Written by Antje Farnier & Leo Paillier — September 18, 2017
This week in AI: start with a rich soup of data science knowledge topped with sweet maple syrup, then follow up with freshly baked TensorFlow with its TensorBoard reduction, and finish with some philosophical reflexion on our future sprinkled with some concrete applications. Don’t wait, dig in!
1- Join the team

Are you on of these people who get pumped up by AI but do not know how to switch from a passion to a full-time job? Don’t worry, we are here for you! With the great number of resources available online today, you don’t need a fancy Master’s degree or Ph.D. anymore. First things first, take a look at Tinniam’s journey which will provide you with a comprehensive list of courses to get in the thick of it. Once you feel you’re proficient enough, put your knowledge to the test using this neat data science interview simulator, and you should be ready to go. You’ll see, it’s not that hard to get a foot in that door! Oh, and more thing, a courtesy of David Fumo: a quick overview of the most influential Deep Learning key-players. It is always good to keep an eye on them.
2- Maple Syrup for everyone

Speaking of which, if you go through the list you’ll soon realize that Canada is home to a lot of them. This trend is not about to stop: Google and Microsoft have already set up headquarters in Montreal to capitalize on the thriving AI ecosystem that is developing there. It’s no surprise then that Facebook just announced that they join the club! Joëlle Pineau will take the lead of the latest FAIR lab that will focus on dialogue systems and reinforcement learning; this sounds a lot like messenger is about to get a chatbot update!
3- TensorFlow on Steroid

Google’s team has been hard at work this summer: they released TensorFlow 1.3.0 in mid-August which introduced new concepts to make your life easier. Unfortunately, Google lacks tutorials. It’s sometimes confusing to implement the latest updates; that’s why Peter Roelants at Onfido released a complete tutorial on estimators, experiments, and datasets to get you started right away!
But that’s not all. They’re back with a new version of TensorBoard, their monitoring tool that goes hand in hand with TensorFlow. The major announcement is that it comes with a new set of APIs to develop custom visualizations: before, you needed to create a GitHub issue on the TensorFlow repo and cross the fingers that the team would deem it interesting enough to implement it at some point. This process can be long, very long. Now, you can take matters into your own hands and build whatever you want! Definitely a step in the right direction.
4- The Future of AI’s Future

Okay, I admit it, we’re cheating a bit with this one; it’s already two weeks old, but it got such an impact on us that we’ve decided to incorporate it anyways. Artificial Intelligence, Artificial General Intelligence, Singularity, sometimes we get mixed up. Throw in all the predictions about how our world is going to change due to IA, how it’s already evolved, and you might end up completely confused by what’s real and what’s fantasy. That’s the exact issue Rodney Brooks, the founding director of MIT’s Computer Science and Artificial Intelligence Lab, is tackling in his latest essay. In his article, he does a complete review of the different biases when evaluating the impact of AI in our lives. It’s a bit lengthy, but it’s worth reading!
5- Real-life impacts
Despite the raging battle between the AI-will-kill-us-all and AI-is-our-BFF camps, we should not overlook the fact that machine learning algorithms already have tangible impacts on our lives; especially in fields where time or expertise are crucial factors. To name only a few recent examples: diagnosing diseases in plants or even in our hearts, discovering the origin of our universe, or more practically simplifying shopping. AI might not be humankind’s future, but you can’t argue against that it allows us to think about it, our future.