Alpenglo Digital’s Post

Ever wonder how machine learning really works? Usama Fayyad, Inaugural Executive Director of Institute for Experiential AI at Northeastern University, explains that it’s not just about advanced algorithms—it’s about using data. He highlights how AI gets smarter with the right data and computing power, but the challenge is capturing and making that data usable. Tap to watch and learn more about the vital role of data in AI! #MachineLearning #AI #DataScience #DataDriven #ArtificialIntelligence

Andrés Corrada-Emmanuel

Industrial scientist and developer focusing on robust AI systems and evaluation frameworks.

1w

Yes. Data is the real asset in AI/ML. Here is a demonstration that I was not able to make 2 weeks ago on how unsupervised evaluation of nearly error independent classifiers works. Previously, I had been using the UCI Adult dataset, a de facto standard dataset in the algorithmic fairness community. It has many problems, the biggest one that it is tiny - 64K rows. Last week I discovered the awesome "folktables" project in Github that surfaces the same US Census data used by the original UCI Adult set. Just one year of the dataset, 2018, has 3 million rows! This allows me to make cool histograms like this comparing my algorithm with the ground truth evaluations. I've had this algorithm for years, it is only when I had more data that I could see/show how well it does.

  • No alternative text description for this image

To view or add a comment, sign in

Explore topics