DAYMN — 5 Sep 2021
Data Articles You Might Need. Subscribe & follow here!
I read a lot of informative and thought-provoking articles every week, and share them immediately in a piecemeal fashion, with friends and colleagues.
One, it is hard to find these articles again — when you need them — despite using a combination of Read-Later tools. And second, the nature of these fleeting shares are quite ephemeral making it hard to retain the insights in this age of information overload.
So, I thought it would be a better practice to batch and share the top 5 articles I read every week along with a short write-up on what was interesting in these articles.
Here are the top 5 articles from this past week — please do share your feedback & thoughts!
1. Analytics Engineering — An Upcoming Discipline
Analytics Engineering is a fairly new role, being recruited by Product companies, focussed on helping their analysts be more efficient with their work. While the job descriptions make them sound like a composite of a data engineer, data modeller and a bit of analyst themselves, this article from Spotify Insights does a great job of explaining how and why it all started.
https://medium.com/spotify-insights/analytics-engineering-at-spotify-f165180a6722
Call out? … wanted to find someone who had the experience of working in analytics and data science and understood the joys and the pains of that job, who was passionate about data quality, documentation and SQL efficiency, who was better at data engineering than most data scientists but who didn’t .want to be a data engineer and who could coach others in these skills.
2. Good Lesson About Perils of Machine Learning
While Machine Learning and AI are often portrayed in Elon Musk-esque or some dystopian SkyNet terms, there are sometimes issues with these algorithms which are often not thought through. Github’s CoPilot AI which was designed to make coding easier and more accessible, turned out to have a fatal flaw — it was built of top of a dataset, that was not exactly copyright-free.
The result? Someone’s hard-thought, sweat-and-tear drenched logic stored on a GitHub repo could now be freely used by anyone using CoPilot AI. Harmless mistake or intentional design flaw?
https://medium.com/geekculture/githubs-ai-copilot-might-get-you-sued-if-you-use-it-c1cade1ea229
Call out? This is the fundamental problem with Copilot. It’s impossible to tell what code it has figured out by itself and what code it has downright copy-pasted from a different source. Another user on Twitter that gathered much attention called out the software for being a means to launder open-source code into commercial works.
3. Understanding AI Ethics — Time to Listen Up
A good segue from the article above is the need to understand the implications of AI Ethics, in a bit more detail. Often times, it is thought of as a theoretical subject relegated to a cursory conversation. But with more algorithms pervading our daily spaces, it is all the more imperative to first listen and then have a honest conversation about ethics in AI.
This wonderfully curated playlist from fast.ai is a great place to start
Call out? Overreliance on metrics is a core problem both in the field of machine learning and in the tech industry more broadly. As Goodhart’s Law tells us, when a measure becomes the target, it ceases to be a good measure, yet the incentives of venture capital push companies in this direction. We see out-of-control feedback loops, widespread gaming of metrics, and people being harmed as a result.
4. Building generalised architecture for AI problems
With the recent news about Dojo from Tesla, the race to build custom architecture to solve massive AI problems has become more specialised. In this wider context, this amazing contribution from DeepMind is a great start. Perceiver IO attempts at building a generalised & versatile architecture that can take in diverse data inputs to solve for a complex real-world problem, which never comes with only one type of data inputs. Great work!
https://deepmind.com/blog/article/building-architectures-that-can-handle-the-worlds-data
Call out? Perceiver IO fixes this problem by using attention not only to encode to a latent array but also to decode from it, which gives the network great flexibility. Perceiver IO now scales to large and diverse inputs and outputs, and can even deal with many tasks or types of data at once. This opens the door for all sorts of applications, like understanding the meaning of a text from each of its characters, tracking the movement of all points in an image, processing the sound, images, and labels that make up a video, and even playing games, all while using a single architecture that’s simpler than the alternatives.
5. I have to vs I choose to — The Essence of Essentialism
I have been reading this amazing book about the power of choice, the need to pick one aspect of your lives to hyper focus and the benefits of doing just one right thing really well. Greg McKeown’s Essentialism: The Disciplined Pursuit of Less is a great book that attempts to teach us, “Only once you give yourself the permission to stop trying to do it all, to stop saying yes to everyone, can you make your highest contribution towards the things that really matter.”
Call out? Half of the troubles of this life can be traced to saying yes too quickly and not saying no soon enough.
Have a wonderful week ahead everyone, hope at least one of these articles is exciting reading material for you, and made you think for a moment!
