AI Without BS — Newsletter #2

Why read this newsletter? There is real progress being made in AI! But, with all the hype, it is difficult to distinguish reality from fiction. This curated newsletter will help you distill the real AI from the noise — stay informed with how AI is being used today to automate well-defined tasks, advancements in platforms/tools to apply AI, and progress in research in defining the art of possible! A weekly newsletter for your Sunday morning reading!

• Using re-enforcement learning is answer the question: Which of the many ads is more likely to appeal to a certain viewer? Finding the right answer to this ads placement question has a significant impact for Web 2.0 companies with thousands of possible ads and millions of visitors. The hands-on article covers “contextual bandit,” an upgraded version of the multi-armed bandit that takes contextual information into account.
• One of the growing use-cases of AI is for customer churn prediction. When combined with a proper sales and marketing process, the short-listed customers can be offered personalized offers and assistance. Check out this interesting hands-on article.
• AI use-cases in marketing and supply chain management — if you are looking for ideas to modernize, an interesting McKinsey report for a quick read.

“I would say, a lot of the value that we’re getting from machine learning is actually happening kind of beneath the surface. It is things like improved search results, improved product recommendations for customers, improved forecasting for inventory management, and literally hundreds of other things beneath the surface,” — Jeff Bezos, Amazon

• Google open-sourced the Language Interpretability Tool (LIT), an interactive platform for NLP model understanding: “As natural language processing (NLP) models become more powerful and are deployed in more real-world contexts, understanding their behavior is becoming increasingly critical. While advances in modeling have brought unprecedented performance on many NLP tasks, many research questions remain about not only the behavior of these models under domain shift and adversarial settings but also their tendencies to behave according to social biases or shallow heuristics.”
• IBM open-sourced tools for AI Data Labeling (as a part of Cloud Annotations project): “Faced with the daunting task of hand labeling thousands of images, developers are looking for an easier way to train their object detection models. Currently, it takes 200–500 samples of hand-labeled images for a model to detect one specific object. Autolabeling images speeds the process and gives developers back valuable time to work on other innovative projects.”

• MIT’s Liquid Time-constant Networks: Introduce time-continuous recurrent neural network models. “Instead of declaring a learning system’s dynamics by implicit nonlinearities, we construct networks of linear first-order dynamical systems modulated via nonlinear interlinked gates. The resulting models represent dynamical systems with varying (i.e., liquid) time-constants coupled to their hidden state, with outputs being computed by numerical differential equation solvers. These neural networks exhibit stable and bounded behavior, yield superior expressivity within the family of neural ordinary differential equations, and give rise to improved performance on time-series prediction tasks.”
• The Machine Learning Model Anonymization tool from IBM: “Traditional data anonymization algorithms don’t consider the specific analysis the data is being used for. What if a 10-year range of ages is too general for an organization’s needs? When these anonymization techniques are applied in the context of machine learning, they tend to significantly degrade the model’s accuracy. The tool anonymizes machine learning models while being guided by the model itself. We customize the data generalizations, optimizing them for the model’s specific analysis — resulting in an anonymized model with higher accuracy. The method is agnostic to the specific learning algorithm and can be easily applied to any machine learning model, making it easy to integrate into existing MLOps pipelines.”

Happy reading and learning! Follow on medium to get notified for the next weekly newsletter!