Wednesday, January 06, 2016

Pass the MOOCs -- Part 5 (Rumsfeld's Three) ... with fifth & final update

First post: 1/23/16 ... Last update: 4/26/16
In early 2002, a few months after the devastating attacks on Wall Street and the Pentagon on 9/11/2001, then Secretary of Defense Donald Rumsfeld was asked about a direct link between Saddam Hussein's government in Iraq and terrorist organizations. He responded with what became his most famous statement:
  • "Reports that say that something hasn't happened are always interesting to me, because as we know, there are known knowns; there are things we know we know. We also know there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don't know we don't know. And if one looks throughout the history of our country and other free countries, it is the latter category that tend to be the difficult ones." 
Rumsfeld's three categories generated a lot of sarcastic commentary -- especially the "unknown unknowns" -- in contrast to the applause that greeted the subsequent publication in 2007 of Nassim Nicholas Taleb's book about "Black Swans" that called attention to the same kinds of unanticipated, highly disruptive developments.  

In my opinion Rumsfeld merely described a phenomenon that not only confronts "free countries"; it confronts most individuals in most societies ... except the lucky (unlucky???) few who never encounter anything in their lives that comes out of the blue and changes everything. 

Indeed, we have long lists of names -- e.g., "gamblers" ... "thrill seekers" ... "adventurers" ... "explorers" ... "entrepreneurs" -- for people who deliberately seek new environments. Their limited knowledge of these new environments guarantees that sooner or later they will encounter powerful, unanticipated factors -- "unknown unknowns" -- that will substantially disrupt whatever they were in the process of doing. 

Black Swan #1 ... unexpected "retirement"

As readers of the first note in this series -- Pass the MOOCs ... Part 1 (New Job) -- may recall, in early 2014 I was suddenly put on permanent sabbatical after forty years at one of the nation's most prominent HBCUs.  It took about a year for me to develop clear ideas about my next career move. I decided to become a part-time consultant/analyst (20 to 25 hours per week), focusing on Black issues in higher education, especially issues related to the use of Internet technologies. In other words, I intended to produce the same kinds of reports I had developed for this blog, but use more powerful "Big Data" statistical procedures, illustrated by more powerful graphics. 

I identified the core skills I needed, noting the ones I already had, my "known knowns", and the ones that I needed to acquire, my "known unknowns". I estimated that it would take about six months for me to acquire the new skills via a series of MOOCs offered by leading universities while studying 10 to 15 hours per week.

Black Swan #2 ... a great MOOC

My first MOOC, The Analytics Edge offered by M.I.T via edX, was my second Black Swan.  I enrolled in this course in May 2015 to explore the outer boundaries of the new skills that I intended to learn. However I didn't anticipate that the course would blow my mind, that it would be one of the two or three best courses on any subject that I had ever taken in my life. 

I didn't anticipate that this course would totally upend/disrupt my new career framework by exposing me to a fascinating zoo of new ideas, unknown unknowns, that would leave me feeling that what was on the other side of my outer boundaries was far more interesting, far more exciting than the intellectual ground on which I was currently standing. Being who I am, I had no choice. I had to move into this new frontier. After adding the course's unknown unknowns to my list of known unknowns, I revised my study time. I now estimated that it would take about nine more months to a year for me to acquire the new skills I needed while studying 15 to 20 hours per week.

Black Swan #3 ... ominous forebodings

My third Black Swan emerged during the series of ten MOOCs on Data Science offered by Johns Hopkins University (JHU) via Coursera in which I am currently enrolled. Each of these courses runs four weeks on the calendar. I have just completed my seventh course, earning a C, a B, and straight A's in my last five courses. The good news is that my list of "known knowns" has grown; but the bad news is that my list of "known unknowns" has grown even faster. 

So where's my third Black Swan? I don't know, but my list of known unknowns has grown so large and grown so fast that I'm sure there has to be at least one very large unknown unknown still out there ... somewhere. I know, like the captain of the little boat in the original Jaws movie knew when his dazed passenger staggered into the captain's cabin to warn the captain that he was "going to need a bigger boat". A few seconds later when the captain himself saw the great white shark, his face did not register surprise because he already knew.

So how do I know? I know from personal experience. I attained professional competence in a few areas over the course of my long career, but I never learned how to do professional quality work by just learning new tools. As a novice I always needed apprenticeships with competent practitioners who helped me distinguish between the ideal toolsets described in the textbooks and the best practices that would enable my efforts to meet my clients' expectations.

In other words, the JHU/Coursera courses are providing excellent introductions to data science "theory"; but once I've completed these courses I will still need to engage in some kind of online apprenticeship before I launch my new career. Fortunately, Udacity's recently announced self-paced nanodegree certificate program in "machine learning engineering" seems to come reasonably close to my requirements. A few features of this program are especially appealing to me:

  • The program is designed for "engineers" not "scientists". Engineers use theories to solve problems, which is what I want to do; scientists develop and test new theories. Preparation for "science" requires prolonged, intensive study, as in PhD programs; whereas engineering skills can be acquired with substantially less investment.
  • Like all Udacity nanodegrees, this program was developed in close collaboration with leading companies in Silicon Valley. In other words it's designed to provide the specific skills sets the tech giants are currently seeking in new hires
  • Student projects receive code reviews from professionals within 24 hours.
  • Students receive hands-on career coaching upon graduation, e.g., preparing for interviews, LinkedIn profiles, etc.
  • Completion of the program within 12 months yields a 50 percent refund of tuition. 
Given the extensive preparation I will have received from the JHU/Coursera courses, I should be able to complete the Udacity program in about six to eight months, studying 20 to 25 hours per week. This will bring my total preparation time to about 18 months, i.e., three times my original six month estimate ... but only time will tell if I will complete this program on my intended schedule ... as only time will tell if my subsequent efforts will enable me to successfully launch another new career before my “final frontier” ... #OldDogStillLearningNewTricks ... :-)

P.S. Update #1 on Thursday 1/14/16

Yesterday, Udacity announced "nanodegree plus" programs. These programs are enhanced versions of its machine learning and a few other nanodegree programs.  Enhancements include guarantees of employment within six months after graduation and full tuition refunds if employment is not gained with six months. 
  • The enhancements will reduce the risks for a student to recoup his or her investment of the time and tuition required to complete the "plus" programs.
  • On the other hand Udacity will reduce its own risks by charging $299 per month for the "plus" programs, instead of the $200 per month charged for its regular nano programs; and Udacity will not refund half the tuition to students who complete the "plus" programs within six months.
  • Of course Udacity can't guarantee that a student will like the job offers that it procures ... So for students like me, for whom job satisfaction is a primary concern, enrolling in a "plus" program poses a risk of paying three times the tuition for unacceptable job offers -- assuming that the student completes the regular $200 per month program within 12 months, receives a 50 percent refund, and thereby pays only $100 per month.

P.P.S. Update #2 on Saturday 1/23/16

IMHO another appealing feature of Udacity's nanodegree programs is the fact that its programs aren't completely "open"; so calling them "MOOCs" is a misnomer, strictly speaking. Udacity rates each program as "beginner" "intermediate" or "advanced" -- labels that encourage honest introspection and self-selection by potential applicants. And some programs -- e.g, "Data Analyst" (intermediate) and "Machine Learning Engineer" (advanced) -- require applicants to pass short online exams before their applications are accepted. 

At this point, I feel like I am about to finish my "sophomore year" in a four-year undergrad program. I have a clear idea of what data science is all about, but I don't have a firm grasp of the fundamentals yet. To paraphrase Einstein, I want to progress as fast as possible, but no faster. Accordingly, I now intend to enroll in Udacity's intermediate level Data Analyst program ("junior year"), instead of its advanced Machine Learning program ("senior year") after I complete the 10-course JHU/Causera program in which I am currently enrolled.

I have added these post-scripts as my personal testimony that obtaining a first rate education via online programs requires the same kind careful planning and candid self-assessment as does obtaining a first rate education via traditional face-to-face programs ... :-)

P(3).S. Update #3 on Wednesday 2/3/16

I just passed my eighth JHU/Cousera course ("Practical Machine Learning"), earning yet another A. I am now fluent in machine learning double talk and can toss around buzzwords like cross-validation, random forests, boosting, generalized additive models, SVM, etc, etc, etc, with the most glib BS artists from both coasts and from the highest high tech centers in between. But when I look in the mirror and ask myself the definitive question: "Would you hire you???" -- the answer roars back, "Hell NO!!!" ... I'm just a smart-mouthed sophomore, a wise fool, and I know it. So when I pass my last two JHU/Coursera courses, I'm off to my "junior" year at Udacity ... :-)

P(4).S. Update #4 on Wednesday 3/16/16
Having passed my ninth JHU/Coursera course ("Developing Data Products") with yet another A, I am now enrolled in the capstone/final course of the "Data Scientist" certificate program. Lasting nine weeks, it's about twice as long as each of the previous courses. As per its name, this final course gives students a chance to pull it all together, i.e., to work on a project on which they can apply everything they've learned so far. I'm having some understandable difficulty getting started, understandable because I know that my efforts in this course will provide substantial confirmation of my previous assessment that I've only learned enough to know what the major topics are all about, especially machine learning -- but not enough to be able to charge clients for the application of my current level of understanding to their problems. So it's on to Udacity. 

I am more motivated than ever to continue my studies after I earn the JHU/Coursera certificate by this week's stunning victory of Google's AlphaGo program in a match in the game of Go with the human world champion, Lee Sedol. As per the extensive news coverage, Go is orders of magnitude more complex than chess. That's why most experts didn't anticipate such a victory until at least ten to twenty years from now. Nevertheless, using advanced machine learning algorithms, AlphaGo beat Sedol in the first three games and split the last two. Wow!!!

P(5).S. Final update #5 on Tuesday 4/26/16
This morning I received notification that I passed the tenth and final "Capstone" course in the Johns Hopkins University "Data Science" specialization certificate program with another "A" grade. This achievement earned two certificates that I immediately posted on my LinkedIn profile, one for the Capstone course and another for the entire Data Science specialization. 

I enrolled in the first course in April 2015, so it's taken me a full year to complete the required ten courses. I earned a C in the first course, a B in the second, and straight A's in the next eight. Nevertheless, as I predicted in my fourth update (above), the Capstone enabled me to pull together much of what I had learned in the previous nine courses. But it did not leave me with a firm enough grasp of the wide range of concepts and procedures covered by the previous courses at a deep enough level for me to feel confident about charging clients for my services as a consultant. 

However, the Capstone changed the metaphor I would use to describe my current understanding. In my fourth update (above) I described myself as a sophomore about to enter my junior year, i.e., halfway through an undergraduate program. Having spent the last twelve months studying about 20 hours each week for one MOOC from M.I.T. then ten from Hopkins, today I feel like a graduate student, halfway through a Masters degree program. Hopefully another six to eight months of even more intensive effort in Udacity's program will enable me to raise my flag from "half mast" to full height!!!

This is my fifth and final update for this series of notes about my views of MOOCs as a lifelong learning student. I'll start a new series when I start the courses in Udacity's program ... #OldDogStillLearningNewTricks ... :-)

Related Notes on this blog: