Recently, I came across a Ted Talk regarding super-chickens. The idea of super-chickens seems like a great idea. In modern societies, the world pushes to have more growth, high competitive drive, and best performers. The ideal of meritocracy is discussed many times in the world of technology. Shouldn’t top companies be run by top employees?
Ms. Margaret Heffernan outlines a Purdue study conducted by William Mure, an evolutionary biologist. The study started with a flock of average producing chickens. The superflock was generated by selecting the highest productive chickens from the average flock to the test group. After each successive generation, the highest productive chickens were again selected from the average group and transferred to the superflock.
After six generations, the average producing flock were more productive than before. However, the test group, the superflock, pecked each other to death. After six generations, only three super-chickens remained. This experiment calls to attention employee selection, management, and evaluation. Specifically, suppression of overall productivity results in the success of super-performers depends upon the failure of others.
I think these types of experiments offer insight to better management of employees. Although the Ted Talk does not name any names, the recommendations are in clear conflict to generally accepted practices of annual reviews, ranking systems, and firing “underperformers.” Even GE, who popularized firing the “bottom” 10 percent annually, is starting to move away from this practice.
Ms. Margaret Heffernan also outlined another social experiment from MIT. Results show more successful groups have more women in the group. If we were to be more “scientific,” shouldn’t we trust research more than an opinionated manifesto by James Damore? Should we overlook or tolerate misbehavior of “high performing” managers or “super stars?” Other questions come to mind regarding this research compared to singular ideology, e.g. meritocracy. My intention is not to settle this discussion in the post, but these types of research should help us rethink common management practices.
After getting a little taste of what data science entails though an online course, I decided to delve deeper into data science by signing up for a 12-week course “bootcamp.” Some schools also call these courses “immersives.” At times while attending the bootcamp, I felt overwhelmed during the lectures and felt lost during the code-alongs. Here are some advice that I would give to my younger self before taking this delightful and sometimes dreadful journey.
Learn some basics of programming. The curriculum will provide details on the main programming language focus. There are tons of resources out there for free that can help you get a jump start on learning a language, if you do not know one already. Although the bootcamp prepares you in many ways, I have found that I will get way behind, when I am not familiar with some of the basic commands. I have to search online what I am trying to accomplish. Good foundation on structure and concepts of object oriented programming can be beneficial. You may want to print out some cheat-sheets before the class begins.
Follow people on twitter. I find that developers are very friendly and active twitter users. Also, following people has led me to follow other people and organizations that provide me useful information. Here are some people that are fairly active on twitter:
Listen to some podcasts. This allows you to be aware of different views and discussions, becoming more familiar into the world of data science. It seems to me that this field is rapidly changing. It is not necessary to listen to everything. Some of podcasts that I listen to are: Not So Standard Deviations, Hanselminutes, Data Skeptic, and Podcast.__init__.
Get familiar with some of the environments that you will be working on. The syllabus provides you the opportunity to look-ahead. That way, you do not need to wait for the instructors to teach you. Most likely, they will encourage you to learn more on your own. They will teach you enough to be dangerous, but the onus is on the students to really investigate into the fine art of data science.
Install as much software as early as possible. Installing software has been a major nuisance and time consuming. Sometimes, my machine ran into difficulties while installing machine learning packages. Other times, I would be able to install, but then the machine will not be able to run those libraries.
For example, my machine could not run pymc3 and multinomial naive bayes. It was frustrating that my computer would hang and I would just have to sit and watch the instructor program on the projector. Eventually, I was able to run both models on Azure notebooks. Guess where I was able to find about Azure notebooks? On Twitter, thanks to @DynamicWebPaige.
It is a beta version, running a Jupyter notebook instance on Azure. Majority of the libraries are installed, but not everything worked for me. However, it ran perfectly on the pymc3 and multinomialnb models.
There many ways to learn, including traditional routes like universities, MOOCs, books, and videos. But, being a little more prepared can go a long way in making the learning experience manageable. I hope this has been helpful to you. If you are heading in the direction of bootcamps, make the most of your time and good luck!