ISSUE 416 · December 13, 2022
The state of AI in 2022—and a half decade in review
The results of this year’s McKinsey Global Survey on AI show how companies are actually using AI and how that use has evolved over the past five years. Adoption has more than doubled since 2017 but success is highly varied. Here's a look at how the leading companies use AI, how they build their teams and how they manage to keep pulling ahead.
Talking About Large Language Models
We have entered an era when technology and philosophy intersect in interesting ways. AI capabilities are rapidly progressing and we're increasingly seeing these systems as possessing human-like qualities. This is a great paper that explores what's really going on.
Combine AI models with the Pinecone vector database to make your applications understand and act on what your users want… without making them spell it out. Pinecone provides the cloud infrastructure that makes vector search easy, fast, and scalable. Understand more about Pinecone and try it →
Tutorials, Projects & Opinions
Searching for Unintended Biases With Saliency
Machine learning models sometimes learn from spurious correlations in training data and when that happens, the model might develop unintended biases that could lead to mistakes. This interactive post explores a technique for detecting these biases by asking for an explanation of how models make decisions.
Simple ML for Sheets
Simple ML for Sheets is an add-on for Google Sheets that enables anyone to use machine learning without being an ML expert. This detailed tutorial shows how to use the add-on to spot abnormal values, train a model, make predictions, evaluate & understand a model, and export a model.
Understanding Convolutions in Probability
This post takes a look at how the mathematical idea of a convolution is used in probability using a simple and slightly ridiculous example.
Data Systems Tend Towards Production
Data teams have substantially larger influence than a decade ago. The surface area of what can go wrong has grown just as fast.
Types of data products
"Data as a product" has become a popular phrase this year but what does that actually mean? This is a great post about the three different types of data products and how to manage them. Each type has different customers, different problems to solve, and unique nuances to be successful.
Ever think of yourself as a fraud? It's easy to do, especially in the data world, where things move fast and there's so much to know. This is a great post about "imposter syndrome" and how to approach that little voice inside your head.
The Normcore Tech Conference - December 15
This online conference is happening this week and it looks amazing! There are a wide variety of talks starting Thursday at 07:30 EST and continuing well into the evening. Check the schedule for details.
Hiring? Join the Data Elixir Talent Collective and get regular drops of outstanding data practitioners and leaders who are open to new opportunities.