
  An alternative publication model for machine learning research...

• Is the Normal Curve Too Good to Be True?
  In this article, I’ll use charts and simulations to demonstrate how traditional methods that rely on normality can fail us. We’ll also look at alternative methods implemented in Python that are more powerful and accurate. All the code will be provided so that you can check the results for yourself...

• Include diagrams in your Markdown files with Mermaid
  A picture tells a thousand words. Now you can quickly create and edit diagrams in markdown using words with Mermaid support in your Markdown files...

Books

• Hands-On Machine Learning with scikit-learn and Scientific Python Toolkits

  
  Integrate scikit-learn with various tools such as NumPy, pandas, imbalanced-learn, and scikit-surprise and use it to solve real-world machine learning problems...
  
  For a detailed list of books covering Data Science, Machine Learning, AI and associated programming languages check out our resources page.
   

  ---

  
  P.S., Enjoy the newsletter? Please forward it to your friends and colleagues - we'd love to have them onboard :) All the best, Hannah & Sebastian