Welcome to the December edition of our best and favorite articles in AI that were published this month. We are a Paris-based company that does Agile data development. This month, we spotted articles about Deep Learning, useful tutorials, and some more philosophical articles. We advise you to have a Python environment ready if you want to follow some tutorials :). Let’s kick off with the comic of the month:
The description of AlphaZero in the paper from Deepmind was quite impressive. However, I did not have the chance to test AlphaZero myself. Nor to see the source code of this wonderful go, chess, and shogi champion. Was the match between AlphaZero and Stockfish a fair one? Cultivating your critical spirit is important! This article explains the shortcomings of the paper from Deepmind. We should refrain from calling AlphaZero a revolution until Deepmind shows us more evidence!
AI is now reaching the stars! An inspiring work by Google and NASA scientists shows us that Neural Networks can find new exoplanets. The spatial telescope Kepler has now recorded enough data to train a neural network to spot planets. AI is coming to help process the vast amount of data and make use of weak signals that are difficult to exploit. The neural network worked on data from only 670 star system to find 2 new exoplanets. It will soon analyze 150 000 more star systems! We should brace ourselves for the incoming avalanche of new worlds that will soon be discovered by AI. The sky is the limit!
Humans are building increasingly smart AI systems. One day, we will probably build one that will be smarter than any human being in all domains. This AI will thus be able to do the same thing just like us: create an AI smarter than itself. That new AI will then create an AI smarter than itself… Soon, super intelligent systems will develop in the so-called intelligence explosion. 50 % of AI researchers still consider this at least as a serious possibility. Sorry for waking you up, but do you know that this dream (or nightmare), is in fact impossible? Don’t read this excellent article if you want to continue to believe in it!
If you want to become a data scientist, you should read this article. If you know someone who wants to become a data scientist, then share this article with him/her. It contains the Hidden Truth that will open your eyes to the reality of this job. Ideas will not transform into products overnight. Likewise, a Machine Learning model will not become a product without some software engineering work. The data scientist that built the model will likely be the one that does it. This article explains why it is important to invest in software engineering skills as a data scientist.
For the end of this year, we want to share with you this excellent summary about Natural Language Processing (NLP). Do you want to know the cutting edge achievements in text summarization and machine translation? Do you want to hear the story of how Radford et al. discovered the “sentiment neuron” by chance? Or maybe you are interested in specialized frameworks for NLP? If you have any interest in NLP you should read this article. You may also be interested to know that we are preparing a blog article explaining how to do Named Entity Recognition for non-English languages using NLKT!
Games are a great tool to measure the problem-solving skill of modern AI systems. AI mastered chess in 1997, and the game of Go last year. The next step is StarCraft II. This article explains the ideas behind this challenge. Best StarCraft II AIs are very weak at the moment, so there is still plenty of room for improvements. Want to participate in beating StarCraft II? Take a look at StarCraft II Learning Environment (SC2LE) from DeepMind and Blizzard!
Have you ever been struggling to find the right SQL request to solve a very simple business problem? This very practical tutorial explains a useful technique for writing simpler and more effective SQL queries. SQL Window Functions can help you solve several common problems. Learn a better way to calculate monthly growth or deal with duplicate data. A very nice tutorial!
Excel is used by so many people. The python package Pandas is so much powerful for data processing automation. Now join them together and you will benefit from the best of the two worlds! This tutorial explains all you need to know to combine effectively Pandas and Excel in a data workflow.
This month, we wanted to highlight a common Machine Learning technique. Principal Component Analysis (PCA) is a dimentionality reduction technique. It can be used to dramatically speed up ML algorithms and also for data visualization. If it is not already, why wouldn’t you put this technique into your toolbox? Discover how simple it is to make use of PCA with scikit-learn!
Dockers are containers that encapsulate an environment and make it easy to share or rebuild it. This article will likely persuade you that any Data Scientist should know the basics of Docker. It provides a very nice introduction to Docker so that you can get started in no time! From sharing your work to deploying systems in production, using Docker can help in many important tasks. By the way, did you read the article we published last month about setting up Tensorflow with Docker in minutes?
Build Your Own Cloud with Kubernetes and Some Raspberry Pi
Managing several Raspberry Pi can be a lot of work. This article will teach you how Kubernetes and Docker can help.
Edge Detection in Opencv 4.0, A 15 Minutes Tutorial
This tutorial will teach you, with examples, two OpenCV techniques in python to deal with edge detection.
Fast Custom KNN in Sklearn Using Cython
Let’s dive into how you can implement a fast custom KNN in Scikit-learn.