Past years’ State of AI reports have been fantastic and this one is no different. This nearly-200-slide deck is probably the single most worthwhile read to understand the progress the industry has made in the prior 12 months. Topics are:
Research: Technology breakthroughs and capabilities.
Talent: Supply, demand and concentration of AI talent.
Industry: Areas of commercial application for AI and its business impact.
Politics: Regulation of AI, its economic implications and the emerging geopolitics of AI.
Predictions: What we believe will happen and a performance review to keep us honest.
Set aside like two hours if you want to go through the whole thing; I’m currently ~50% through. Seriously, though, worth it.
Another “state of the world” post that comes out every year (why is this month the big month?!). Matt’s summary is fantastic, and if you’re looking for a data industry logo-jungle-infographic this is definitely the most authoritative one in the industry.
The success of a data science project depends on the business outcome it inflects, not the ROC. When evaluating a new data science endeavor, it’s critical to measure how the product impacts actual business KPIs.
The above sentence is the central tenet of the post, and it’s one that I hard-agree with. Every new data scientist should read this post.
This is slightly off the beaten path, but really such an important topic. It turns out that talking about data (and other technical topics!) is quite challenging—IMO the main barrier to achieving successful project outcomes is typically communication, not technical competence.
This article is a short guide for you to run more effective meetings where you’re discussing technical topics. Doing this well is hard, and not something you can just pick up from reading a post, but this one captures some of the fundamentals nicely.
This is maybe not that profound of a point, but I really really loved this infographic for a very specific reason. Often, data people spend most of their time focusing on analysis, on modeling. This feels like the highest-leverage use of our time, it’s certainly where we get to demonstrate our cleverness.
But the above analysis is dead-freaking-simple. What’s clever about it is that it asks a great question, and goes and gets a really perfect dataset to answer that question. Really, that’s the magic of the trade. Can you ask good questions? Can you be resourceful in sourcing data to answer them?
The market for solutions for developers will get ten times larger and hundred times better.
This is another slightly unusual thing for me to link to, as it’s focused on software engineering and not specifically data. But I’m an avid reader of Erik’s blog and loved this post.
His statement (above) is absolutely true. The flywheel of tooling for software engineers is spinning fast and it’s only going to spin faster. As such, it is an absolutely fantastic time to be someone who writes code for a living—your tools are increasing your individual capabilities explosively. This is one of the reasons that I asserted, four years ago now, that analytics is a subfield of software engineering. Modern humans process data using computers, and the most expressive way of instructing a computer is to write code.
If you’re writing code to with data, you should think of yourself as a software engineer.