Insight. Applied.
By editor
It appears that the singularity is here.
Google Deepmind have created an adaptation of the DQN algorithm—which combines Q-learning with a deep neural network—to enable a computer to beat us at 31 Atari classics (including Space Invaders).
Any other task that AI could be applied to now seems so unimportant…
By editor
The Social and Behavioral Science Team, established by Barack Obama, has produced its first annual report. The unit applies insights from behavioral science in an attempt to make government more effective.
David Cameron, the UK Prime Minister, set up a similar unit in 2010 which has now been spun off as an independent entity.
Questions studied by the US team include
It’s all very “micro”, but maybe that’s where the gains are to be had. UK Olympic cycling coach Dave Brailsford brought home seven goal medals by improving everything tiny thing related to cycling by one percent. Sometimes the Devil is in the details.
Update: The 2016 report is now available.
By editor
Microsoft are developing an open-source distributed analytics platform—codename “Prajna”.
It’s apparently inspired by Spark, but will also make it easy for developers to deploy cloud servies that exploit the processing capabilities of the platform.
It’s written in F# and the project is hosted on GitHub.
By editor
Scala has gained prominence in the data science community due to the rapid growth of Spark. Functional languages are well-suited to data science work so, if you find Scala productive, you might be interested in looking at F#. There’s a project devoted to promoting F# as an effective data science language—FsLab.
Those with Scala experience who wish to dip a toe into the F# waters should have a look a Alfonso Garcia-Carot’s F# for Scala Developers presentation.
He even used F# to create the presentation. That’s dedication.
By editor
O’Reilly have published their 2015 Data Science Salary Survey.
SQL, Excel, R and Python remain the top tools. Spark (and, consequently, Scala) have been the big climbers—and people with these skills are getting paid well.
R remains popular, but might be an intermediate step to other tools. Those using Python and Spark are less likely to use R. This makes intuitive sense. R is a great statistical analysis tool, but developers tend to prefer more traditional software development languages. If you’ve invested in learning an “enterprise” development language you are likely to stick with it.