The NYU Courant MS-CEI Program
I was on the board of NYU for almost a decade until recently stepping down. I learned a ton about NYU during that time and one of the things I learned was that NYU’s Courant Institute of Mathematical Sciences is one of the most prestigious math schools in the world.
In addition to a world class math program, Courant houses one of NYU’s three computer science programs (the others are at the Tandon School of Engineering and the Stern School of Business), and has a top notch machine learning faculty, including Yann LeCun, who also leads Facebook’s AI Research Team in NYC.
Courant is a special place where math, machine learning, and computer science come together.
And now Courant is offering a new Masters degree in Entrepreneurship in partnership with NYU’s Stern business school. It is called the MS-CEI Program.
The MS-CEI program is a Master’s degree in Computing, Entrepreneurship and Innovation, combining computer science courses from the Courant Institute (Graduate School of Arts and Science) and business courses from the Stern School of Business…. The MS-CEI is designed for computer science students and technology professionals interested in pursuing entrepreneurship or assuming leadership roles in innovative technology based organizations.
Here is a link to the program overview and sample courses.
The initial class will enroll in the Summer of 2018 and applications are due by December 2017.
If you are interested in this program and want to learn more, please share your contact info with NYU here.
It looks like a melting pot of proficiencies, and that’s a very good formula.
Agreed. And as important as I believe CS is, I am becoming more and more convinced that really it’s math that is the most critical, and have adjusted my own emphasis with my own kids as such.And Fred wasn’t joking about the timezones!
NY heritage as the melting pot. I like that and the roots of the city.Now to cycle in the arts ala Black Mountain Style which is a natural offshoot.Black Mountain College–a visionary paradigm of learning http://arnoldwaldstein.com/…
one of the greatest art shows i have ever seen was one on black mountain college at the Hammer in LA. wow, wow, wow. that was some place
Yes indeed.This rewired my brain in the best possible way.
I am sure it is a great course but as a non “computer science students and technology professionals” MD, I am searching for great courses to bridge the gap between AI/Machine learning and other professions. That, I believe, is sorely needed. its really difficult as an MD trying to understand some of the stuff and even the fairly basic courses are quite technical.
There was a good mathematician, Leo Breiman. He was first a high end guy in academic probability theory — super nice stuff going back to A. Kolmogorov in 1933 and then H. Lebesgue near 1900. Breiman was a student of M. Loeve at Berkeley, long one of the best such probability guys on the planet, e.g., competed with J. Doob at U. Illinois, Doob’s student P. Halmos long at U. Chicago, W. Feller, long at Princeton, and Kolmogorov student E. Dynkin long at Cornell. Should include U. Grenander at Brown’s Division of Applied Math and E. Cinlar at Princeton. Sure, H. Mckean at Courant.Well, Breiman got into trying to make sense out of medical data. So, he tried to fit multidimensional planes, that is, generalizations via several variables (medicine is awash in variables) of ordinary planes. Can see some popular articles by/about Breiman.Of course, medical data is too complicated for such a fit to do well. So, Breiman fit in patches. Or he did a fit, found where his plane didn’t fit, treated that data separately, and built a tree of such separations and fits. Ah, triage!He and others wrote it up as Classification and Regression Trees, CART. The “regression” part is the fitting of the planes.Supposedly CART made some progress on some medical data, that is, got some fits and, thus, for a given patient, could make some predictions about sick or well (the “classification” part) or how sick (the “regression” part).Breiman is/was a bright guy and a very clear writer.Well, first cut, CART is the single most important foundation of current AI/ML.What Breiman did is fine, but, as I’m sure Breiman would readily agree, much more is possible.For more, read the easy “One Trick Pony” article I referenced elsewhere today. There see the stuff about recognizing digits in hand writing. So, just set up a regression model; or in an important sense liinearity is the land of milk and honey in applied math, and regression exploits that. Right from Yale and Courant, there was Dunford and Schwarz, Linear Operators, nice book I used in grad school!Well, it’s also possible to do some non-linear things, still close to linear, e.g., logistic regression, and that’s close to the AI/ML neural network curve fitting arithmetic.So, for recognizing those hand written digits, take an image with 100 x 100 pixels, regard that as regression or curve fitting with 100 x 100 = 10,000 variables; as in regression, get a lot of data; and fit. Supposedly it could do relatively well recognizing hand written digits.For so far, that’s about it: Lots of variables, thousands, maybe millions, lots of data, millions, maybe trillions, and FIT.There’s nothing intelligent about it.In practice, it may have some utility in image processing, maybe reading some medical images.But that’s just less than 1% of the huge ocean of beautifully done (if want to say that AI/ML arebeautifully done) applied math.
Great that is a melting pot. I would encourage them to add real experience to the program.
Yann LeCun. It’s interesting how people most of us have probably never heard of are doing things that could have a most profound influence over the future of the societies we live in. I hope he takes the time to contemplate the further undemocratic power he is almost certainly handing to Facebook and its founder. It definitely feels like power in the wrong hands.#stopfacebookbeforeitistoolate
* https://www.theatlantic.com…* MIT Technology Review on Deep Learning being a one-trick ponyhttps://www.technologyrevie…* Google SVP of Search says bias is the real danger:https://www.technologyrevie…@fredwilson:disqus — MIT TR notes this: “Deep learning in some ways mimics what goes on in the human brain, but only in a shallow way—which perhaps explains why its intelligence can sometimes seem so shallow. Indeed, backdrop wasn’t discovered by probing deep into the brain, decoding thought itself; it grew out of models of how animals learn by trial and error in old classical-conditioning experiments.”That’s BF Skinner’s “rat-in-the-box” experiments from 1948. It means AI folks have built machines that mimic human intelligence as if we’re rats-in-boxes.* https://www.theatlantic.com…
I can’t resist! I suspect I know who are the rats in the boxes! Hint: They write software and do numerical experiments but mostly don’t prove theorems! Ah, they push up the stock of NVIDIA.
Oh this is what happened. https://uploads.disquscdn.c…
why in the US do people say “math” and not “maths”?
Especially when one says United Kingdom and the other says United States.
the UK has only the House of Windsor, and therefore only one kingdom.there is only one Spanish state, but it is not that united 🙂
I heard the United Kingdoms are undergoing serious reconsideration as well…
that would be the Republic of Scotland…i’m not even going to contemplate the status of Northern Ireland. my brain would melt.
Because I’m in the US, say “math”, and have never been in England!!!!!
yeah, but your language has, and so that’s what i don’t get about “math”. import tax?
In school, the subject was always “math”. So, it was all “math” to me! I heard “maths” for the first time only in the last few years but tried not to sweat the small stuff!
must be a TSA thing.
Hmm, yes, Courant has long been one of the best math shops in the world.Looking at some of the courses for the MS-CEI Program, I don’t see much from the Courant quality math!Instead the program emphasizes computing, computer science, and business. I touch on each of these three below:ComputingThe computing stuff looks like some overview rehashing of some of what is common and popular in computing now. I taught courses like that. Maybe some of the students got a start good enough to keep them going.I’m not impressed. My guess is that students will be mostly wasting their time.Here’s why:For computing, there’s:(1) Routine stuff common in current practical computing. In the US, that’s commonly self-taught, maybe with a start in some high school, community college, trade school, ugrad courses.(2) More advanced stuff in current practice. Here I would point to:(A) The standard stuff in algorithms, data structures, and computational complexity, e.g., Knuth, Ullman, Sedgewick.That stuff is nearly universal in ugrad CS programs.(B) Designing and running the server farms at Citibank, FedEx, Google, Facebook, Microsoft, and Amazon. Similarly for the networks. Also consider the power and HVAC. Or, how do they do that?Of course, roughly the answer is, “As fast and best they can.”, but there are no doubt some lessons to be learned. And if want to set up such a server farm, then likely need to know the lessons.Sadly, it’s not easy for an academic shop to know this stuff. In few years, a lot of that stuff changes significantly. The people who do know the stuff, don’t much write, document, or teach it.(C) How the heck to manage the software and other development work — high level design, lower level designs, principles for reliability, documentation, testing, and revision, coding and documentation techniques and standards.Again, the people who actually do this stuff don’t much teach it, and I’m reluctant to believe that an academic shop will know much about it.(3) New stuff, that is, new in some or all of problems being solved, products and services, markets, data, architectures, UI/IX, algorithms, secret sauce, etc.That’s, call it, applied research. Well, that’s a special case of, uh, sorry ’bout that, research.It’s not easy to do or teach research.My view is that the most important prerequisite for good applied research in the future of computing is math. Some of the courses do touch on, that is, try to make use of, math.IMHO, sadly, from those courses, the students will not be well served. Why? Over and over, the more promising courses keep talking about, right, linear algebra, “vector calculus,” optimization, probability, and statistics. But these topics are not taught in the program. And the program does not assume or require that the students have this material. Also, the students should have some abstract algebra. And “vector calculus” is not enough or even very close to the right direction. The long standard is W. Rudin, Principles of Mathematical Analysis, a.k.a. Baby Rudin, i.e., in strong contrast to several other much more advanced books from Rudin.Here’s the truth: The students need to roll back to ugrad school and get a good ugrad pure math major. That’s the minimum. Then they should get a carefully selected Master’s in pure/applied math. Then they should get a Ph.D. in applied math, e.g., learn how to do some research.Then they could take the courses in the Courant program that need some math. And then the students will commonly find that the profs fumble terribly with the math, the other students are nearly lost, and neither the students nor the profs have much hope of doing much in research using math.By analogy, an auto mechanic needs a good set of socket wrenches, and he won’t make his own.For applied research in computer science, it’s narrow. E.g., might look atJames Somers, “Is AI Riding a One-Trick Pony?”, MIT Technology Review, September 29, 2017.as athttps://www.technologyrevie…Right, “narrow” as in “one trick pony”.Uh, much more is possible, guys, uh, lots of tricks, dozens of them, thousands, ….Yup, a core problem that explains this article is that (i) the main key to good applied research in computer science is math and (ii) the computer science profs didn’t ever learn at all well the material in a good ugrad math major. Sorry ’bout that.Uh, once at a high end research university, there was a grad program that had a Ph.D. qualifying exam in mathematical analysis. But, they didn’t teach such a course and assumed that, well, somehow their entering students would know it, learn it, absorb it, get it by sleeping with a suitable text under their pillow, etc.Well, nearly all the students did poorly on that exam except one. That one had, well before going to the grad school, gone through Baby Rudin, all three editions, carefully, plus a stack of about a dozen closely related and similarly respected texts. There’s no royal road, guys.Or, how to get a silly requirement for a brain-dead course in economics set aside: After the first lecture, alone with the prof, ask him seriously just what he was assuming about his free hand supply and demand curves: Lebesgue measurable, continuous, differentiable, continuously differentiable, infinitely differentiable, Lipschitz, upper/lower semi-continuous, convex, pseudo-convex, quasi-convex, or what? Presto, bingo, no longer need the silly econ course!Why? Those are all legitimate questions about assumptions about curves, especially about curves where we are about to consider estimation and optimization, and the econ prof didn’t know enough math to respond and didn’t want to be embarrassed in class.It’s an old story and standard: In nearly all fields, especially in the STEM fields, the best research mathematizes the field, and profs in all those fields struggle terribly for all their careers with too little background in math.Computer ScienceBroadly, computer science is a dead field, is stuck-o on some old problems it can’t solve, has too little in good connections with practice to pick good, new problems, and has too little background, especially in math, for how to get new solutions and push the field ahead.In part a solution is for computer science to be more clinical, like medicine, that is, be close to practice and take a lot of motivation from practice.E.g., go to some of the leading research/teaching hospitals and see how they connect research with practice. Do the same for some of the US national security labs.BusinessGee, here are some first stops on business!Read a book on organizational behavior!Go to a tax accountant and get a tax accounting tutorial 101.Read a book on business law 101.Read the stuff at AVC.com “MBA Mondays” and from JLM. Maybe watch the YCombinator videos. Read the Paul Graham essays.Call it DONE!