This year The Algorithm’s AI In 2018: A Year In Review really struck a chord with me. It opens with a large infographic detailing the major events in AI this year.
The article goes through the very long list of incidents with this quote summarizing it best:
All these events pushed a growing wave of tech criticism, which focused on the unaccountable nature of these systems… To date, we’ve seen no real movement from Washington.
The Algorithm “Ai in 2018: A Year In Review”The question is, how did it come to this? Why did no one foresee what would happen if we don’t take responsibility for what we are developing?
And the bigger question is, how long will it continue?
Until there is a big(ger) push in legislation (California Consumer Privacy Act is a good start) to protect ourselves from misuse of our data, bias in AI and generally new threats in this new age we are living in, we are on the brink of even more disaster.
We are used to seeing IT companies as some kind of “good guy”, moving us forward, bringing us new conveniences, but at what cost?
Using data to answer interesting questions is what researchers are busy doing in today’s data driven world. Given huge volumes of data, the challenge of processing and analyzing it is a big one; particularly for statisticians or data analysts who do not have the time to invest in learning business intelligence platforms or technologies provided by Hadoop eco-system, Spark, or NoSQL databases that would help them to analyze terabytes of data in minutes.
The norm today is for researchers or statisticians to build their models on subsets of data in analytics packages like R, MATLAB, or Octave, and then give the formulas and data processing steps to IT teams who then build production analytics solutions.
One problem with this approach is that if the researcher realizes something new after running his model on all of the data in production, the process has to be repeated all over again.
What if the researcher could work with a MongoDB developer and run his analysis on all of the production data and use it as his exploratory dataset, without having to learn any new technology or complex programming languages, or even SQL?
No matter what you call it– you can’t dispute it’s convenient.
I can’t decide which one I want more. They’re both pretty cute.
So what is it, already?
It’s a Word document you can code in.
In all seriousness, it’s a web application, designed for use in the data sciences.
iPython is a sub-project under Project Jupyter (which supports a whole toooooon of languages, not just Python). Project Jupyter was started in 2014, so although its Jupyter notebooks now, a lot of people still call them iPython notebooks. And yes, it takes some time to get used to spelling Jupyter.
Let’s work our way down from the top.
And take a look at the menu…
I just want to go through things that I think are useful, because most of it is very self-explanatory. So let’s start with Jupyter’s time travel device.
“Save and Checkpoint” and “Revert to Checkpoint” are its own built-in version control system, which is more convenient than saving it before every major change.
That brings us to the Edit menu.
The Edit menu allows you to easily edit the metadata.
The rest of the menu bar is pretty easy to understand, which brings us to a fun feature: Widgets.
Widgets can make your notebook come alive by adding interactivity to your notebook. Here’s some examples:
An interactive scatterplot you can explore.
A 3-D visualization you can explore from all angles.
Adding these widgets and more (customized maps!) make it a lot easier to turn your ideas into engaging data visualizations with Jupyter.
Now for buttons…
The Save button is obviously there, but what is that little cross next to it?
It adds a cell to your notebook (shown below).
The thing that is unique about Jupyter is that you run your code directly in the application, with the output coming out all nice and neat in its little box.
Another useful feature is that you can run the code section by section so that if you (GOD FORBID!) have a mistake, you don’t have to run your environment all over again from the beginning.
Next are the handy yet dangerous Cut, Copy and Paste buttons.
They cut, copy and can paste your cells, but be sure to be saving and adding checkpoints in case you hit the cut button by accident.
The next buttons move cells up and down. Pretty easy to use.
The next buttons are pretty important. The first one runs your currently selected cell, the stop button stops your kernel and the reset button resets it.
The next menu formats your cells. There are three options: Code, Markdown and Raw NBConvert.
The next button opens up a handy little search menu of all the commands for Jupyter Notebook.
The next button is accessible two ways. Through this button:
And the View Menu:
They both open a menu with a couple of different options to view the Notebook. I’m not sure why it’s really there.
In previous posts I talked about SQL and NoSQL, and I want to go into a little more detail (while keeping it simple) what makes them different.
Scalability>>> Think making big things small. In SQL data is stored vertically (so typically all on one server- expensive!). NoSQL stores it horizontally (many servers==ok).
Schema>>> Technically schema means a representation of some model. In programming land, it is used to refer to a structure of a database. So think because you can’t see a database (at least I hope you can’t) you have to think how that structure is represented. In SQL, the schema is fixed, columns must be decided ahead of time, and you have to put data in every column. Remember that wine shelf? You can’t really be adding a new column to your shelf after you’ve built it…it will probably look like all the images when you google “shelf fail.”
I don’t know why, but this shelf is kind of cute.
Also, you have to put a bottle in every slot. Someone’s going to be a happy wine collector.
NoSQL deals with schema in a very different way. It just says “Nope.” and walks away. You can add (or leave out) anything you want, anytime you want. Now that’s flexibility.
Data>>> Finally let’s get to the data. In SQL all rows contain one specific entry. For example, in a row containing information about a bottle of wine you might have “Year”,”Location”,”Winery” etc. You can’t have two years for a bottle of wine, or two locations. In NoSQL, that’s A-OK. You can have two wineries (maybe it was a collaboration?) or no wineries. If that’s what you want.