Can you believe that 1/4th of the year is almost over? It’s going by so fast that we flew through February which brings us to our “March” Newsletter! As always, a few notables from the Alectio side of the world, followed by some selects from the Alectio reading playlist. Enjoy!
WHAT WE’VE BEEN DOING
What computing forgetfulness on unseen data taught us
When a model changes its prediction, it’s telling us something about our data. The question is what exactly it’s telling us.
An interview with our founder about active learning
Jennifer Prendki stopped by to speak with Pete Soderling of Data Council. It’s a wide-ranging discussion that covers everything from the rudiments of active learning to its challenges and where we’ve seen the biggest successes.
9 Reasons Why Active Learning Isn’t Widely Adopted
Active learning is an elegant solution to a lot of the most pervasive issues in ML today. Yet, it is both rarely used in the ML/AI industry and commonly misunderstood. Alectio CEO Jennifer Prendki explains reasons why it is not being adopted – and also, helps explain how you can actually integrate active learning into your own ML lifecycle.
An introduction to Data Prep Ops
While many of us have heard of MLOPs, fewer are familiar with the idea of Data Prep Ops. In this whitepaper, we explain what Data Prep Ops is and how you can implement smart practices at your organization. Download at:
WHAT WE’VE ENJOYED
An AI that makes mistakes on purpose
Unlike Deep Blue, this chess algorithm tries to anticipate human error. Could it be used to help fight medical misdiagnosis and problems a bit more important than losing your rook?
Who should stop unethical AI?
We all know AI will be transformative. But what happens when that transformation makes society worse, not better?
An AI that keeps learning after it’s trained
Fascinating research that looks to solve a pervasive ML problem: the need to constantly retrain models that frequently require novel inputs.
An AI that draws abstract concepts
Remember when you were asked to draw “Courage” in art school? Well, now there is an AI to do that. Multimodal neurons just like the ones found in the human brain allow those next-gen models to cluster abstract concepts into high-level themes, just like we do!
Space exploration with AI
We just landed on Mars (again)! So, let’s talk about this NASA project leveraging a less intimate approach to exploration.
Have something worth reading that we’d find interesting? Want a deeper dive into Alectio? Give us a shout at firstname.lastname@example.org. Until next month, take care.